Intel to bring forward Conroe release date

[mearle73]

http://www.theregister.co.uk/2005/12/21/intel_reschedules_conroe/

[saaya]

:rocker:

[hicookie]

Pentium M+new netburst? expect!

[saaya]

no, it has nothing to do with netburst.

its a yonah on steroids :D

[Cooper]

Pentium M+new netburst? expect!

I totaly agree :)

[Carfax]

Nah, this isn't even Yonah on steroids. It's something better and different.

4 wide issue core, enhanced prefetch, memory disambiguation, macro-ops fusion, faster, and wider caches, and best of all, it should be VERY overclockable :D

The only thing I don't like about Conroe/Merom is that it still uses the same old bus architecture. Intel needs to get with the times, and quickly!

Especially if the FSB is only 1066 at launch, that will be a shame. It should be atleast 1333, or even 1600.

Oh well, if I'm very impressed with the performance and I buy it, I doubt I'll keep it at stock FSB :D

[grimREEFER]

hmm, any chance of it working on 945p?

[nfm]

I just wanna know one thing, how many pipelines? If less than 15 pipes then Intel will be my choice for 2006

[Nasgul]

no, it has nothing to do with netburst.
its a yonah on steroids


Conroe takes key NetBurst features and integrates them into a new core developed on the back of what the company learned when it created today's Pentium M chips.

Why is it that everyone cannot face that fact?

Intel has been saying that since August 24, 2005 (http://216.239.37.104/translate_c?hl=de&ie=UTF-8&oe=UTF-8&langpair=de%7Cen&u=http://www.computerbase.de/news/hardware/prozessoren/intel/2005/august/idf_intel_details_architektur/&prev=/language_tools), and they even show you a pictureS:


http://i6.photobucket.com/albums/y218/Nasgul/7.jpg


Long live NETBURST!

PPPPSHHHHH!!!!!!!!! Carfax, check this out: Conroe @ 1333fsb (http://forums.extremeoverclocking.com/showpost.php?p=2030374&postcount=12) on DDR3 1333 (http://www.geek.com/news/geeknews/2005Aug/bch20050826032040.htm)

[savantu]

I just wanna know one thing, how many pipelines? If less than 15 pipes then Intel will be my choice for 2006

14.

[Carfax]

PPPPSHHHHH!!!!!!!!! Carfax, check this out: Conroe @ 1333fsb (http://forums.extremeoverclocking.com/showpost.php?p=2030374&postcount=12) on DDR3 1333 (http://www.geek.com/news/geeknews/2005Aug/bch20050826032040.htm)

Well thats good to know man. Although, DDR3 will not be available when Conroe is introduced next year. In 2007, DDR3 should be out and Intel should be using CSI, rather than FSB.

Bloody hell! I just checked The Register and CSI won't be available till 2008 :(

[Donnie27]

Well thats good to know man. Although, DDR3 will not be available when Conroe is introduced next year. In 2007, DDR3 should be out and Intel should be using CSI, rather than FSB.

Bloody hell! I just checked The Register and CSI won't be available till 2008 :(

DDR3 will launch when Conroe does. That's why Intel invested almost a one Billion dollars in Hynix, Elpida, Samsung and Micron. They did the same thing to help bring DDR2 to the market.

http://cc.msnscache.com/cache.aspx?q=2762280301955&lang=en-US&mkt=en-US&FORM=CVRE

http://www.elpida.com/en/news/2005/11-14.html

http://www.vaata.com/?news=250

Add to this Intel is saying Conroe is smoother than normal rumors are June-06 for Conroe to hit the market, DDR3 is no big deal at all. Why manny speculate AMD skipping DDR2 and going straight to DDR3. But also, AMD using DDR2 has built-in higher sales as DDR2 becomes obsolete very quickly.

Donnie27

[Donnie27]

Why is it that everyone cannot face that fact? Intel has been saying that since........

Yup, but you know nothing is hated more by many folks than Netburst LOL! Maybe folks don't want to be jumpped by AMD ______s you fill in the blank for pointing out that Netburst is NOWHERE nearly as bad as they profess. It's why I don't post much here anymore. Conroe will have just as much Netburst as Dothan/Banias, hell Intel has always said that.

Dothan already proves Integrated Memory Controller is not a must, sheesh! That also makes an already expensive Processor even more exspensive.

Donnie27

[saaya]

4 wide issue core, enhanced prefetch, memory disambiguation, macro-ops fusion, faster, and wider cachesthats what i mean with a yonah on steroids... 3 way to 4way issue core, enhanced preftetch, faster and wider caches and some added extras, thats not a new core, thats an improved core, hence its more a yonah on steroids if you ask me.

and nasgul, thats a marketing wallpaper your pointing out there, did you even read what intel mentiones as the netburst tech they integrated into the new core design? quad pumped bus and 64bit tech... if thats netburst than you could also claim that dothan and yonah are netburst as well because dothan and yonah both use the quad pumped interface and yonah even supports 64bit wide memory addressing...

whats CSI?

[saaya]

Conroe will have just as much Netburst as Dothan/Banias, hell Intel has always said that.exactly :)
they have nothing to do with netburst at all... :D


Dothan already proves Integrated Memory Controller is not a must, sheesh! That also makes an already expensive Processor even more exspensive.hows the imc making it more expensive?



Add to this Intel is saying Conroe is smoother than normal rumors are June-06 for Conroe to hit the market, DDR3 is no big deal at all.
what?

Why manny speculate AMD skipping DDR2 and going straight to DDR3. But also, AMD using DDR2 has built-in higher sales as DDR2 becomes obsolete very quickly. built in higher sales?
your post doesnt make any sense, can you please explain what your trying to say?

[Carfax]

thats what i mean with a yonah on steroids... 3 way to 4way issue core, enhanced preftetch, faster and wider caches and some added extras, thats not a new core, thats an improved core, hence its more a yonah on steroids if you ask me.

Well, do you think the K8 is a K7 on steroids, or is it a new core? The K8 had substantial improvements over the K7, and although many similarities exist between them, I wouldn't classify the K8 as a K7 on steroids, because the differences are too striking. Things like the IMC, SSE2, longer pipeline, twice the number of GPR etc..

Conroe has even more differences between it and Yonah, than the K8 compared to the K7.

The K8 is a 3 issue core, just like the K7. Conroe is 4 issue however, and Yonah is 3 so it's wider.

Then, Conroe has 64-bit capability, whilst Yonah doesn't. All of these things require a ton of architectural changes, more than just a makeover, but since I am no engineer, I cannot go that far indepth as to specific details :D

BTW, CSI stands for "Common serial interconnect" and it's basically Intel's version of Hypertransport.

It's supposed to be much better than the current HT that AMD employs, but by the time CSI is due out, AMD will also have something much better..

[Nasgul]

Conroe will have just as much Netburst as Dothan/Banias, hell Intel has always said that.
Of course!

Netburst Rules!

Kinda sad that people still thinks Netburst = Socket 478 Prescott.


Dothan already proves Integrated Memory Controller is not a must, sheesh! That also makes an already expensive Processor even more exspensive.
Don't forget the useless 3DNow Pro!!!, the on-goning lame lawsuit, the company's CEO which middle name is "Jesus".
Their fab in a Foreign country.
The company they always run to for help (IBM) because they have no clue of what to do next and have to pay them massive ammounts of money for their technology because they don't have own of their own.
And so on.....

[grimREEFER]

i understand that it uses socket 775, but will 945/955/975 boards work with it? i dont wanna part with my ddr2 just yet....

[Donnie27]

exactly :)
they have nothing to do with netburst at all... :D

hows the imc making it more expensive?
what?
built in higher sales?
your post doesnt make any sense, can you please explain what your trying to say?

Exactly!

Donnie27

[Donnie27]

Of course!

Netburst Rules!

Kinda sad that people still thinks Netburst = Socket 478 Prescott.

Don't forget the useless 3DNow Pro!!!, the on-goning lame lawsuit, the company's CEO which middle name is "Jesus".
Their fab in a Foreign country.
The company they always run to for help (IBM) because they have no clue of what to do next and have to pay them massive ammounts of money for their technology because they don't have own of their own.
And so on.....

It's as simple as this. An Integrated Memory Controller is nice. But the more stuff you pack on die, the chance or odds something will go wrong increases. This decreases yields and make the few than can pass bins, more expensive per unit. Just maybe Intel knew of this when they killed Timna and haven't went back to it.

Geesh dewd! I love it when folks on message boards know more than Intel themselves, WOW! Here's one for ya' maybe Intel is lying and just making a Pentium-M on Steroids like these guys are saying?

Donnie27

[Donnie27]

Originally Posted by Donnie27
Why manny speculate AMD skipping DDR2 and going straight to DDR3. But also, AMD using DDR2 has built-in higher sales as DDR2 becomes obsolete very quickly.

For saaya!

Each memory change means AMD will have to sell a different/new processor since the IMC will not be compatible with DDR2 and DDR3. Get it? Higher sales numbers!

Donnie27

[zakelwe]

Conroe is going to be the real deal in 2006 I think. Assuming it is much more overclockable than the mobile variant (Memrom?) due to it not having as many power saving features I think we could find it more than AMD can handle.

More and more I seem to think I will be weaving back to Intel come next summer :)

Regards

Andy

[NiCKE^]

I wonder if I should go for a Cedar Mill or wait for the Conroe, the Conroe sounds like a killer indeed if you believe what you hear about 'em.

[mearle73]

I wonder if I should go for a Cedar Mill or wait for the Conroe, the Conroe sounds like a killer indeed if you believe what you hear about 'em.

Think Im going to wait for Conroe,its not going to be that long a wait after the release dates for Cedar Mill.Then I will be able to find out what revision of the 975x boards are going to support it

[NiCKE^]

Probably those which is out now will support Conroe.

[Shnorque]

Intel is to align the launch of its next-generation desktop processor, 'Conroe', and its upcoming 965 family of chipsets, delaying the latter and bringing forward the former.

wasnt 965 due out with 975? if so does this mean 975 isnt due till Q3 instead of Q1?

[saaya]

Conroe has even more differences between it and Yonah, than the K8 compared to the K7.

The K8 is a 3 issue core, just like the K7. Conroe is 4 issue however, and Yonah is 3 so it's wider.

Then, Conroe has 64-bit capability, whilst Yonah doesn't. All of these things require a ton of architectural changes, more than just a makeover, but since I am no engineer, I cannot go that far indepth as to specific details :D
adding 64bit support is not as big of a deal as far as i know.
intel added 64bit to prescott in a hurry, it only took them a few months.
and making a 3 issue core a 4 issue core isnt that big of a deal either. would you say r420 is a new vpu design? its a r3xx on steroids imo.
r3xx had 2 quads, r420 had 4... very similar if you ask me, and i wouldnt call it a new design, neither would i call conroe a new design.
yonah is MUCH more of a new design than conroe is.
the changes from dothan top yonah are signinificant, from yonah to conroe its the same basic architecture with tweaks.
and yes, if the a64 wouldnt have an imc id say its an athlonxp on steroids.
but the imc and crossbar switch with the system request manager is a totally new component and a significant new component.

whats new in conroe? 64 bit, thats just widening the registeres, its nothing really new. making it process 4 instead of 3 instructions per clockcylce isnt a new design, its just adding more of the already existing logic and tweak the setup. so whats new in conroe?


BTW, CSI stands for "Common serial interconnect" and it's basically Intel's version of Hypertransport.

It's supposed to be much better than the current HT that AMD employs, but by the time CSI is due out, AMD will also have something much better..its supposed to be much better? well for something comming in 2 years beeing better than an already 2 year old technology doesnt really sound promising.

thats like saying in 2 years im going to build a rig that performs faster than a an athlon xp 3200+ with a 9800pro!
does this sound like a kick4ss system to you? ^^



Conroe will have just as much Netburst as Dothan/Banias, hell Intel has always said that.
what do you consider netburst then?
please give me your definition.



Netburst Rules!what is netburst?

Don't forget the useless 3DNow Pro!!!, the on-goning lame lawsuit, the company's CEO which middle name is "Jesus".
Their fab in a Foreign country.
The company they always run to for help (IBM) because they have no clue of what to do next and have to pay them massive ammounts of money for their technology because they don't have own of their own.
And so on.....and this has what to do with conroe?
dont make this an amd vs intel thread or im going to send you to a nice christmas vacation. you keep posting how you love intel and amd, good for you! but can you please keep this to yourself instead of pushing thread after thread into offtopic amd vs intel discussions? thank you


i understand that it uses socket 775, but will 945/955/975 boards work with it? i dont wanna part with my ddr2 just yet....from a technology point of view, it should work with any chipset that is dual core certified by intel... my guess would be no though, because it looks like its comming with high fsb speeds, so only the top end boards currently available should support them at full speeds.


It's as simple as this. An Integrated Memory Controller is nice. But the more stuff you pack on die, the chance or odds something will go wrong increases. This decreases yields and make the few than can pass bins, more expensive per unit. Just maybe Intel knew of this when they killed Timna and haven't went back to it
hmmm yeah, thats true, but the imc is really tiny on the a64, so i doubt it has a big impact on yields. on amds dual core chips the imc only takes 2% of the die space afaik... so i doubt this has a big impact on yields.


For saaya!

Each memory change means AMD will have to sell a different/new processor since the IMC will not be compatible with DDR2 and DDR3. Get it? Higher sales numbers!

Donnie27
hmmm thats true... but it doesnt necessarily mean higher sales. just think of how long amd stuck to ddr1. if they would have wanted to push sales they could have gone ddr2 a long time ago and have forced people to buy new cpus and boards.
and amd is trying to get ddr3 backwards compatible so it can be used in ddr2 systems and with ddr2 imcs and cpus from amd.
and uess whos trying to change the pinout of the ddr3 sticks just to make them not compatible? intel...


Probably those which is out now will support Conroe.dont think so... maybe some high end boards, but i dont think 945 chipset based boards will support it... but lets see...


wasnt 965 due out with 975? if so does this mean 975 isnt due till Q3 instead of Q1?
xbit said intel will delay 965 so it can launch it together with conroe, sounds like a hint to me that conroe wont work with current boards, at least not at full fsb speeds.

[Donnie27]

and uess whos trying to change the pinout of the ddr3 sticks just to make them not compatible? intel...

Why not?
Who in the hell spent 1 Billion for these folks to develope DDR3?
How much did AMD, the EU and the German Government spend or invest in DDR2 or DDR3?
Ohhhh but AMD should get a free and clear ride uh?

Maybe, just maybe if AMD, the German Government or the EU should kept Micron, Hynix and others out of the red they'd have a say. Then someone whould listen to them when they try to "get ddr3 backwards compatible" someone *might listen. I don't agree with what Intel did with DDR3 but I'm also not sure if would be backward compatible anyway.


if they would have wanted to push sales they could have gone ddr2 a long time ago and have forced people to buy new cpus and boards.

So you think M2 and etc.. will not support DDR2? If it does, then AMD is doing just what you say they aren't doing. They will go through the different RAM speeds just as they did with PC-1600 through PC-3200. Yes, folks bought all of those different Chipsets as well. Why would you think they wouldn't now? Is this revisionist history? No, it doesn't excuse the BS Intel has pulled. It sucks when THEY (Intel and AMD) do this.

Donnie27

[Donnie27]

hmmm yeah, thats true, but the imc is really tiny on the a64, so i doubt it has a big impact on yields. on amds dual core chips the imc only takes 2% of the die space afaik... so i doubt this has a big impact on yields.

Oh, it is not the size but if it works or not, doesn't matter if it only took up .01%. It is a value added feature.

Donnie27

[Shnorque]

xbit said intel will delay 965 so it can launch it together with conroe, sounds like a hint to me that conroe wont work with current boards, at least not at full fsb speeds.

this sounds logical. although i really dont think intel will launch these procs with higher than 1066fsb? maybe 1333fsb on the extreme edition chips? - but a P5WD2 would run at 333fsb wouldnt it?

i expected intel to move to a 1066fsb on all the chips along time ago? after the release on the extreme edition northwood?

i dont see that intel pushing a new pin count on the DDR3 DIMMs is a big problem - you would need a new mobo to support the speeds anyway, so i dont see a big difference buying a new mobo to support the DIMM configuration. so long as intel dont make us pay royalties like they did with RAMBUS it wont worry me.

[Victorshen]

this sounds logical. although i really dont think intel will launch these procs with higher than 1066fsb? maybe 1333fsb on the extreme edition chips? - but a P5WD2 would run at 333fsb wouldnt it?

i expected intel to move to a 1066fsb on all the chips along time ago? after the release on the extreme edition northwood?

i dont see that intel pushing a new pin count on the DDR3 DIMMs is a big problem - you would need a new mobo to support the speeds anyway, so i dont see a big difference buying a new mobo to support the DIMM configuration. so long as intel dont make us pay royalties like they did with RAMBUS it wont worry me.
Conroe is officially 1333 fsb

[saaya]

Why not?
Who in the hell spent 1 Billion for these folks to develope DDR3?
How much did AMD, the EU and the German Government spend or invest in DDR2 or DDR3?
Ohhhh but AMD should get a free and clear ride uh?hahahah lol, what the h3ll does the german government and the eu have to do with this? :rofl:

but good point, your right...
i think its kinda lame to change the pinout just so ddr3 cant be used in ddr2 boards... intel customers would benefit from this as well as ddr2 and 3 can be compatible if you build the boards right afaik... but your right, intel paid a big part of the development cost so they can decide how the interface should look.


Maybe, just maybe if AMD, the German Government or the EU should kept Micron, Hynix and others out of the red they'd have a say. Then someone whould listen to them when they try to "get ddr3 backwards compatible" someone *might listen. I don't agree with what Intel did with DDR3 but I'm also not sure if would be backward compatible anyway. nono, you got that wrong, it IS backwards compatible with ddr2 afaik, its just about the pins on the ddr3 sticks, intel wants it to be incompatible with ddr2, maybe so people have to buy new memory, new boards, new amd cpus.
thats afaik at least.

but again donnie, what does the eu or germany have to do with this?
i dont get it... and they should have kept hynix and micron out of the red?
what red? please explain what you mean.



So you think M2 and etc.. will not support DDR2? If it does, then AMD is doing just what you say they aren't doing. They will go through the different RAM speeds just as they did with PC-1600 through PC-3200. Yes, folks bought all of those different Chipsets as well. Why would you think they wouldn't now? Is this revisionist history? No, it doesn't excuse the BS Intel has pulled. It sucks when THEY (Intel and AMD) do this.

Donnie27of course amds new socket and chips will be ddr2, the name has been changed btw afaik, but how can you blame amd for that move? the alternative would be to keep using ddr1, and that would hold them back performence wise, and will be a problem because afaik dram makers are shutting down their ddr1 production fabs and are moving to ddr2 and some to ddr3 already. so they HAVE to move to ddr2. and if there was a way to move to ddr2 without forcing people to a new plattform i bet theyd have done it, but ddr2 has a different pin number and layout and theres no way to fit them into a 939 board.
well actually on via chipset based 939 boards you can run ddr2 with todays a64 chips, you just need an adapter so the memory fits into the slots.
again, if ddr2 sticks wouldnt have a different pin number and layout they could work in current mainboards and all you would need was a new cpu, and not a new mainboard as well.

but its not amd who decided what layout ddr2 and ddr3 have, so dont blame them for it. its not them who are forcing people to upgrade, its the dramurai+intel


Oh, it is not the size but if it works or not, doesn't matter if it only took up .01%. It is a value added feature.

Donnie27true, BUT since its a small size its very unlikely an error will occure in this particlar space of the die, and even if theres an error, the imc is 146 bits wide or something like that, and i think amd can disable a part of the imc and still have a 128bit imc functional chip.


this sounds logical. although i really dont think intel will launch these procs with higher than 1066fsb? maybe 1333fsb on the extreme edition chips? - but a P5WD2 would run at 333fsb wouldnt it?
it wont support a higher fsb that 975, that wouldnt make any sense imo.
and whats the max speed 975 supports officially? 1066fsb right? maybe it does 1333... but not officially... so id say normal conroe chips will be 1066 or even only 800fsb. most likely 1066 though.


i expected intel to move to a 1066fsb on all the chips along time ago? after the release on the extreme edition northwood?
yeah, i never got that either... their boards can take it, their chipsets can take it, their cpus can take it, the mem can take it... so what are they waiting for? maybe higher fsb speeds would add 10 or 20W more heat, and for prescott and the pentium d that was a nono?


i dont see that intel pushing a new pin count on the DDR3 DIMMs is a big problem - you would need a new mobo to support the speeds anyway, so i dont see a big difference buying a new mobo to support the DIMM configuration. so long as intel dont make us pay royalties like they did with RAMBUS it wont worry me.well not really, you could run ddr3 on a current ddr2 board and run it with a memory divider.
the divider will mean it wont perform as well as 1:1, but ddr3 has the same low latency as ddr1 and even higher bandwidth as ddr2, its really the best of both ddr2 and ddr1 plus x :D


Conroe is officially 1333 fsb
conroe extreme edition afaik...

[Nasgul]

what do you consider netburst then?If you don't know, then why are you in the INTEL side of the forum?



what is netburst?Google it!



and this has what to do with conroe?
dont make this an amd vs intel thread or im going to send you to a nice christmas vacation. you keep posting how you love intel and amd, good for you! but can you please keep this to yourself instead of pushing thread after thread into offtopic amd vs intel discussions? thank youI was just fallowing up with Donnie27, that's all. I don't see why anyone else should be so touchy about that. After all I quoted him, in reference to him alone. And tell me which "THREAD after THREAD" I'm pushing off topic? I'm dying to see how many and when have I said I luv AMD? :confused:





For saaya!

Each memory change means AMD will have to sell a different/new processor since the IMC will not be compatible with DDR2 and DDR3. Get it? Higher sales numbers!

Donnie27hmmm thats true... but it doesnt necessarily mean higher sales. just think of how long amd stuck to ddr1. if they would have wanted to push sales they could have gone ddr2 a long time ago and have forced people to buy new cpus and boards.
and amd is trying to get ddr3 backwards compatible so it can be used in ddr2 systems and with ddr2 imcs and cpus from amd.
and uess whos trying to change the pinout of the ddr3 sticks just to make them not compatible? intel...

Now, what does that ^^^^ have to do w/ Conroe, to me it sound like an Intel vs AMD.......

And I guess it hasn't stopped there> hasn't it? (http://www.xtremesystems.org/forums/showpost.php?p=1198983&postcount=32)

[Shnorque]

well not really, you could run ddr3 on a current ddr2 board and run it with a memory divider.
the divider will mean it wont perform as well as 1:1, but ddr3 has the same low latency as ddr1 and even higher bandwidth as ddr2, its really the best of both ddr2 and ddr1 plus x :D

yes, i never considered that? although at the moment the intel chipsets only support DDR800? so this is similar to my point about running a 975 board at 333fsb, possible but not officially supported...


dramurai+intel

dramurai?


it wont support a higher fsb that 975, that wouldnt make any sense imo.
and whats the max speed 975 supports officially? 1066fsb right? maybe it does 1333... but not officially... so id say normal conroe chips will be 1066 or even only 800fsb. most likely 1066 though.

as you say i think it will be 1066, there is alot of hype about conroe on 1333fsb, but i agree with you and think it will only be the xtreme procs.

although i believe that intel will release a new version of 975 chipset - 975xe perhaps? that will support 1333fsb and DDR3 ram. they have made similar changes to chipsets like this in the past i think?

[grimREEFER]

will hyperthreading be standard on all conroe models(im assuming they are all going to be dual core)?
cause if it is, im gonna buy a low end conroe, overclock it, and run cinebench and PWN lol....if it works on my board in the first place lol, probly not if it has such a high fsb...

[Donnie27]

hahahah lol, what the h3ll does the german government and the eu have to do with this?

Now it's my turn to :rofl:
The Germans, your tax dollars, Paid for Fab 30, it cost 2.5 billions.
The Germans, your tax dollars, Paid for Fab 36, 1.45 Billion with another 600 million coming from the EU, still feel like laughing?:p:

http://www.semiconductor-technology.com/projects/amd/

AMD Saxony represents one of the largest US projects in East Germany, with total investment expected to reach $2.5 billion by 2003. Before re-unification Dresden was part of East Germany, and unemployment is high.

http://news.cnet.co.uk/desktops/0,39029662,39193194,00.htm

Government loans and subsidies come to about $1.5bn, according to AMD.

http://techreport.com/onearticle.x/6241

I would like to thank the taxpayers of Germany, Saxony, and the EU for subsidizing cheap chips for us all.

I could give at least 60 more links if you like. These are Grants and Loans, some call it porkbarrel.

Higher sales figures when you get to sell the same processors to the same folks. Just as folks did and have done since Intel and AMD started making processors. The difference between Intel and AMD is that AMD is a Parts supplier and Intel is a whole systems supplier. No, I don't think it is fair when Intel has to compete against the Germans and the EU, that's BS! What if the U.S. did the same for Intel? Still laughing, hehehe?

Donnie27

[Shnorque]

Now it's my turn to :rofl:
The Germans, your tax dollars, Paid for Fab 30, it cost 2.5 billions.
The Germans, your tax dollars, Paid for Fab 36, 1.45 Billion with another 600 million coming from the EU, still feel like laughing?:p:

So what you were originally saying is that if the EU or the Germans gave grants to hynix et al for the development of DDR2/3 then AMD should have a say in its pinout etc?

[cirthix]

egad, this thread turned ugly fast

[Donnie27]

So what you were originally saying is that if the EU or the Germans gave grants to hynix et al for the development of DDR2/3 then AMD should have a say in its pinout etc?

YES! They are expecting ROI afterall. That'd give them (and AMD) the Right not only ask, but demand compatability=P These folks are lieing back, then waiting for Intel and other companies to R&D the tech then *Freeload. Conroe will launch with DDR2 then move to DDR3.

Be warned, Intel has pulled some good April Fool's launches.

Donnie27

[Donnie27]

egad, this thread turned ugly fast

No, we don't have ugly threads when folks come with facts instead of feelings for Company-A or Company-B. Chevy doesn't help Ford, Coke doesn't help Pepsi, and FedX doesn't help UPS. Why in Blue Blazes should Intel be expected to help AMD? Both companies have done great things and both have had taken turns at screwing the consumers. But Fans don't like to see or hear the true suckyness of their favorite. Then the flames start. That's whacky!

Intel said Conroe would "use the best tech from Pentium 4 and Pentium M". Intel showed slides of as Nasgul linked to. So I don't know why anyone would say otherwise and that's what started all of this. Conroe's launch has been moved up. i965 launch has been moved back to launch with it.

Donnie27

[IluvIntel]

although i believe that intel will release a new version of 975 chipset - 975xe perhaps? that will support 1333fsb and DDR3 ram. they have made similar changes to chipsets like this in the past i think?

A similar change happened with 925X and 925XE chipsets, about 5 months apart, 925X supported 800mhz fsb and 925XE - 1066mhz fsb.

[Sentential]

I kinda find alot of this to be humerous. What gives you the idea that Conroe will be DDR3? In addition who told you that AM2 will by DDR3? ... lol :D

I dont want to sound rude or anything but you guys really dont have a clue. I can tell all of you right now that AM2 will be DDR2, so will Conroe. I do know that AM2 is 100% going to use DDR2 this "skipping" scenerio is simply not going to happen period.

Perhaps when S1207 comes out they might change it to DDR3 but its not going to happen anytime soon.

DDR3 while socket compatible has a different "bios" to the PCB itself (Brian would know more, thats just where my info goes fuzzy). You cant simply plug DDR3 into a DDR2 based box. The whole layout via the memory controller would have to be significantally changed to use it.

Not only is that going to be an issue but Elpeda hasnt shown any interest to us that DDR3 is in the immediate future. DDR2 has atleast another year, probably 2 before we will make that transition since that is what all the road maps say.

I applaud the effort but DDR3 just isnt going to happen on desktop platforms for quite some time. K9 wont use it, perhaps K10 / P5 will, but it certianly wont be until atleast mid 2007 before it will start to be phased to DDR3.

As for commentary on Conroe Ive got mixed feelings on it. I feel a bit disspointed that they went with such a lax cache layout, but we will see how high they scale. As for the rumor of a 333 bus im very skeptical. They are too far into the design phase to make a significant change like that. Most likely it will be only 266 if that.

AMD on the other hand Ive grown more impressed the more I know about AM2. Everyone is in for a real treat when it comes this summer. I have high hopes that we may see 600 1:1 for the first time ever. The next-gen of DDR2 will certianly make that possible as its starting to appear already.

[Shnorque]

YES! They are expecting ROI afterall. That'd give them (and AMD) the Right not only ask, but demand compatability

OK, I understood your original refferance to the EU and Germany, but wasnt sure if it had anything to do with RAM...


I can tell all of you right now that AM2 will be DDR2, so will Conroe. I do know that AM2 is 100% going to use DDR2 this "skipping" scenerio is simply not going to happen period.

Perhaps when S1207 comes out they might change it to DDR3 but its not going to happen anytime soon.

Last I checked Soceket M2 and 1207 where the same thing? I am sure the pincount on M2 is around that?

Also, Sentinal, Im not sure if you realise but in reality conroe should be P5 - I cant see that Intel could make such significant changes to the arch and still call them P4s. So by your own logic it is possible it could use DDR3 - P5 is really only 1 generation away whereas K10 is two?
I wont venture into RAM discussions anymore, as I really know little about them, not to mention how far off topic it is.

[Shnorque]

A similar change happened with 925X and 925XE chipsets, about 5 months apart, 925X supported 800mhz fsb and 925XE - 1066mhz fsb.

this is what i base my theory on. didnt 925xe also support DC, whereas 925x didnt?

[Who]

Last I checked Soceket M2 and 1207 where the same thing? I am sure the pincount on M2 is around that?

Nope, the last you checked you checked incorrectly. AM2 has 940 pins. Socket F, the new server/workstation socket, will have 1207.

[vapb400]

this is what i base my theory on. didnt 925xe also support DC, whereas 925x didnt?
nope, just fsb bump

[Sentential]

Last I checked Soceket M2 and 1207 where the same thing? I am sure the pincount on M2 is around that?
Nah. AM2 is still S940 last I heard. S1207 comes later I think.


Also, Sentinal, Im not sure if you realise but in reality conroe should be P5
No it really isnt. CSI is what was supposed to make Conroe P5. Now instead they scrapped that idea for a multi-hub based NB system for later next year. That is P5, this is just a refresh.


- I cant see that Intel could make such significant changes to the arch and still call them P4s. So by your own logic it is possible it could use DDR3 - P5 is really only 1 generation away whereas K10 is two?
I wont venture into RAM discussions anymore, as I really know little about them, not to mention how far off topic it is.
I believ K9 was effectively scrapped and merged into S939 (X2 was supposed to be K9), so they are both one generation away.

[Donnie27]

Nah. AM2 is still S940 last I heard. S1207 comes later I think.

No it really isnt. CSI is what was supposed to make Conroe P5. Now instead they scrapped that idea for a multi-hub based NB system for later next year. That is P5, this is just a refresh.

I believ K9 was effectively scrapped and merged into S939 (X2 was supposed to be K9), so they are both one generation away.


Conroe will launch with DDR2 then move to DDR3.

AMD went against calling or naming any of their Processors K9. They didn't like the references to "Dog" K9=Dog or etc... But you're right, from what I saw, X2 was supposed to be the K9 if they would have used that name.

Socket 940 and M2-940 are not compatible with eachother. Just as 940 and 939 aren't or 603/604. 1207 is for higher end stuff anyway.

The reasons for the heated post ( Flames) and hurtful tone of some of the posts? The seemly endless Double standards that are excepted by AMD fans/majority (of the posters), on the Intel side of the forums is one reason. When Intel does this, they are scumbags and lowest of the low lifes. When AMD does the very same thing, it's accepted and defended. They both might have had a very good reason, but you'd never know it. You'd still get the same old tired AMD is the white hat wearing good guy and Intel is more hated than many of the virus writers.

Donnie27

[Donnie27]

OK, I understood your original refferance to the EU and Germany, but wasnt sure if it had anything to do with RAM...

Last I checked Soceket M2 and 1207 where the same thing? I am sure the pincount on M2 is around that?

No biggie, certain things are not for the 75% of posters who like, follow and even worship AMD. Negative News seems to be tied to advertising Dollars or something.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2476

Introducing M2

The big news of course is the upcoming transition from socket 939 to socket M2. M2 will have 940 pins, and while the actual layout of motherboards with the new socket will be very similar to socket 939, processors will obviously not be pin compatible. The same goes for current socket 940 processors: they won't work or fit in socket M2. That makes sense, considering that M2 will require the use of DDR2 memory. The similarity in layout will allow motherboard makers to quickly adapt 939 designs to M2, however.

Initially slated to support up to DDR2-667, the socket M2 processors will launch in the second quarter of 2006. Codenamed Orleans for the single core processors

Note the DDR2 support here? Also notice the nice Spin? If Intel were changing anything, everyone at Anandtech will instantly see it as negative. Nothing AMD ever does is wrong or for less than honest reasons.

To AMD's credit, they had to do this because they know the corner cutting fans would try the New Processors in their old boards. Look at the power figures of 95A?

i925 was complete and waiting in the wings whe Intel rushed 8xx-D. It didn't have a chance to have the support it needed for 8xx. The main parts from the Arbitraction Chip and beefier power support.

Donnie27

[NiCKE^]

Saya, what I meant when I said those which is out now I meant the i955X chipset and nothing else.

[saaya]

If you don't know, then why are you in the INTEL side of the forum?
Google it! :rolleyes:
"i errr... i know it i! i really know it, i swear!!
"so what is it?"
"i....ill never teeeeelll :P"
reminds me of kids in kingergarten :stick:
there is nothing to know here, its like what you consider a good car, theres no right or wrong.
sooo, what do you consider netburst?
i bet you dont even know what to say now, you probably just repeat what you hear and read elsewhere :D


I was just fallowing up with Donnie27, that's all. I don't see why anyone else should be so touchy about that. After all I quoted him, in reference to him alone. And tell me which "THREAD after THREAD" I'm pushing off topic? I'm dying to see how many and when have I said I luv AMD? :confused:you walk around and bash amd and praise intel, you rarely back up your statements. this all wouldnt be a problem if you wouldnt post your opinion in such a provoking manner, wich repeatedly incited flames and provoked other members.


Now, what does that ^^^^ have to do w/ Conroe, to me it sound like an Intel vs AMD.......

And I guess it hasn't stopped there> hasn't it? (http://www.xtremesystems.org/forums/showpost.php?p=1198983&postcount=32)you didnt get my point either.
sorry if i didnt make it clear.
my point was that ,if the rumor i heard is true, its stupid of intel to change the pinout of ddr3 dimms just to hurt amd, because if it would be identical intel customers would benefit more than anybody else, because intel already has a widespread ddr2 plattform, amd doesnt.
so if they would keep them compatible people could go and use ddr3 in their ddr2 boards. they probably wouldnt be able to run them at full speed and get as much performence out of it as on the latest boards, but thats thir decision. making it impossible for their own customers to do this, just to try to hurt amd is imo stupid...

do you see anybody besides you post "i love intel/amd" in bold letters with font size 8?
i posted that i heard intel would be pushing a new pinout for ddr3 just to make it incompatible with ddr2, wich i said, i dont like and think is a bad move.
you just posted "Don't forget the useless 3DNow Pro!!!, the on-goning lame lawsuit, the company's CEO which middle name is "Jesus".
Their fab in a Foreign country.
The company they always run to for help (IBM) because they have no clue of what to do next and have to pay them massive ammounts of money for their technology because they don't have own of their own.
And so on....."

what does this have to do with conroe? nothing..
its not even amd vs intel, its not even a debate, its just you standing on a soap box :cussing:

its not informative, its not an interesting debate, it isnt even a debate, its just annoying and people keep complaining about it.
so again, stop it, and please move on.



dramurai?a name given to the dram cartell, samsung hynix and micron and i think one or two others.

even though it cant be prooved most of the time, everbody i the industry knows this cartell exists.


Now it's my turn to :rofl:
The Germans, your tax dollars, Paid for Fab 30, it cost 2.5 billions.
The Germans, your tax dollars, Paid for Fab 36, 1.45 Billion with another 600 million coming from the EU, still feel like laughing?:p:
yeah, i still dont get what this has to do with ddr2 and ddr3...
why should the eu or german government invest in ddr2 or ddr3???
and about tax dollars spent for fabs of big companies, what world do you live in? this is happening worldwide, and its called capitalism.
lets please get back on topic, i dont see what the eu and germany have to do with ddr2 and ddr3 and conroe :P




I could give at least 60 more links if you like. These are Grants and Loans, some call it porkbarrel.

Higher sales figures when you get to sell the same processors to the same folks. Just as folks did and have done since Intel and AMD started making processors. The difference between Intel and AMD is that AMD is a Parts supplier and Intel is a whole systems supplier. No, I don't think it is fair when Intel has to compete against the Germans and the EU, that's BS! What if the U.S. did the same for Intel? Still laughing, hehehe?

Donnie27
dont get your point... why do you care about this?
jobs for people in dresden, and some support for amd, wich improves the competition betwen intel and amd and will lead the faster and cheaper cpus for all of us, already has lead to this...
and do you think this is the first time a country has paid a company to invest into a local plant? again, what world are you living in?
almost all fabs that get built are supported by the region its built in...

now what does this have to do with conroe?


YES! They are expecting ROI afterall. That'd give them (and AMD) the Right not only ask, but demand compatability=P These folks are lieing back, then waiting for Intel and other companies to R&D the tech then *Freeload. Conroe will launch with DDR2 then move to DDR3.
dont forget it was amd who pushed ddr, not intel ;)
and yes, trying to make the competition do the work is ntohign new in the business either... ati and nvidia are doing it all the time.


Be warned, Intel has pulled some good April Fool's launches.

Donnie27like what?


No, we don't have ugly threads when folks come with facts instead of feelings for Company-A or Company-B. Chevy doesn't help Ford, Coke doesn't help Pepsi, and FedX doesn't help UPS. Why in Blue Blazes should Intel be expected to help AMD? Both companies have done great things and both have had taken turns at screwing the consumers. But Fans don't like to see or hear the true suckyness of their favorite. Then the flames start. That's whacky!thanks for geting this thread back on track.
i never said intel should help amd with ddr2/ddr3, all i said was that changing the layout of the ddr3 dimms just to hurt amd is a bad move imo. intel and amd, and intels and amds customers noth have a disadvantage from this.
you saw it as an offense towards intel, wich is not what i meant...
if intel makes ddr3 dimms incompatible with ddr2 just to hurt amd, then it means people will have to buy new boards for conroe to run ddr3 once its out, wich wont be that long after conroes launch.

so i hope this rumos isnt true and ddr2 and ddr3 dimms will hopefully have the same layout, or, ddr3 gets launched together with conroe, wich would be really sweet, but unfortunately unlikely.



Intel said Conroe would "use the best tech from Pentium 4 and Pentium M". Intel showed slides of as Nasgul linked to. So I don't know why anyone would say otherwise and that's what started all of this. Conroe's launch has been moved up. i965 launch has been moved back to launch with it.
Donnie27i say otherwise. conroe doesnt use the best of netburst, conroe is a new design wich is optimized for performence per watt, netburst is optimized for speed and performence no matter what. you cant tak the best of netburst and make a new cpu wich is suddenly a great performence per watt cpu.

its like saying you too the best out of a server to build a laptop.
its contradicting.... a server runs how, consumes a lot of power and is designed to handle big loads of work and power consumption and heat output isnt that important.
a laptop has to consume as little power as possible by offering just about the performence you ned, nothing more, as it would cost power wich means shorter battery life.

the things that come to my mind when i think of netburst, the important aspects and strong points of the design, are not used in conroe afaik, and it wouldnt make any sense to use it in conroe.

now again, what do you think of netbursts strong and most important points?
what comes to your mind?
im curious about what you guys are seing as netbursts strong points.


I kinda find alot of this to be humerous. What gives you the idea that Conroe will be DDR3? In addition who told you that AM2 will by DDR3? ... lol :D

I dont want to sound rude or anything but you guys really dont have a clue. I can tell all of you right now that AM2 will be DDR2, so will Conroe. I do know that AM2 is 100% going to use DDR2 this "skipping" scenerio is simply not going to happen period.

Perhaps when S1207 comes out they might change it to DDR3 but its not going to happen anytime soon.

DDR3 while socket compatible has a different "bios" to the PCB itself (Brian would know more, thats just where my info goes fuzzy). You cant simply plug DDR3 into a DDR2 based box. The whole layout via the memory controller would have to be significantally changed to use it.

Not only is that going to be an issue but Elpeda hasnt shown any interest to us that DDR3 is in the immediate future. DDR2 has atleast another year, probably 2 before we will make that transition since that is what all the road maps say.

I applaud the effort but DDR3 just isnt going to happen on desktop platforms for quite some time. K9 wont use it, perhaps K10 / P5 will, but it certianly wont be until atleast mid 2007 before it will start to be phased to DDR3.

As for commentary on Conroe Ive got mixed feelings on it. I feel a bit disspointed that they went with such a lax cache layout, but we will see how high they scale. As for the rumor of a 333 bus im very skeptical. They are too far into the design phase to make a significant change like that. Most likely it will be only 266 if that.

AMD on the other hand Ive grown more impressed the more I know about AM2. Everyone is in for a real treat when it comes this summer. I have high hopes that we may see 600 1:1 for the first time ever. The next-gen of DDR2 will certianly make that possible as its starting to appear already.
who said am2 will be ddr3?
hmmm i didnt know about this dr2 ddr3 compability thing, as i said, its just a rumor i heard. so teres non way to build a ddr2 board to be ddr3 compatible?
or is it just too much of an effort?
i don really know anything about tis, i just ead that ddr2 and ddr3 are basically compatible, wich amazed me, and i thougt we might see boards that run ddr2 and later ddr3, just like videocards that used gddr2 and then gddr3.

and 600 1:1 sounds nice, but then again, thats just a high number :D
lets wait and see how it performs, im curious how a low latency optimized architecture wil work with high latency memory :D

welcome to XtremeSystems Who :toast:


Nah. AM2 is still S940 last I heard. S1207 comes later I think.


No it really isnt. CSI is what was supposed to make Conroe P5. Now instead they scrapped that idea for a multi-hub based NB system for later next year. That is P5, this is just a refresh.


I believ K9 was effectively scrapped and merged into S939 (X2 was supposed to be K9), so they are both one generation away.
afaik 1207 will be fbdimm, so itl be ddr1 ddr2 ddr3, xdr, xdr2, rambus, whatever you want :D



The reasons for the heated post ( Flames) and hurtful tone of some of the posts? The seemly endless Double standards that are excepted by AMD fans/majority (of the posters), on the Intel side of the forums is one reason. When Intel does this, they are scumbags and lowest of the low lifes. When AMD does the very same thing, it's accepted and defended. They both might have had a very good reason, but you'd never know it. You'd still get the same old tired AMD is the white hat wearing good guy and Intel is more hated than many of the virus writers.

Donnie27

donnie, please just stop this amd intel fanboy stuff and lets go back on topic, k?

No biggie, certain things are not for the 75% of posters who like, follow and even worship AMD. Negative News seems to be tied to advertising Dollars or something.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2476


Note the DDR2 support here? Also notice the nice Spin? If Intel were changing anything, everyone at Anandtech will instantly see it as negative. Nothing AMD ever does is wrong or for less than honest reasons.

To AMD's credit, they had to do this because they know the corner cutting fans would try the New Processors in their old boards. Look at the power figures of 95A?

i925 was complete and waiting in the wings whe Intel rushed 8xx-D. It didn't have a chance to have the support it needed for 8xx. The main parts from the Arbitraction Chip and beefier power support.

Donnie27
could you please stop this whining about how everybody loves amd and hates intel?

[Nasgul]

i bet you dont even know what to say now, you probably just repeat what you hear and read elsewhereOh! like you're the only one that knows something? please!

Like I said, Google it! when in doubt.




you walk around and bash amd and praise intel, you rarely back up your statements. this all wouldnt be a problem if you wouldnt post your opinion in such a provoking manner, wich repeatedly incited flames and provoked other members.bash amd? so, telling some REAL facts is "bashing" all of the sudden? BTW, if other members are provoked by some facts, then they need to get a clue, it ain't my fault that they don't know or understand. I don't get so hung-up on Intel bashing by other fellas, could careless or when I get quoted and rant about it.



you didnt get my point either.
sorry if i didnt make it clear.
my point was that ,if the rumor i heard is true, its stupid of intel to change the pinout of ddr3 dimms just to hurt amd, because if it would be identical intel customers would benefit more than anybody else, because intel already has a widespread ddr2 plattform, amd doesnt.
so if they would keep them compatible people could go and use ddr3 in their ddr2 boards. they probably wouldnt be able to run them at full speed and get as much performence out of it as on the latest boards, but thats thir decision. making it impossible for their own customers to do this, just to try to hurt amd is imo stupidSo it is Intel's fault that amd refused to use DDR2 and now that DDR3 will come out, it is Intel's fault that they'll make DDR3 not compatible in DDR2 RIMMs? Well, can't you blame Intel for not making DDR2 compatible with DDR1?

Can you blame Intel for not making PCI-Ex16 compatible with AGP also? What else can we blame Intel for not making it compatible with what AMD refuses to change? How about BTX?

If you ask me Intel does what's best for them, so why blame Intel for not depending on other major corporations? Is anyone blaming Intel for not having 3DNow Pro? I thoguht so.



do you see anybody besides you post "i love intel/amd" in bold letters with font size 8?And what's wrong with saying I luv Intel?
Here: I luve Intel. :oscar: , btw, why do you always exaggerate? Font size 8? pushing thread after thread off topic?
I'll always favor Intel over any other, it's my choice.


i posted that i heard intel would be pushing a new pinout for ddr3 just to make it incompatible with ddr2, wich i said, i dont like and think is a bad move.Yes, you're repeating yourself, that's cool I understand. So let's keep blaming Intel for not making Socket 478 backwards compatible with Socket 754 and LGA with socket 939. What else is there to blame Intel for? Hope you haven't forgotten that Intel had sockets compatible with amd CPUs back in the day.



its not informative, its not an interesting debate, it isnt even a debate, its just annoying and people keep complaining about it.
so again, stop it, and please move on. A little fact is always unpleasant. While it may not be an interesting debate, I wasn't debating, just quoting Donnie, just like you keep quoting him,which is not interesting at all. But hey! you're entitled to your opinion and so does he, except he knows more than....... And so I move on :devil:

[FUGGER]

you walk around and bash amd and praise intel, you rarely back up your statements. this all wouldnt be a problem if you wouldnt post your opinion in such a provoking manner, wich repeatedly incited flames and provoked other members.


If we stand up for Intel we have to defend ourselves. Pretty sad that this happens here.

No one can convince me AMD makes a better gaming rig, thats for sure.

[Willis]

If we stand up for Intel we have to defend ourselves. Pretty sad that this happens here.

No one can convince me AMD makes a better gaming rig, thats for sure.
I know that you know that this is not true ..
they equalise ;)

[Pinnacle]

:rolleyes:
"i errr... i know it i! i really know it, i swear!!
"so what is it?"
"i....ill never teeeeelll :P"
reminds me of kids in kingergarten :stick:
there is nothing to know here, its like what you consider a good car, theres no right or wrong.
sooo, what do you consider netburst?
i bet you dont even know what to say now, you probably just repeat what you hear and read elsewhere :D

you walk around and bash amd and praise intel, you rarely back up your statements. this all wouldnt be a problem if you wouldnt post your opinion in such a provoking manner, wich repeatedly incited flames and provoked other members.

you didnt get my point either.
sorry if i didnt make it clear.
my point was that ,if the rumor i heard is true, its stupid of intel to change the pinout of ddr3 dimms just to hurt amd, because if it would be identical intel customers would benefit more than anybody else, because intel already has a widespread ddr2 plattform, amd doesnt.
so if they would keep them compatible people could go and use ddr3 in their ddr2 boards. they probably wouldnt be able to run them at full speed and get as much performence out of it as on the latest boards, but thats thir decision. making it impossible for their own customers to do this, just to try to hurt amd is imo stupid...

do you see anybody besides you post "i love intel/amd" in bold letters with font size 8?
i posted that i heard intel would be pushing a new pinout for ddr3 just to make it incompatible with ddr2, wich i said, i dont like and think is a bad move.
you just posted "Don't forget the useless 3DNow Pro!!!, the on-goning lame lawsuit, the company's CEO which middle name is "Jesus".
Their fab in a Foreign country.
The company they always run to for help (IBM) because they have no clue of what to do next and have to pay them massive ammounts of money for their technology because they don't have own of their own.
And so on....."

what does this have to do with conroe? nothing..
its not even amd vs intel, its not even a debate, its just you standing on a soap box :cussing:

its not informative, its not an interesting debate, it isnt even a debate, its just annoying and people keep complaining about it.
so again, stop it, and please move on.

a name given to the dram cartell, samsung hynix and micron and i think one or two others.

even though it cant be prooved most of the time, everbody i the industry knows this cartell exists.


yeah, i still dont get what this has to do with ddr2 and ddr3...
why should the eu or german government invest in ddr2 or ddr3???
and about tax dollars spent for fabs of big companies, what world do you live in? this is happening worldwide, and its called capitalism.
lets please get back on topic, i dont see what the eu and germany have to do with ddr2 and ddr3 and conroe :P



dont get your point... why do you care about this?
jobs for people in dresden, and some support for amd, wich improves the competition betwen intel and amd and will lead the faster and cheaper cpus for all of us, already has lead to this...
and do you think this is the first time a country has paid a company to invest into a local plant? again, what world are you living in?
almost all fabs that get built are supported by the region its built in...

now what does this have to do with conroe?


dont forget it was amd who pushed ddr, not intel ;)
and yes, trying to make the competition do the work is ntohign new in the business either... ati and nvidia are doing it all the time.

like what?

thanks for geting this thread back on track.
i never said intel should help amd with ddr2/ddr3, all i said was that changing the layout of the ddr3 dimms just to hurt amd is a bad move imo. intel and amd, and intels and amds customers noth have a disadvantage from this.
you saw it as an offense towards intel, wich is not what i meant...
if intel makes ddr3 dimms incompatible with ddr2 just to hurt amd, then it means people will have to buy new boards for conroe to run ddr3 once its out, wich wont be that long after conroes launch.

so i hope this rumos isnt true and ddr2 and ddr3 dimms will hopefully have the same layout, or, ddr3 gets launched together with conroe, wich would be really sweet, but unfortunately unlikely.


i say otherwise. conroe doesnt use the best of netburst, conroe is a new design wich is optimized for performence per watt, netburst is optimized for speed and performence no matter what. you cant tak the best of netburst and make a new cpu wich is suddenly a great performence per watt cpu.

its like saying you too the best out of a server to build a laptop.
its contradicting.... a server runs how, consumes a lot of power and is designed to handle big loads of work and power consumption and heat output isnt that important.
a laptop has to consume as little power as possible by offering just about the performence you ned, nothing more, as it would cost power wich means shorter battery life.

the things that come to my mind when i think of netburst, the important aspects and strong points of the design, are not used in conroe afaik, and it wouldnt make any sense to use it in conroe.

now again, what do you think of netbursts strong and most important points?
what comes to your mind?
im curious about what you guys are seing as netbursts strong points.


who said am2 will be ddr3?
hmmm i didnt know about this dr2 ddr3 compability thing, as i said, its just a rumor i heard. so teres non way to build a ddr2 board to be ddr3 compatible?
or is it just too much of an effort?
i don really know anything about tis, i just ead that ddr2 and ddr3 are basically compatible, wich amazed me, and i thougt we might see boards that run ddr2 and later ddr3, just like videocards that used gddr2 and then gddr3.

and 600 1:1 sounds nice, but then again, thats just a high number :D
lets wait and see how it performs, im curious how a low latency optimized architecture wil work with high latency memory :D

welcome to XtremeSystems Who :toast:


afaik 1207 will be fbdimm, so itl be ddr1 ddr2 ddr3, xdr, xdr2, rambus, whatever you want :D



donnie, please just stop this amd intel fanboy stuff and lets go back on topic, k?

could you please stop this whining about how everybody loves amd and hates intel?


DAMN! saaya make your posts shorter!! I need a cup of coffe to go through it all, and I dont even drink coffee hehe :)

[tritium]

Will DDR2 really increase performance on the AMD platform? How much and why?

[Pinnacle]

Very little from what I hear!

[saaya]

Oh! like you're the only one that knows something? please!

Like I said, Google it! when in doubt. haha, yeah right, see, as i said, you dont even know what your talking about :D
im through trying to discuss things with you, you dont reply to questions and dont back up your statements, believe whatever makes you happy.


bash amd? so, telling some REAL facts is "bashing" all of the sudden? BTW, if other members are provoked by some facts, then they need to get a clue, it ain't my fault that they don't know or understand. I don't get so hung-up on Intel bashing by other fellas, could careless or when I get quoted and rant about it. there are good things and bad things about all companies and all products, you constantly point out the bad things, thats bashing.
and dont try to disctract by pointing at others.


So it is Intel's fault that amd refused to use DDR2who said that?

and now that DDR3 will come out, it is Intel's fault that they'll make DDR3 not compatible in DDR2 RIMMs?thats a rumor i heard wich is appreantly wrong

Well, can't you blame Intel for not making DDR2 compatible with DDR1?i never heard that it was intel who tried to make ddr2 incompatible with ddr1 on purpose, did you?


Can you blame Intel for not making PCI-Ex16 compatible with AGP also?i see you know a lot about serial interfaces...
its impossible to make agp and pciE compatible

What else can we blame Intel for not making it compatible with what AMD refuses to change? How about BTX?btx is dead... so lets skip this


If you ask me Intel does what's best for them, so why blame Intel for not depending on other major corporations? Is anyone blaming Intel for not having 3DNow Pro? I thoguht so.right, whats best for them, not whats best for their customers. most big companies do this, and imo its stupid and huting themselves...


And what's wrong with saying I luv Intel?
Here: I luve Intel. :oscar: , btw, why do you always exaggerate? Font size 8? pushing thread after thread off topic?
I'll always favor Intel over any other, it's my choice.and here you proved that your biased towards intel, you will always buy intel no matter what, even if amd or another company has better products for less.
this isnt IntelSystems, is XtremeSysems.
i didnt check, so what font size was it? 5? 4?
does it matter?
the point is that nobody cares...
you dont discuss a point, you just abuse xs to post your blog, what you like, what you dont like etc... please go and open your own blog site to do that.


Yes, you're repeating yourself, that's cool I understand. So let's keep blaming Intel for not making Socket 478 backwards compatible with Socket 754 and LGA with socket 939. What else is there to blame Intel for? Hope you haven't forgotten that Intel had sockets compatible with amd CPUs back in the day.i didnt blame intel for anything... read my posts.
i only posted about a rumor i heard and posted my opinion about it and explained why i think that way.


A little fact is always unpleasant. While it may not be an interesting debate, I wasn't debating, just quoting Donnie, just like you keep quoting him,which is not interesting at all. But hey! you're entitled to your opinion and so does he, except he knows more than....... And so I move on :devil:more than you? yes :D
debating with donnie is fun, because he actually makes a point and he likes intel but hes not biased and would buy intel or any other companies products no mater what. he has an opinion based on points, wich he explains and discusses. hs very stubborn and doesnt like to admit he was wrong about something, but so what :D

oh, i see you called charles for help again?... :lol:


If we stand up for Intel we have to defend ourselves. Pretty sad that this happens here.you dont stand up for intel, you dont say intel has great products, you say intel products > amd products!
thats not standing up for intel...


No one can convince me AMD makes a better gaming rig, thats for sure.thats what makes you biased.
you could never be convinced that amd has good products even if they had the better product.
theres nothing wrong with this, as long as you dont try to force your will onto others by claiming your opinion = truth, period.
and i hope it will remain like this :)


DAMN! saaya make your posts shorter!! I need a cup of coffe to go through it all, and I dont even drink coffee hehe :)
my bad :sofa:
i got a cold so first i was in bed and couldnt reply for some time, and now i hav enough time to reply to all posts and couldnt resist :D
atm i cant really do anything but lie in bed and post, anyways :D


Will DDR2 really increase performance on the AMD platform? How much and why?thats what im curious about as well, from the x2s we know that a4s are anything but bandwith limited, so whats the point of adding more bandwidth?
maybe amd redesigned their cores to be more bandwidth hungry?
but i dont really think so... in the end they do scale with higher memory bandwidth, but not that much.
depending on the applications ddr400 to ddr500 gave a 0-5% boost iirc.
the maximum boost i xpect from ddr on amd systems will be 10%, and not in all apllications.

now, lets get back to conroe, shall we? :p:

whats netburst according to intel?
rapid exectution engine (double pumped alus)
hyper pipeline long pipline clocked to high speeds/low ipc
trace cache/micro ops fusion/tiny l1 cache that can only store already decoded micro ops)
afaik conroe doesnt feature any of this...

what do i think of as the strong and characteristical points of netburst?
what dfferes the design from other designs, imo, its the following things additional to the above mentioned definition from intel:
large l2 cache
hyper threadding
high bandiwidth optimized quad pumped bus

conroe uses ht and the qpb, but since intel uses the qpb on their server and mobile designs as well id rather not name it a cpu design specific feature.
just as much as ddr isnt a design specific feature...
and ht was introduced in xeons first, and is used used in none netburst designs as well, ibm cpus also have ht... you just double some parts of the logic wich improves the eficiency of the cpu. bu its still an architectural charaveristic of netburst imo... as its done inside the core and the design has to b built in a certain way to make it possible.
so the only thing that conroe and netburst have in common is HT in my opinion

but does this make conroe a netburst cpu?
or a netburst derived design?
most of the pentium ds dont even have ht enabled, and until the b revision of nothwood there was no ht on netburst cpus either...
so its more of an add on like sse3 or 64bit instructions or larger cache in my opinion. if the pentium m would have ht, wich would have been a rally nice feature btw, would that have made it a netburst cpu?
imo no...


now what do conroe and yonah have in common from what we know so far?
shared l2 cache
on die system interface hub
short pipeline clocked to medium/high speeds speeds + high ipc

[NiCKE^]

Please guys, can't you just keep the messages short or do this in PM? Ruin such a "good" thread :) Let us who are looking forward to Conroe to get released discuss it here and no fanboy and childish behaviour please!

Saaya should maybe consider to use the "Post Quick Reply" button more often :D

[Nasgul]

im through trying to discuss things with you, you dont reply to questions and dont back up your statementsYou're not trying to discuss things, just asking what is netburst? :rolleyes: . Sorry but you're not the only one that knows "something".


you constantly point out the bad things, thats bashing.
and dont try to disctract by pointing at others.
you just posted "Don't forget the useless 3DNow Pro!!!, the on-goning lame lawsuit, the company's CEO which middle name is "Jesus".
Their fab in a Foreign country.
The company they always run to for help (IBM) because they have no clue of what to do next and have to pay them massive ammounts of money for their technology because they don't have one of their own.
And so on....."So is that bashing? AFAIK, those are facts. And I'll still reiterate it. Now tell me which of that is not true? I quoted Donnie, Donnie quoted me, in between some fella gets upset? big deal move along, don't get distracted.


and here you proved that your biased towards intel, you will always buy intel no matter what, even if amd or another company has better products for less.Finally, you've cracked the case :clap: . & AMD has not a better product and never will (http://www.xtremesystems.org/forums/showpost.php?p=1190471&postcount=11). Also I don't buy from corporations that bases their businesess abroad, I'm 100% hard core American. I put in my PC a CPU from an Ameican Corporation that has plenty of FABs here in the US, just like I love my Hot-Dogs with real American Cheese and not Sour-Krout.



this isnt IntelSystems, is XtremeSysems.There are "choices" in the menu, how hard is that to understand? FYI, this is the Intel section, and thus it will remain that way.




i didnt check, so what font size was it? 5? 4?
does it matter?
the point is that nobody cares.So, you just like to exaggerate then. LOL. Funny how "no one" cares but yet it gets pointed out.


you dont discuss a point, you just abuse xs to post your blog, what you like, what you dont like etc... please go and open your own blog site to do that.I was discussing a point with Donnie which is a fact, why is that part so hard to understand? I don't abuse XS, I'm not the one with 20+ post per day, I'm no where near that record, I like to point out to the preacher, to practice what he preaches. BTW, let me know when this becomes "saayaxssystems", okay?


i didnt blame intel for anything... read my posts.
i only posted about a rumor i heard and posted my opinion about it and explained why i think that way.What differnece is that from posting my opinion as well? You think you're the only one entitled to post your own opinions, thoughts, ideas.....? I don't need to explain the self-explanatory to Donnie, I don't undermine a persons knowledge.


debating with donnie is fun, because he actually makes a point and he likes intel but hes not biased and would buy intel or any other companies products no mater what.K, so you like him basically because he's not biased? Well, now you know me. Is that the reason for the animosity? Hey! I'm 100% biased, so what? what do you have against it? Right now I like Intel better for so many reasons, not just CPU alone, if in the future I decide to change, no one will ever know. I luv Intel.


oh, i see you called charles for help again?Actually, never have & never will. Hey! maybe some other BIASED person reported it and asked you to keep it short and on topic, use PMs if discussing with Donnie makes you happy. After all this is the Intel section of the forum (FYI). Wouldn't be easier to go administrate the amdzone.com?

[.sentinel]

Quote:
Originally Posted by saaya
and here you proved that your biased towards intel, you will always buy intel no matter what, even if amd or another company has better products for less.

Finally, you've cracked the case . & AMD has not a better product and never will. Also I don't buy from corporations that bases their businesess abroad, I'm 100% hard core American. I put in my PC a CPU from an Ameican Corporation that has plenty of FABs here in the US, just like I love my Hot-Dogs with real American Cheese and not Sour-Krout.

Okay so you just insulted all the germans, all the british, all the africians on this WHOLE board. What shoes do you wear, what clothes do you where? I bet 100 bucks that it is made in china or some other country.

[FunkyRider]

I would guess Nasgul is just a stubborn who will never listen to other's opinion, so why bother trying to convince him?

Just let him live in his own dream then.

BTW, sorry to wake you up from your dream but... Pure 4.02ghz is just too plain for any P4 6 series

[Sentential]

hmmm i didnt know about this dr2 ddr3 compability thing, as i said, its just a rumor i heard. so teres non way to build a ddr2 board to be ddr3 compatible?
or is it just too much of an effort?

Its completly and totally not possible. It would require the whole platform to be retrofitted

[Donnie27]

thats what im curious about as well, from the x2s we know that a4s are anything but bandwith limited, so whats the point of adding more bandwidth?
maybe amd redesigned their cores to be more bandwidth hungry?
but i dont really think so... in the end they do scale with higher memory bandwidth, but not that much.

Fact, each time a multiplier is lowered, latency is too. Each time a Clock Gen has to divide, it increases latency. The number of clock gens add latency if I understand it correctly. DDR3 brings many improvements over DDR1 and 2. It makes an already great gaming system's integrated memory controller even better.

AMD's own negative PR for something they didn't have at the time Kept them from moving to DDR2 sooner. The did this with SSE 1, 2, and 3, Strained Silicon, and etc... Same goes for the Mod and Overclocking friendly BTX BTW. "if we don't have it, we must on speak poorly of it". Something AMD and Intel does BTW. Both have taken turns eating crow and were left with egg on their face.

IMHO, the folks talking about the Yonah and increased latency from Pentium M's 10 to Yonah's 14 blew it.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2627&p=3

Something else might be at fault here, not what the Great Anand said. Why? Maybe it's just that Smart Cache, Arbitration Built in, and etc.. causes a hit to performance. If Yonah wasn't a combind Core (.Exe unit) using one L2, it would be two full cores on one package like Presler. Then you'd have Two processors with a combind 20ns instead of 14. So 14 is a savings, not lost, its good not bad. There's plenty else to hurt Single Thread performance here.


I would guess Nasgul is just a stubborn who will never listen to other's opinion, so why bother trying to convince him?

It is not a matter of listening, it has all the rights in the world to his opinions as well. Many folks agree to disagree and move on. It's my opinio that there a clear BIAS for AMD on the Intel side of the forum, that seems strange to me. Here's my just completed system.

Super Flower 301 Black Case and Drives
PSU Antec 550W True Power 2.0 SLI
Asus A8N SLI Standard
Athlon64 3500+ Venice
Zalman CNPS-7000 CU
1GB PQI PC-4000
Retail ATI X800XL
SoundBlaster X-Fi Xtreme Music
Hauppauge-TV Card
Western Digital SATA-II 250GB HDD 2500KS 16mb cache
NEC 3540 DL 16X DVD-RW
LiteOn 5232k Combo 52X CD-R/16X DVD-ROM
Sony Floppy
MS MM Natrual MM KB and MS PS2 Mouse
Dual CRTs, 21" and 19"
Will run XP Pro but Win2K Pro installed now. Then this rig will be updated to Opteron or X2 for sure next year.

So if I'm an Intel employee as I was called, way to pro Intel as I'm branded and way Intel BIASed in general, something is wrong. Hell yes I've been wrong and pointed out some blunders I've made. If I thought AMD sucked I wouldn't have built the above system though.

Donnie27

[Carfax]

Will DDR2 really increase performance on the AMD platform? How much and why?

From what an AMD rep said, DDR2 should increase the performance of the K8 by 10 > 15% on average. That sounds kinda high to me (he may have been referring to the boost for server oriented tasks which are bandwidth sensitive), but when you think about it, with DDR2-667, the X2 or dual core Opterons will be getting quite a bit more bandwidth per core than they do with regular DDR.

So in bandwidth limited circumstances, DDR2 should improve performance significantly I should say.

[JoeBar]

It just keeps getting harder to follow this thread. :mad:
Why u ppl can't solve your unsolved diffs over PM's? Are we supposed to gain anything from your argue? NO!

There are many ppl here who want to learn and DON'T care if their intel system is better than amd or the opposite...

[Donnie27]

Tomorrow Intel will launch Presler and Cedar Mill. In June, they'll launch Conroe and i965 chipset. I'm kind of shocked they didn't launch Presler and Cedar Mill for the Xmas sales run.

The problems started when folks jumpped in and tried to claim AMD will do or nearly do the same launchwise with 65nm (Cheerleaders). On some forums, nothing can be said in an negative manner about AMD, even if it is factual, with links and etc... (AMD Defenders come to the resue).

Folks talking about Conroe didn't need to bring up AMD at all. But message boards are about info and not about Cheerleading for Intel or AMD. That old addage of, "if you are not for me, you're against me" doesn't fit in this market. Nothing on this earth is perfect. I post on the Intel side of the forum because 70% of AMD Fans do believe AMD is Godly and very perfect.

To the Other guy, it's called ROI or Return On Investment. Both the Germans and the EU didn't do this because they like AMD. Israel is said to be doing the same thing for Intel.

Donnie27

[grimREEFER]

Will DDR2 really increase performance on the AMD platform? How much and why?
no, but its cheaper, faster, and better, and u can get 1gb stix of it and not cry about how crappy it overclocks

[vanovich]

doese conroe have duel core? what makes it better than presley ? specs anyone "link" . are conroe good oc? socket 775? how many watts doese it burn ? i like the results on presely at the moment . please all , hold this forum about conroe .takes hell of a timme reading the posts . thanks .

[Who]

welcome to XtremeSystems Who :toast:

Thanks :toast:

vanovich: Yes, Conroe will have dual cores. It's better than Presler because it has a more efficient design based on Netburst and Pentium M architecture. AFAIK, nobody's overclocked a Conroe yet and it'll be socket 775.

[vanovich]

thnks who . what about l2cashe on conroe?

[Who]

I believe some Conroes will have 4mb shared L2 and some will have 8.

[vanovich]

I believe some Conroes will have 4mb shared L2 and some will have 8.
nice . probably quite expensive then? any retail prices yet?

[FunkyRider]

here's some info about Conroe

* Initially no HT like current Pentium D, so two cores, two parallel threads
* It had TDP of 6x Watts, in contrast to 120W of PD 8xx and 8x W of PD 9xx
* It migrated the power efficient designs existing in Pentium M (Banias arch.)
* It will be built with 65nm manufacture and will have a shrink to 45nm some years later
* It will have a short pipeline, hence lower frequency than current P4/PD
* Current ES chips (Engineer Sample) works @ around 1.8GHz for a general interest to you guys
* Some rumers mentioned this "2.93GHz", not quite sure whether its the starting frequency or highest stock frequency.
* Current 975X platform needs a "platform refreshment", a.k.a. revision to support Conroe processors
* As for price, the price when it hits the shelf will be expensive, cuz this would be recoginzed to as a "new generation processor" from Intel.

[IluvIntel]

no, but its cheaper, faster, and better, and u can get 1gb stix of it and not cry about how crappy it overclocks

And cooler, 1.8v default for basic DDR2 sticks V 2.65v basic DDR sticks. :D

[AkXb70]

here's some info about Conroe

* Initially no HT like current Pentium D, so two cores, two parallel threads
* It had TDP of 6x Watts, in contrast to 120W of PD 8xx and 8x W of PD 9xx
* It migrated the power efficient designs existing in Pentium M (Banias arch.)
* It will be built with 65nm manufacture and will have a shrink to 45nm some years later
* It will have a short pipeline, hence lower frequency than current P4/PD
* Current ES chips (Engineer Sample) works @ around 1.8GHz for a general interest to you guys
* Some rumers mentioned this "2.93GHz", not quite sure whether its the starting frequency or highest stock frequency.
* Current 975X platform needs a "platform refreshment", a.k.a. revision to support Conroe processors
* As for price, the price when it hits the shelf will be expensive, cuz this would be recoginzed to as a "new generation processor" from Intel.
HT may or may not be supported at first, but eventually ALL intel processors will have it (i remember reading an intel reply something along the lines of "HT is going the way of MMX. invisible to the customer insofar as advertising, but still there nonetheless)
it will have 14 pipeline stages (compared to 12 for dothan, 31 for prescott).
as for the 2.93GHz thing, i would think that would be the higest frequencey on launch (or near top tier). maybe itl use .5 increments for multipliers since the fsb is going to be so high?

[FunkyRider]

yeah, .5 increment of multi seems reasonable

shame that Conroe still uses 1066FSB but at the same time this is a good thing cuz it keeps multi high enough for us to OC :clap:

[Sentential]

There might be more truth to the "1333FSB" rumor than what I initially gave it credit for.

I was looking at a i975XBX and I saw something curious at the bottom....

http://www.theinquirer.net/images/articles/presler426overclk.jpg

FSB Frequency Override [1333Mhz] ?

No thats not an easter egg or a photochop, that is really what it has listed. I know the last "official" info on Conroe suggested a 1066 part but this has me wondering... Maybe at the very least the EE has a 1333 FSB? Atleast it would explain the need for a platform refresh.

[Willis]

you dont stand up for intel, you dont say intel has great products, you say intel products > amd products!
thats not standing up for intel...


thats what makes you biased.
you could never be convinced that amd has good products even if they had the better product.
theres nothing wrong with this, as long as you dont try to force your will onto others by claiming your opinion = truth, period.
quoted for truth

[AkXb70]

There might be more truth to the "1333FSB" rumor than what I initially gave it credit for.

I was looking at a i975XBX and I saw something curious at the bottom....

http://www.theinquirer.net/images/articles/presler426overclk.jpg

FSB Frequency Override [1333Mhz] ?

No thats not an easter egg or a photochop, that is really what it has listed. I know the last "official" info on Conroe suggested a 1066 part but this has me wondering... Maybe at the very least the EE has a 1333 FSB? Atleast it would explain the need for a platform refresh.
i can see the EE as having a 1333MHz fsb. also, all xeon derivitives are supposed to have 1333fsb's (dual independent ones for 2+ cpu's)

with amd going to a 1333 effective bus (ram bandwidth wise, not HTT, i know that) intel would probably want to 'keep up with the jonses' in that respect

[saaya]

nasgul, stop making provoking comments.
yhpm
sorry guys, you right, we should have taken this to pm earlier.

funkyrider, i dont want to convince nasgul of anything, i dont care whether he likes amd or intel, i just dont wanna hear about it every time i read one of his posts.

carfax, amd said they would go ddr2 as soon as it hits 800 speeds because thats where the increased bandwidth makes up for the added latency.
then later they said they will do it when ddr2 hits 1066 speeds...
i think they will just add more dividers to allow fsb 200 mem 400 (ddr800) and fsb 200 mem 533 (ddr1066)

grimREEFER, no amd vs intel discussion here, please stop it.

* Some rumers mentioned this "2.93GHz", not quite sure whether its the starting frequency or highest stock frequency.
afaik highest frequency intel is aiming for, for the launch

* Current 975X platform needs a "platform refreshment", a.k.a. revision to support Conroe processors
afaik no?

* As for price, the price when it hits the shelf will be expensive, cuz this would be recoginzed to as a "new generation processor" from Intel.
id guess, yeah, very expensive, especially since its probbably going to beat amd, so they can ask for a lot

iluvintel, less volts doesnt mean it runs cooler, ddr2 consumes more power than ddr1. pentium pro cpus used 3.3v vcore and consumes 10W, current pentium d cpus have a vcore of 1.35 and consume up to 150W.
vcore alone or vdimm alone doesnt show anything :)

[saaya]

ahhh i think i got it :D
conroe needs a new 975x revision, yes and no.
the current chipset is compatible afaik, BUT it cant provide the 1333fsb...
thats what a new revision is needed for.
id still guess that only the XE of conroe will run with 1333fsb, so other conroes should be 1066 and they should work in current 975 and 955 chipsets

[zakelwe]

IMO Conroe does not need a 1333 bus, sounds more like marketing one upmanship than anything. That's not to say it will not have it of course.

If it does I'd be interested to know how much more performance it gets at 1333 than 1066.

Regards

Andy

[grimREEFER]

nasgul, stop making provoking comments.
yhpm
sorry guys, you right, we should have taken this to pm earlier.

funkyrider, i dont want to convince nasgul of anything, i dont care whether he likes amd or intel, i just dont wanna hear about it every time i read one of his posts.

carfax, amd said they would go ddr2 as soon as it hits 800 speeds because thats where the increased bandwidth makes up for the added latency.
then later they said they will do it when ddr2 hits 1066 speeds...
i think they will just add more dividers to allow fsb 200 mem 400 (ddr800) and fsb 200 mem 533 (ddr1066)

grimREEFER, no amd vs intel discussion here, please stop it.

* Some rumers mentioned this "2.93GHz", not quite sure whether its the starting frequency or highest stock frequency.
afaik highest frequency intel is aiming for, for the launch

* Current 975X platform needs a "platform refreshment", a.k.a. revision to support Conroe processors
afaik no?

* As for price, the price when it hits the shelf will be expensive, cuz this would be recoginzed to as a "new generation processor" from Intel.
id guess, yeah, very expensive, especially since its probbably going to beat amd, so they can ask for a lot

iluvintel, less volts doesnt mean it runs cooler, ddr2 consumes more power than ddr1. pentium pro cpus used 3.3v vcore and consumes 10W, current pentium d cpus have a vcore of 1.35 and consume up to 150W.
vcore alone or vdimm alone doesnt show anything :)
i didnt say anything about amd vs intel, is that a typo?:confused:

[FunkyRider]

Conroe's performance will most likely to be held back by the old fashioned FSB. Despite such a high perf. core x 2, the 1333 FSB would barely enough to feed the core with data stream when required.


1066 is way too slow (533 for each IF divided by 2)


... Intel will be damned if it uses 800 FSB for conroe

[Carfax]

1333 FSB provides 10.8 GB/S of bandwidth, which should be enough I think. That would run in perfect sync with DDR2-667, which also provides 10.8 GB/S.

Overclock the FSB to 1600, which many of us will do, and you can pair Conroe up with DDR2-800 @ 1:1 ratio.

BTW, the archaic FSB architecture won't really matter to us, because the FSB can easily feed just two cores. More than two cores however, and AMD's method is much more efficient.

[FunkyRider]

LOL

If you buy a Conroe with 2.66GHz, that's 266x10, you most likely to have a locked multi @ 10

If you want 1600 FSB, which is 400MHz real clock, I doubt Conroe can bust at whoppy 400x10 = 4GHz LOL. If it does, it will blow every CPU away on this planet

[Carfax]

LOL

If you buy a Conroe with 2.66GHz, that's 266x10, you most likely to have a locked multi @ 10

If you want 1600 FSB, which is 400MHz real clock, I doubt Conroe can bust at whoppy 400x10 = 4GHz LOL. If it does, it will blow every CPU away on this planet

Why does it have to be 266x10? If it uses 333 FSB, which it should, it would be 333x8. Bump that up to 400x8 and you've got 3.2ghz, which should be attainable given that it's on a 65nm process, uses very little power and has a medium length pipeline.

High end air coolers next year may be able to manage that kind of overclock, and good watercooling should definitely be able to.

[coldpower27]

here's some info about Conroe

* Initially no HT like current Pentium D, so two cores, two parallel threads
* It had TDP of 6x Watts, in contrast to 120W of PD 8xx and 8x W of PD 9xx
* It migrated the power efficient designs existing in Pentium M (Banias arch.)
* It will be built with 65nm manufacture and will have a shrink to 45nm some years later
* It will have a short pipeline, hence lower frequency than current P4/PD
* Current ES chips (Engineer Sample) works @ around 1.8GHz for a general interest to you guys
* Some rumers mentioned this "2.93GHz", not quite sure whether its the starting frequency or highest stock frequency.
* Current 975X platform needs a "platform refreshment", a.k.a. revision to support Conroe processors
* As for price, the price when it hits the shelf will be expensive, cuz this would be recoginzed to as a "new generation processor" from Intel.

I apologize if some of this information has already been mentioned.
Conroe has a stated TDP of 65W, in contrast to the 130W of the Pentium D 830, 840, Pentium EE 840 and 955.

Yes there will be two versions of the chip for Desktop, Conroe with 4MB of Shared Cache, and Allendale with 2MB of Shared Cache.

Yes the high end frequency is 2.93GHZ which would state a FSB of 1066.

The launch of Conroe has been paried up with the i965 Chipset Series, which supports 533/800/1066 FSB with Dual Channel DDR2 533/667/800. My guess is Conroe will be 1066FSB, while an Extreme Edition could possibly have 1333 FSB.

The 45nm versions Ridgefield 6MB Shared Cache, and Wolfdale 3MB Shared Cache, are guess to have 1333 FSB.

My guess of the lineup is thus:

2.4GHZ/2MB/1066FSB
2.4GHZ/4MB/1066FSB
2.66GHZ/2MB/1066FSB
2.66GHZ/4MB/1066FSB
2.93GHZ/2MB/1066FSB
2.93GHZ/4MB/1066FSB

Extreme Edition 3.00GHZ/4MB/1333FSB (Heh, wishful on my part)

It would depend, if Intel can luanch an entire lineup of CPU's from top to bottom as I suggest above some could be realtively cheap.

The current die size of Conroe is a bit smaller 140mm2, with Allendale probably larger then Cedar Mill's 70mm2. So it wouldn't be extremely expensive for Intel to product these CPU's.

I also think they will keep the 266 intervals, as by that time AMD will have moved Athlon 64 to DDR2 which will be in 333MHZ intevals to match up with DDR2-667.

[AkXb70]

why not:
2.33GHz/333X7
2.5GHz /333X7.5
2.66GHz/333X8
2.83GHz/333X8.5
3.0GHz/333X9
(identical to amd's huh?)

given intels current 1MB/2MB lineup of all the same clock speed, a 2MB and 4MB version of each (as you have)

[coldpower27]

why not:
2.33GHz/333X7
2.5GHz /333X7.5
2.66GHz/333X8
2.83GHz/333X8.5
3.0GHz/333X9
(identical to amd's huh?)

given intels current 1MB/2MB lineup of all the same clock speed, a 2MB and 4MB version of each (as you have)

The only problem with that is that from what infromation I read the i965 Chipset doesn't have 1.33GHZ FSB Support. So I am going to assume a 1066FSB for Conroe/Allendale. Also Intel has already displaced the old Pentium 4 5xx Series with the 6xx Series as the 6xx is basically cheaper then the 5xx with the same clockspeed anyway.

I am going by that I rmemeber hearing multiple cache sizes for the desktop NGMA processor so 2MB/4MB seemed about right.

[AkXb70]

The only problem with that is that from what infromation I read the i965 Chipset doesn't have 1.33GHZ FSB Support. So i am going to assume a 1066FSB for Conroe/Allendale.
youre probably more right...1066 for base chips and 1333 for EE (aka 3.73EE and i925xe / i915 lacking 1066 support)

but i can still want a 1333 bus right?:D

[coldpower27]

youre probably more right...1066 for base chips and 1333 for EE (aka 3.73EE and i925xe / i915 lacking 1066 support)

but i can still want a 1333 bus right?:D

Sure, from what I gather the 45nm versions will have it by default :). Then we can probably look at 1 last Extreme Edition of 1.6GHZ FSB before intel finally ditiches the NetBurst FSB for the CSI in 2008.

[AkXb70]

it also makes sense from the perspective that intel wants headroom to up the fsb in the future. jumping the fsb would limit the advances that could be made in that area

[coldpower27]

it also makes sense from the perspective that intel wants headroom to up the fsb in the future. jumping the fsb would limit the advances that could be made in that area

Yes. I agree, it also may help keep power consumption in check.

[Carfax]

I would seriously hope that the only difference between the EE version and regular Conroe is the FSB. If thats the only difference, I sure as hell ain't paying over a thousand bucks for an EE processor.

There will more than likely be other differences though, to justify the price difference.

[RoyaL]

i hope it will come with 1066 fsb
u guys r crazy hoping for 1333
WTF higher fsb = lower multi

[idiotec]

I would seriously hope that the only difference between the EE version and regular Conroe is the FSB. If thats the only difference, I sure as hell ain't paying over a thousand bucks for an EE processor.

There will more than likely be other differences though, to justify the price difference.
What was the difference between the EE and other Prescotts?

Oh, higher FSB and double cache.... I am sure they'll do the same crap with Conroe. The EE's will always be a rip.......

[Pinnacle]

So, is it a for sure thing, that the Conroe WONT have HyperThreading?

Because that would really suck the left nut :slap: . I was looking forward to it. :(

[FunkyRider]

Conroe comes with 2 cores, that's 2 threads already. This will make a difference if you are from a "Solo" land.

[Carfax]

HT will be in Conroe, just not in the first revision. HT should be much better in Conroe than it was in the P4. The P4 didn't have enough execution units to truly take advantage of multithreading. Conroe will be super wide though, the widest x86 design to my knowledge and it will have more execution units and OoO capabilities to really utilize SMT.

[Carfax]

i hope it will come with 1066 fsb
u guys r crazy hoping for 1333
WTF higher fsb = lower multi

I don't get why so many people are against having a high FSB. WIth the prevalence of DDR2, achieving a high FSB isn't nearly as problematic as it was in the past with just regular DDR, having to use various ratios to limit the speed of the memory because it couldn't reach a higher FSB.

DDR2-800 is meant to be used in conjunction with a 1600 FSB! And DDR2-667 should be paired with 1333.

Whats the problem? And it's not like DDR2 can't overclock well either..

[IluvIntel]

iluvintel, less volts doesnt mean it runs cooler, ddr2 consumes more power than ddr1. pentium pro cpus used 3.3v vcore and consumes 10W, current pentium d cpus have a vcore of 1.35 and consume up to 150W.
vcore alone or vdimm alone doesnt show anything :)

I'd have to disagree with you there, If you touch a DDR2 module @ default voltage and touch a DDR module @ default v. you can tell the DDR2 is cooler to touch, thus less heat circulating in your case. ;)

[GoThr3k]

saaya i am sorry for you dude
dealing with nasgul isnt easy (not even in the movies)
i just wanna congratz you for you patience, and always replying to his post
you are a worthy admin :clap:

[coldpower27]

I would seriously hope that the only difference between the EE version and regular Conroe is the FSB. If thats the only difference, I sure as hell ain't paying over a thousand bucks for an EE processor.

There will more than likely be other differences though, to justify the price difference.

Well my guess is the extra 66MHZ to bring it to 3GHZ, and like the Dual Core Extreme Editions you should have an unlocked upper multiplier similar to the Athlon FX series.

[Donnie27]

Well my guess is the extra 66MHZ to bring it to 3GHZ, and like the Dual Core Extreme Editions you should have an unlocked upper multiplier similar to the Athlon FX series.

First of all, both i975 and i965 support 1333 FSB settings=P Several Websites talked about it and even Sentinal posted Screenshots of the BIOS showing that speed. Now, only the EE *may have official1333 support but there's no reason to believe Intel couldn't launch that speed for all 4MB versions and 1066MHz for all of the 2MB Single Core and 800MHz FSB for the 512K L2 Celeron versions. If Intel is even telling a half-truth, the Celeron version should be faster than Northwood. If Intel were going to, I'm sure they would have done the same thing they did with 800MHz.

Last but not least, the only things wrong with Hyperthreading the lack of Optimized software and what Intel charges for it on Dual Cores that don't need it.

TO: saaya

Unless I totally missed it, DDR2 uses less power and runs cooler than DDR1. DDR3 is supposed to be even better.

Donnie27

[coldpower27]

First of all, both i975 and i965 support 1333 FSB settings=P Several Websites talked about it and even Sentinal posted Screenshots of the BIOS showing that speed. Now, only the EE *may have official1333 support but there's no reason to believe Intel couldn't launch that speed for all 4MB versions and 1066MHz for all of the 2MB Single Core and 800MHz FSB for the 512K L2 Celeron versions. If Intel is even telling a half-truth, the Celeron version should be faster than Northwood. If Intel were going to, I'm sure they would have done the same thing they did with 800MHz.

Last but not least, the only things wrong with Hyperthreading the lack of Optimized software and what Intel charges for it on Dual Cores that don't need it.

TO: saaya

Unless I totally missed it, DDR2 uses less power and runs cooler than DDR1. DDR3 is supposed to be even better.

Donnie27

Well care to link some of those websites?, From what I read so far I only see 1.06GHZ FSB support, I did read about the 1.33GHZ FSB overide settings in BIOS for i975, but that only makes me think that the high end chipset will have 1.33GHZ FSB support which sites said the i965 will have 1.33GHZ FSB support?

In addition here are some roadmaps on the Broadwater i965 Chipset
http://tweakers.net/nieuws/40068

[Donnie27]

Well care to link some of those websites?, From what I read so far I only see 1.06GHZ FSB support, I did read about the 1.33GHZ FSB overide settings in BIOS for i975, but that only makes me think that the high end chipset will have 1.33GHZ FSB support which sites said the i965 will have 1.33GHZ FSB support?

In addition here are some roadmaps on the Broadwater i965 Chipset
http://tweakers.net/nieuws/40068

Yes and before it or anything else launches or ships, the features could easily change=P There might even be another i975EX like Intel did with 925/E and etc....

Speculation is all any of this is. As I said though, maybe it's for an High-End procs but the FSB override is there=P The news items are at overclockers.com, the Inq, hell, even at AMDtech, erum Anandtech. i965 will ship in about 4 flavors and a Cheaper i963 as the Inq posted.

http://www.theinquirer.net/?article=28602


Conroe will be be a new core built from the ground up. Basically it is a four issue wide 65nm Pentium M inspired chip. You can read about the details and not much has changed since then. The Achilles Heel of the chip, or heels as is the case here, are the front side bus (FSB) and the Integrated Memory Controller (IMC), or lack thereof. The FSB, 1333 for Conroe, is a bit creaky old, and slow. Add in the lack of an IMC, and memory latency becomes a sore spot. Both should be worlds ahead of the Pentium 4, but not up to the performance of the AMD A64.

Besides the AMD fan Crap-O-La, there is talk and some folks saying 1333MHz FSB is an option. Some speculate that Intel may not move to 1333 FSB until 45nm reached.


Moving to a lower and far less extreme chip, is the sucessor to Conroe called Ridgefield. This one looks to be the 45 nanometre shrink of Conroe. It's slated to bump the cache from 4MB to 6MB, and up the front side bus (FSB) to 1333MHz. Not Extreme, but tasty anyway. Let's just call it the Wild Strawberry Jell-O of Intel's 2007 lineup.

Donnie27

[saaya]

i didnt say anything about amd vs intel, is that a typo?:confused:
my bad :D
sorry


1333 FSB provides 10.8 GB/S of bandwidth, which should be enough I think. That would run in perfect sync with DDR2-667, which also provides 10.8 GB/S.

Overclock the FSB to 1600, which many of us will do, and you can pair Conroe up with DDR2-800 @ 1:1 ratio.

BTW, the archaic FSB architecture won't really matter to us, because the FSB can easily feed just two cores. More than two cores however, and AMD's method is much more efficient.
hmmmmm yonah has a system bus router inside the core, wich is half an onboard memory controller on die, its a little like amds x-crossbar wich manages the bandwidth between both cores...

atm intel is playing with dual fsb for high end servers, maybe they will intrdocue this for their desktop plattfors as well?
then its pretty much dual dual channel memory and even at lower speeds it would provide a lot of bandwidth.
two channels from the cpu to the chipset, and the chipset has 4 channels to the memory hmmm
this would mean you need at least 4 sticks of memory though for the full potential... hmmmm

so maybe they will have a dual channel between the cpu and chipset and a dual channel between the memory and the chipset?
hmmmm but then it wouldnt work in todays mainboards... so scratch what i said ^^


LOL

If you buy a Conroe with 2.66GHz, that's 266x10, you most likely to have a locked multi @ 10

If you want 1600 FSB, which is 400MHz real clock, I doubt Conroe can bust at whoppy 400x10 = 4GHz LOL. If it does, it will blow every CPU away on this planet
i think there have been northwoods at 10x400 iirc :D


Why does it have to be 266x10? If it uses 333 FSB, which it should, it would be 333x8. Bump that up to 400x8 and you've got 3.2ghz, which should be attainable given that it's on a 65nm process, uses very little power and has a medium length pipeline.

High end air coolers next year may be able to manage that kind of overclock, and good watercooling should definitely be able to.
i thougt conroe is only 65W?
so you woulsdnt need high end cooling...

hey coldpower27, welcome to XtremeSystems :toast:
thx for the infos :)
so you think amd will go for 266 mhz steps betwen their cpus now?
hmmm are you sure?
id rather think the memory speed would run 266 but the fsb would remain at 200, so they just use a different divider. because 266mhz steps would mean between 2ghz and 3ghz they could only have 4 different cpu models, actually only 3! so that would tighten their lineup quite a bit.
even with different cache sizes they would only have 8 different cpu models maximum then... maybe 12 if they go for a dual core sempron... but thats a tiny amount of chips compared to their current lineup.

conroe is 140mm²? hmmmm
thats around as big as a newcastle and northwood in 130nm...
130mm² ...thats not too big, i dont think they will be too expensive...
but conroe will cost more than dothan chips, yeah...
did you hear that conroe is 140mm² or did you calculate it? ^^
because it makes sense, but it rather have guessed conroe is more than 140mm²....
yonah is 96mm²... yonahs cache is ~40% of the core... so yonah with 4mb shared l2 cache would be 134.4m transistors...
but conroe is supposed to have ht and be a 4 issue core instead of 3 issue yonah... so either ht is already implemented, or yonah is already 4 issue but has a part of the logic disabled, maybe because of yield issues?
i cant imagine that intel can make the core 4 issue and add ht with adding just 10% more transistors.... hmmm
did you har about 140mm² somewhere or did you calculate it somehow like me? :D




The only problem with that is that from what infromation I read the i965 Chipset doesn't have 1.33GHZ FSB Support. So I am going to assume a 1066FSB for Conroe/Allendale. Also Intel has already displaced the old Pentium 4 5xx Series with the 6xx Series as the 6xx is basically cheaper then the 5xx with the same clockspeed anyway.

I am going by that I rmemeber hearing multiple cache sizes for the desktop NGMA processor so 2MB/4MB seemed about right.
6xx is cheaper to mfg than 5xx? but its larger? :confused:


Sure, from what I gather the 45nm versions will have it by default :). Then we can probably look at 1 last Extreme Edition of 1.6GHZ FSB before intel finally ditiches the NetBurst FSB for the CSI in 2008.
i still dont get why some people see conroe as a netburst cpu :confused:
intel officially said they dumped netburst months ago...
i have yet to see anybody explain why conroe would be netburst, or even netburst related :P


I would seriously hope that the only difference between the EE version and regular Conroe is the FSB. If thats the only difference, I sure as hell ain't paying over a thousand bucks for an EE processor.

There will more than likely be other differences though, to justify the price difference.
ht maybe... or larger cache?


HT will be in Conroe, just not in the first revision. HT should be much better in Conroe than it was in the P4. The P4 didn't have enough execution units to truly take advantage of multithreading. Conroe will be super wide though, the widest x86 design to my knowledge and it will have more execution units and OoO capabilities to really utilize SMT.
what do you mean with "it didnt have enough execution units to fully take advantage of multithreadding"?
can you explain what you mean or link me to something that explains this?
thx :)

carfax, people dont want such a high stock fsb because the only way to oc intel chips is through the fsb... :D


I'd have to disagree with you there, If you touch a DDR2 module @ default voltage and touch a DDR module @ default v. you can tell the DDR2 is cooler to touch, thus less heat circulating in your case. ;)

http://www.micron.com/products/dram/syscalc.html



saaya i am sorry for you dude
dealing with nasgul isnt easy (not even in the movies)
i just wanna congratz you for you patience, and always replying to his post
you are a worthy admin :clap:thx, i should have taken this to pm though to not mess this thread up :D
he hasnt replied to my pm though... but at least this thread is more calm now :)
oh and i removed that one word in you post, please be more polite even if you dont like him :)

coldpower27 yeah, ets hope intel finally sets unlocked multis for their ee chips to make them worth the money they cost.
a fully unlocked ee conroe would be a dream! :slobber:



TO: saaya

Unless I totally missed it, DDR2 uses less power and runs cooler than DDR1. DDR3 is supposed to be even better.

Donnie27

975 has 1333 support?
thats a guess your taking, 975 does NOT have official 1333 support!
if it does then link me plz =P

afaik the current boards dont, only a nw revision of them has 1333 support, the 975xe is what they will call it i guess... maybe the delay in 975x boards means intel will wait and tries to jump to 975xe right away or cancel 975xe and make all 975x boards 1333 capable? makes sense seeing as conroe comes in just a few months and at least some are 1333 :)

and according to micron and some tests ive seen and what several mainboard and memory makers told me at cebit almost a year ago, ddr2 does consume MORE than ddr1. but that may vary from chip to chip i guess... samsungs ddr1 and winbonds ddr1 are very power hungry for example...

http://www.micron.com/products/dram/syscalc.html

sdram power consumtion 444.5mw (per chip?)
ddr sdram power consumption 349.2mw (per chip?)
ddr2 sdram power consumption 365.4mw (per chip?)

its all 256mb sticks, ddr 400 and ddr2 400 iirc
since ddr2 clocks much higher than 400 effective speed it consumes even more power at those high speeds

sounds like theinq fell for the amd 2000mhz fsb marketing or at least enforces it as they call 1333fsb slow?
its by now means slow LOL :D

amd = 200fsb up to 266mhz aka ddr1 533 possible with dividers on the latest cpus and boards without ocing the fsb
intel = 266fsb up to 500mhz aka ddr2 1000 possible with dividers without ocing the fsb

amd am2 = (most likely) 200fsb and up to 500 mhz aka ddr2 1000 memory possible with dividers without ocing the fsb
so its more like amd will catch up with intels memory speeds!
they just have the advantage of beeing more efficient with the memory bandwidht with the integrated memory controler, so they didnt need the fast and high bandwidth memory before.

intel will move to 333fsb wich means at least 666mhz :devil: aka 1333mhz ddr2 memory speed, so this is not slow at all, its just a question of how well intel will use this massive bandwidth, how efficient the new onboard outer can distribute the bandwidth between both cores.

with the shared l2 cache wich reduces the bandwidth needed/makes the bandwidth usage more efficient, and the on die router, wich is half a memory controller on die if you want, PLUS the massive raw bandwidth of 1333 ddr2 i think intel will beat amd bandwidth wise.

so its kinda weird how theinq calls 1333 fsb slow ^^
or maybe it was meant ironic? :p:

the big question remains though, at what price will intel gain the top spot, at least memory bandwidth wise, back?
ddr2 1333mhz wont be cheap, the 1333fsb cpus wont be cheap either, and neither will the boards be cheap...

[Pinnacle]

my bad :D
sorry


hmmmmm yonah has a system bus router inside the core, wich is half an onboard memory controller on die, its a little like amds x-crossbar wich manages the bandwidth between both cores...

atm intel is playing with dual fsb for high end servers, maybe they will intrdocue this for their desktop plattfors as well?
then its pretty much dual dual channel memory and even at lower speeds it would provide a lot of bandwidth.
two channels from the cpu to the chipset, and the chipset has 4 channels to the memory hmmm
this would mean you need at least 4 sticks of memory though for the full potential... hmmmm

so maybe they will have a dual channel between the cpu and chipset and a dual channel between the memory and the chipset?
hmmmm but then it wouldnt work in todays mainboards... so scratch what i said ^^


i think there have been northwoods at 10x400 iirc :D


i thougt conroe is only 65W?
so you woulsdnt need high end cooling...

hey coldpower27, welcome to XtremeSystems :toast:
thx for the infos :)
so you think amd will go for 266 mhz steps betwen their cpus now?
hmmm are you sure?
id rather think the memory speed would run 266 but the fsb would remain at 200, so they just use a different divider. because 266mhz steps would mean between 2ghz and 3ghz they could only have 4 different cpu models, actually only 3! so that would tighten their lineup quite a bit.
even with different cache sizes they would only have 8 different cpu models maximum then... maybe 12 if they go for a dual core sempron... but thats a tiny amount of chips compared to their current lineup.

conroe is 140mm²? hmmmm
thats around as big as a newcastle and northwood in 130nm...
130mm² ...thats not too big, i dont think they will be too expensive...
but conroe will cost more than dothan chips, yeah...
did you hear that conroe is 140mm² or did you calculate it? ^^
because it makes sense, but it rather have guessed conroe is more than 140mm²....
yonah is 96mm²... yonahs cache is ~40% of the core... so yonah with 4mb shared l2 cache would be 134.4m transistors...
but conroe is supposed to have ht and be a 4 issue core instead of 3 issue yonah... so either ht is already implemented, or yonah is already 4 issue but has a part of the logic disabled, maybe because of yield issues?
i cant imagine that intel can make the core 4 issue and add ht with adding just 10% more transistors.... hmmm
did you har about 140mm² somewhere or did you calculate it somehow like me? :D




6xx is cheaper to mfg than 5xx? but its larger? :confused:


i still dont get why some people see conroe as a netburst cpu :confused:
intel officially said they dumped netburst months ago...
i have yet to see anybody explain why conroe would be netburst, or even netburst related :P


ht maybe... or larger cache?


what do you mean with "it didnt have enough execution units to fully take advantage of multithreadding"?
can you explain what you mean or link me to something that explains this?
thx :)

carfax, people dont want such a high stock fsb because the only way to oc intel chips is through the fsb... :D



http://www.micron.com/products/dram/syscalc.html


thx, i should have taken this to pm though to not mess this thread up :D
he hasnt replied to my pm though... but at least this thread is more calm now :)
oh and i removed that one word in you post, please be more polite even if you dont like him :)

coldpower27 yeah, ets hope intel finally sets unlocked multis for their ee chips to make them worth the money they cost.
a fully unlocked ee conroe would be a dream! :slobber:



975 has 1333 support?
thats a guess your taking, 975 does NOT have official 1333 support!
if it does then link me plz =P

afaik the current boards dont, only a nw revision of them has 1333 support, the 975xe is what they will call it i guess... maybe the delay in 975x boards means intel will wait and tries to jump to 975xe right away or cancel 975xe and make all 975x boards 1333 capable? makes sense seeing as conroe comes in just a few months and at least some are 1333 :)

and according to micron and some tests ive seen and what several mainboard and memory makers told me at cebit almost a year ago, ddr2 does consume MORE than ddr1. but that may vary from chip to chip i guess... samsungs ddr1 and winbonds ddr1 are very power hungry for example...

http://www.micron.com/products/dram/syscalc.html

sdram power consumtion 444.5mw (per chip?)
ddr sdram power consumption 349.2mw (per chip?)
ddr2 sdram power consumption 365.4mw (per chip?)

its all 256mb sticks, ddr 400 and ddr2 400 iirc
since ddr2 clocks much higher than 400 effective speed it consumes even more power at those high speeds

sounds like theinq fell for the amd 2000mhz fsb marketing or at least enforces it as they call 1333fsb slow?
its by now means slow LOL :D

amd = 200fsb up to 266mhz aka ddr1 533 possible with dividers on the latest cpus and boards without ocing the fsb
intel = 266fsb up to 500mhz aka ddr2 1000 possible with dividers without ocing the fsb

amd am2 = (most likely) 200fsb and up to 500 mhz aka ddr2 1000 memory possible with dividers without ocing the fsb
so its more like amd will catch up with intels memory speeds!
they just have the advantage of beeing more efficient with the memory bandwidht with the integrated memory controler, so they didnt need the fast and high bandwidth memory before.

intel will move to 333fsb wich means at least 666mhz :devil: aka 1333mhz ddr2 memory speed, so this is not slow at all, its just a question of how well intel will use this massive bandwidth, how efficient the new onboard outer can distribute the bandwidth between both cores.

with the shared l2 cache wich reduces the bandwidth needed/makes the bandwidth usage more efficient, and the on die router, wich is half a memory controller on die if you want, PLUS the massive raw bandwidth of 1333 ddr2 i think intel will beat amd bandwidth wise.

so its kinda weird how theinq calls 1333 fsb slow ^^
or maybe it was meant ironic? :p:

the big question remains though, at what price will intel gain the top spot, at least memory bandwidth wise, back?
ddr2 1333mhz wont be cheap, the 1333fsb cpus wont be cheap either, and neither will the boards be cheap...


OK, this is Ridiculuos

moderation sayaa, MODERATION. Too much info to grasp HEHE :D

We know your a smart fellow, but make it your New Years Resolution to write in smaller chuncks

;)

[NiCKE^]

Come on saaya, stop writing a essay as answer to each one that wants to argue please ;)

I'm thrilled about the Conroe release and hope we will see many samples in the hands of XS members as soon as possible!

[Who]

Someone found a roadmap: http://www.xtremesystems.org/forums/showthread.php?t=84866

[NiCKE^]

Interesting, i985X, never heard of that chipset before.

[AkXb70]

Someone found a roadmap: http://www.xtremesystems.org/forums/showthread.php?t=84866
that roadmap is screwy...

[FunkyRider]

http://img4.picsplace.to/img4/17/IDF_2005_cpus.jpg

That's what I'm talking about man... If some one can do a little Photoshop trick, it shouldn't be hard to "estimate" the core area of Conroe. I will do it soon anyway LOL

[FunkyRider]

Just finish, hope this "very rough estimate" isn't too bad

http://img5.picsplace.to/img5/16/conroe_die.gif

[afireinside]

Is 965 the little brother of 975 or is it faster? This is really confusing me...

[coldpower27]

my bad :D
sorry


hmmmmm yonah has a system bus router inside the core, wich is half an onboard memory controller on die, its a little like amds x-crossbar wich manages the bandwidth between both cores...

atm intel is playing with dual fsb for high end servers, maybe they will intrdocue this for their desktop plattfors as well?
then its pretty much dual dual channel memory and even at lower speeds it would provide a lot of bandwidth.
two channels from the cpu to the chipset, and the chipset has 4 channels to the memory hmmm
this would mean you need at least 4 sticks of memory though for the full potential... hmmmm

so maybe they will have a dual channel between the cpu and chipset and a dual channel between the memory and the chipset?
hmmmm but then it wouldnt work in todays mainboards... so scratch what i said ^^


i think there have been northwoods at 10x400 iirc :D


i thougt conroe is only 65W?
so you woulsdnt need high end cooling...

hey coldpower27, welcome to XtremeSystems :toast:
thx for the infos :)
so you think amd will go for 266 mhz steps betwen their cpus now?
hmmm are you sure?
id rather think the memory speed would run 266 but the fsb would remain at 200, so they just use a different divider. because 266mhz steps would mean between 2ghz and 3ghz they could only have 4 different cpu models, actually only 3! so that would tighten their lineup quite a bit.
even with different cache sizes they would only have 8 different cpu models maximum then... maybe 12 if they go for a dual core sempron... but thats a tiny amount of chips compared to their current lineup.

conroe is 140mm²? hmmmm
thats around as big as a newcastle and northwood in 130nm...
130mm² ...thats not too big, i dont think they will be too expensive...
but conroe will cost more than dothan chips, yeah...
did you hear that conroe is 140mm² or did you calculate it? ^^
because it makes sense, but it rather have guessed conroe is more than 140mm²....
yonah is 96mm²... yonahs cache is ~40% of the core... so yonah with 4mb shared l2 cache would be 134.4m transistors...
but conroe is supposed to have ht and be a 4 issue core instead of 3 issue yonah... so either ht is already implemented, or yonah is already 4 issue but has a part of the logic disabled, maybe because of yield issues?
i cant imagine that intel can make the core 4 issue and add ht with adding just 10% more transistors.... hmmm
did you har about 140mm² somewhere or did you calculate it somehow like me? :D




6xx is cheaper to mfg than 5xx? but its larger? :confused:


i still dont get why some people see conroe as a netburst cpu :confused:
intel officially said they dumped netburst months ago...
i have yet to see anybody explain why conroe would be netburst, or even netburst related :P


ht maybe... or larger cache?


what do you mean with "it didnt have enough execution units to fully take advantage of multithreadding"?
can you explain what you mean or link me to something that explains this?
thx :)

carfax, people dont want such a high stock fsb because the only way to oc intel chips is through the fsb... :D



http://www.micron.com/products/dram/syscalc.html


thx, i should have taken this to pm though to not mess this thread up :D
he hasnt replied to my pm though... but at least this thread is more calm now :)
oh and i removed that one word in you post, please be more polite even if you dont like him :)

coldpower27 yeah, ets hope intel finally sets unlocked multis for their ee chips to make them worth the money they cost.
a fully unlocked ee conroe would be a dream! :slobber:



975 has 1333 support?
thats a guess your taking, 975 does NOT have official 1333 support!
if it does then link me plz =P

afaik the current boards dont, only a nw revision of them has 1333 support, the 975xe is what they will call it i guess... maybe the delay in 975x boards means intel will wait and tries to jump to 975xe right away or cancel 975xe and make all 975x boards 1333 capable? makes sense seeing as conroe comes in just a few months and at least some are 1333 :)

and according to micron and some tests ive seen and what several mainboard and memory makers told me at cebit almost a year ago, ddr2 does consume MORE than ddr1. but that may vary from chip to chip i guess... samsungs ddr1 and winbonds ddr1 are very power hungry for example...

http://www.micron.com/products/dram/syscalc.html

sdram power consumtion 444.5mw (per chip?)
ddr sdram power consumption 349.2mw (per chip?)
ddr2 sdram power consumption 365.4mw (per chip?)

its all 256mb sticks, ddr 400 and ddr2 400 iirc
since ddr2 clocks much higher than 400 effective speed it consumes even more power at those high speeds

sounds like theinq fell for the amd 2000mhz fsb marketing or at least enforces it as they call 1333fsb slow?
its by now means slow LOL :D

amd = 200fsb up to 266mhz aka ddr1 533 possible with dividers on the latest cpus and boards without ocing the fsb
intel = 266fsb up to 500mhz aka ddr2 1000 possible with dividers without ocing the fsb

amd am2 = (most likely) 200fsb and up to 500 mhz aka ddr2 1000 memory possible with dividers without ocing the fsb
so its more like amd will catch up with intels memory speeds!
they just have the advantage of beeing more efficient with the memory bandwidht with the integrated memory controler, so they didnt need the fast and high bandwidth memory before.

intel will move to 333fsb wich means at least 666mhz :devil: aka 1333mhz ddr2 memory speed, so this is not slow at all, its just a question of how well intel will use this massive bandwidth, how efficient the new onboard outer can distribute the bandwidth between both cores.

with the shared l2 cache wich reduces the bandwidth needed/makes the bandwidth usage more efficient, and the on die router, wich is half a memory controller on die if you want, PLUS the massive raw bandwidth of 1333 ddr2 i think intel will beat amd bandwidth wise.

so its kinda weird how theinq calls 1333 fsb slow ^^
or maybe it was meant ironic? :p:

the big question remains though, at what price will intel gain the top spot, at least memory bandwidth wise, back?
ddr2 1333mhz wont be cheap, the 1333fsb cpus wont be cheap either, and neither will the boards be cheap...

Thank you for the welcome.

In regards to the the AMD steps, I am sorry if I gave you information to draw the wrong conclusion, I actually think AMD will be going to 333MHZ steps rather then 266MHZ steps, so hence why I don't think Intel will need half multiplier. Thereason as such is because AMD's highest DDR2 support with M2 will be 667MHZ or at least that is what we think anyway.

Athlon 64x2 2.0GHZ/2x512KB
Athlon 64x2 2.33GHZ/2x512KB
Athlon 64x2 2.33GHZ/2x1MB
Athlon 64x2 2.667GHZ/2x512KB

Athlon FX 60 AM2 2.667GHZ/2x1MB

Eventaully they will also have the 3GHZ version so that adds to the lineup to 7 chips.

AMD doesn't need that many Dual Core, but we will see.

Well from a picture I saw regarding Conroe though, now that I have heard Presler is about 162mm2, from Anandtech and Cedar Mill is 81mm2, that means. Conroe should be at or below 160mm2 probably in the 140mm2 to 160mm2 range with 4MB of LV2 cache, there is also another version planned with lower cache called Allendale, with 2MB LV2 cache so that will have a smaller die as well. I actually now think that Conroe is probably more then 140mm2 now that I know what Preslers size is, but my estimates were based on the information prior that Presler was 140mm2.

From what I have read Yonah is 90.3mm2 this is from Intel slides themselves. So Yonah isn't that large but Yonah being a first generation 65nm product seems to have quite a low transistor density for a 65nm processor, I would assume that a more mature 65nm product would have better density somewhere like Cedar Mill's level of 2.3 Mil per mm2.

Well, The only possible sense of leaving Prescott-1M in the production line now, is for legacy support, as Prescott-2M is a better processor to produce for Intel as it has lower power consumption thatnks to the mildly increased die size to 135mm2, but I think Intel only left Prescott-1M because of how Smithfield is produced. You can salvage a Prescott from a Smithfield if one core doesn't work. Not only that but the pricing structure of Prescott-2M would suggest Intel wants you to buy those processor instead of Prescott. As they are basically the same price to the consumer so from a consumer standpoint it makes no sense to buy Prescott-1M at all.

Sorry, that wasn't what I said. I don't see Conroe as a NetBurst processor at all. I said Intel's NGMA (Conroe) will switch over to CSI in 2008. For the moment it will be using NetBurst based FSB technology. We have seen that Conroe fits into LGA775 socket, so it basically confirms the use of the NetBurst based FSB technology.

Conroe is based on the performance per watt philosophy, and is a high IPC architecture. It has some elements pioneered by the NetBurst architecture.
Namely, the FSB technology and EM64T, I only call it the NetBurst FSB as it was first used on Pentium 4 Willamette for the FSB, and Prescott for EM64T.
It will also have some elements of the Pentium M architecture, like the more integrated nature to work together like in Yonah, as well as the shared cache architecture pioneered in Yonah. The new architecture will also introduce some new improvements of it's own like L1 to L1 links, as well as macro-ops fusion.

I also beleive high IPC architectures aren't really starved for memroy bandwidth, so for the moment 1066FSB should be sufficient where Dual Channel DD2-533 is enough to supply the bandwidth. I also doubt that Dual Independent Bus is going to be implemented as that would basically need Quad Channel DDR which is much to expensive. If 1333FSB is used Dual Channel DDR2-667 will be sufficient.

To me though the HyperTransport 2GHZ number doesn't mean anything in regards to memory bandwidth, actual memory bandwidth will only be 10.5GB/s with Dual Channel DDR2-667 on Socket AM2 processors. It's purely dependent on what the memory is running at and what config rather then what HyperTransport number AMD displays. That number is for I/O bandwdith only, with 2GHZ representing the bi-directional nature of the link. It's actually 1GHZ each way.

From what I have generally seen Intel's efficiency on it's off die memory controller is in the low 70%'s for memory bandwidth, but latency is quite alot higher.

[coldpower27]

Just finish, hope this "very rough estimate" isn't too bad

http://img5.picsplace.to/img5/16/conroe_die.gif

Only problem with this is that Prescott-2M is 135mm2. Rather then Prescott-1M's 112mm2.

[FunkyRider]

Oops, I must made some mistakes, in the die photo it was actually marked the Prescott core as "90nm Pentium 4", so this must be Prescott-1M, not 2M, my bad :D

[saaya]

OK, this is Ridiculuos

moderation sayaa, MODERATION. Too much info to grasp HEHE :D

We know your a smart fellow, but make it your New Years Resolution to write in smaller chuncks

;)

ok, ill try to shorten every post after i finsihed writing it from now on :D


Come on saaya, stop writing a essay as answer to each one that wants to argue please ;)i cant resist :D
i have no doubt there will be conroes on xs before they launch :D

FunkyRider, your wrong, 112mm² is the size of the 1m prescott
the 2m prescott is 135mm²

the package is 37.5 x 37.5mm, so id guess that its pretty much exactly 10mm wide and pretty much exactly 14mm long, so its around 140mm² like mentioned earlier... hmmmm
yonahs logic is around 60mm²
if the cache in conroe is as dense as in yonah and the ache architecture is the same, e they just doubled the cache array, wich is what it sounds like, then thise means intel only has 6mm² extra logic in conroe compared to yonah, wich is 10%.
thats as much transistors you need for ht alone, but it now seems conroe doent have ht, so 10% for turning the 3way yonah into a 4way conroe sounds about right... they really squeezed the logic there, but it sounds about right :D

well pressler isnt really 160mm², id say its 2x cedar mill wich is 81mm... but thats a different thing, bigger size means worse yields, since pressler is made of two independant cores its not the same as one 160mm² core and will cost a lot less to manufacture tha a 160mm² die :D
wich is a smart move by intel, no doubt!
i asked amd why they didnt do this and they said it doesnt work for them because this increases the package prices, plus it increases the latency from one core to the other, since amd has the imc this would be deadly i suppose

and yeah, conroe with only 140mm² is weird isnt it?
ive seen a yonah with my own eyes though :D
and its smaller than the conroe core on that pic, and i doubt they added more logic to conroe later on... hmmmm
making a second conroe die with less cache doesnt make sense either though... making a conroe with less cache first and then adding more cache doesnt make sense either, as they copied the cache block from yonah, making each block smaller and then making it bigger again doesnt make sense... and only using one cache block would mean they need to change how they arrange the logi in each conroe with more and less cache, so that doesnt make sense either...

i think conroe will only come as a 4mb design and some chips will have half the cache damaged or disabled... and i think conroe is really only 140mm² and that the pic shows the final conroe chip.

wich is some damn good logic tweaking done by intel :clap:

coldpower, i guess thats what intel meant with low power 65nm, they didnt put the transistors that tight wich means less hot spots and less leakage/heat...
hmmmm but yonah is up to ~60W already, and thats 2.4ghz+... conroe is supposed to be only 85W max wich must be the 3200mhz one... so conroe is consuming less power than yonah!?!
hmmmmmm now thats interesting.... hmmmm
2ghz yonah is supposed to consume 35W max iirc, the desktop 2.4ghz chip is above 55W, so around 60W
well i guess they are shoving a pretty high vcore up yonahs 4ss to make it run at 2.4ghz and higher...
i guess conroe clocks higher because of some tweaks and needs less vcore so it can run 3200mhz and only consume 85W max

about prescott 2m, its not running cooler because of the larger die afaik, it needs less volts and is tweaked overall to run cooler...
im not sure if smithfield chips can be split into seperate prescott 1m cores...
i guess yes, but then this brings up some problems...
as anandtech revelead a few days ago smithfield actually is a prescott 2m with cut l2 cache, and not a double prescott 1m as everybody thougt before...

smithfields cache has the same high latency as the 2m prescotts...
so a 5xx based on a dead or cut in half smithfield would perform notably slower than a "real" prescot 1m!
unless intel chan change the cache latency... but if they could why would they set smithfield with such a high cache latency even thouzgh it has a rather low cache capacity of just 1mb? so i suppose they cant.

and yes, conroe uses the qpb, but does it make it a netburst cpu? :D
and remember, the difference between yonah and conroes logic is only around 10%, and its not really much new but just some units got doubled.
and the cache design is basically identical as well.
and it now looks theres no ht.
so if you ask me conroe is not netburst at all... not a tiny bit
macro ops fusion is already present in dothan afaik and the cpus interface doesnt define the architecture its based on if you ask me... doesnt the itanium use the qpb as well?

ad yeah, that amd 2ghz fsb was really going on my nerves, im glad almost everybody ignored it.
and what upset me even more is the stupid "with hyper transport technology" listen on the amd boxed cpus, wich made people think this is like intels hyper threadding technology, wich is something completely different...
hyper transport technology... like thats a feature of a cpu wich makes it better pffff
to some degree, yes, it gets rid of the fsb bottleneck, but that amd marketing bs is realy misleading

[flakpyro]

hmmmm but yonah is up to ~60W already, and thats 2.4ghz+


There is a 2.4Ghz Yonah? Either im understanding you wrong or my head must be under a rock! :slobber:

[ingentingmendeg]

got a quick question, will conroe be compatible with 955x chipsets?

[Pinnacle]

i asked amd why they didnt do this and they said it doesnt work for them because this increases the package prices, plus it increases the latency from one core to the other, since amd has the imc this would be deadly i suppose


You talk to AMD? You got some inside connections :p:

[ingentingmendeg]

i thought they stopped making the P4, no?

[ingentingmendeg]

dont know if anyone already saw this but its interesting

http://images.google.com/imgres?imgurl=http://www.bit-tech.net/news_images/intel_processors_conroe/article_img.jpg&imgrefurl=http://www.bit-tech.net/news/2005/08/23/intel_processors_conroe/&h=250&w=300&sz=23&tbnid=HWvQMqD9DGkJ:&tbnh=92&tbnw=111&hl=en&start=4&prev=/images%3Fq%3Dintel%2Bconroe%26svnum%3D10%26hl%3Den %26lr%3D%26client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:official_s%26sa%3DG

The bottom line? Next year, Intel should have a kick ass chip that should be fantastic for gaming, whilst keeping power and heat to a minimum. Can I get a 'w00t'?

[AkXb70]

i thought they stopped making the P4, no?
cedar mill is a 65nm p4 (pressie 2m)....due out relatively soon, and the 9xx chips are 2 cedar mills, so no, they havnt stopped

[coldpower27]

ok, ill try to shorten every post after i finsihed writing it from now on :D

i cant resist :D
i have no doubt there will be conroes on xs before they launch :D

FunkyRider, your wrong, 112mm² is the size of the 1m prescott
the 2m prescott is 135mm²

the package is 37.5 x 37.5mm, so id guess that its pretty much exactly 10mm wide and pretty much exactly 14mm long, so its around 140mm² like mentioned earlier... hmmmm
yonahs logic is around 60mm²
if the cache in conroe is as dense as in yonah and the ache architecture is the same, e they just doubled the cache array, wich is what it sounds like, then thise means intel only has 6mm² extra logic in conroe compared to yonah, wich is 10%.
thats as much transistors you need for ht alone, but it now seems conroe doent have ht, so 10% for turning the 3way yonah into a 4way conroe sounds about right... they really squeezed the logic there, but it sounds about right :D

well pressler isnt really 160mm², id say its 2x cedar mill wich is 81mm... but thats a different thing, bigger size means worse yields, since pressler is made of two independant cores its not the same as one 160mm² core and will cost a lot less to manufacture tha a 160mm² die :D
wich is a smart move by intel, no doubt!
i asked amd why they didnt do this and they said it doesnt work for them because this increases the package prices, plus it increases the latency from one core to the other, since amd has the imc this would be deadly i suppose

and yeah, conroe with only 140mm² is weird isnt it?
ive seen a yonah with my own eyes though :D
and its smaller than the conroe core on that pic, and i doubt they added more logic to conroe later on... hmmmm
making a second conroe die with less cache doesnt make sense either though... making a conroe with less cache first and then adding more cache doesnt make sense either, as they copied the cache block from yonah, making each block smaller and then making it bigger again doesnt make sense... and only using one cache block would mean they need to change how they arrange the logi in each conroe with more and less cache, so that doesnt make sense either...

i think conroe will only come as a 4mb design and some chips will have half the cache damaged or disabled... and i think conroe is really only 140mm² and that the pic shows the final conroe chip.

wich is some damn good logic tweaking done by intel :clap:

coldpower, i guess thats what intel meant with low power 65nm, they didnt put the transistors that tight wich means less hot spots and less leakage/heat...
hmmmm but yonah is up to ~60W already, and thats 2.4ghz+... conroe is supposed to be only 85W max wich must be the 3200mhz one... so conroe is consuming less power than yonah!?!
hmmmmmm now thats interesting.... hmmmm
2ghz yonah is supposed to consume 35W max iirc, the desktop 2.4ghz chip is above 55W, so around 60W
well i guess they are shoving a pretty high vcore up yonahs 4ss to make it run at 2.4ghz and higher...
i guess conroe clocks higher because of some tweaks and needs less vcore so it can run 3200mhz and only consume 85W max

about prescott 2m, its not running cooler because of the larger die afaik, it needs less volts and is tweaked overall to run cooler...
im not sure if smithfield chips can be split into seperate prescott 1m cores...
i guess yes, but then this brings up some problems...
as anandtech revelead a few days ago smithfield actually is a prescott 2m with cut l2 cache, and not a double prescott 1m as everybody thougt before...

smithfields cache has the same high latency as the 2m prescotts...
so a 5xx based on a dead or cut in half smithfield would perform notably slower than a "real" prescot 1m!
unless intel chan change the cache latency... but if they could why would they set smithfield with such a high cache latency even thouzgh it has a rather low cache capacity of just 1mb? so i suppose they cant.

and yes, conroe uses the qpb, but does it make it a netburst cpu? :D
and remember, the difference between yonah and conroes logic is only around 10%, and its not really much new but just some units got doubled.
and the cache design is basically identical as well.
and it now looks theres no ht.
so if you ask me conroe is not netburst at all... not a tiny bit
macro ops fusion is already present in dothan afaik and the cpus interface doesnt define the architecture its based on if you ask me... doesnt the itanium use the qpb as well?

ad yeah, that amd 2ghz fsb was really going on my nerves, im glad almost everybody ignored it.
and what upset me even more is the stupid "with hyper transport technology" listen on the amd boxed cpus, wich made people think this is like intels hyper threadding technology, wich is something completely different...
hyper transport technology... like thats a feature of a cpu wich makes it better pffff
to some degree, yes, it gets rid of the fsb bottleneck, but that amd marketing bs is realy misleading

Anandtech has revealed that Smithfield's LV2 cache has the same latency as Presler/Cedar Mill/Prescott-2M, nothing more, this doesn't mean the cores are based on 2 cut down Prescott-2M processors at all. Slowing down the cache would be the logical solution as Smithfield is still 2 Prescott-1M with arbiter logic. The slowing down of the cache should help ensure good yields and power consumption. Since the Prescott-1M cores, are designed to run at the lower latency in single core configuration, it shouldn't be a problem, and in single core configuration, power consumption can be kept in check.

There is plenty of sense in creating another die with reduced cache size, Allendale with 2MB of LV2 cache will be a smaller product and hence higher yielding and easier to make for Intel, AMD does this too, as they created Toledo first, with the 512KB parts using damaged, disabled cores, when yeilds improved they introduced the Manchester core to get more usable cores per wafer. Intel needs as much 65nm production capacity as it can get, as it needs to produce Yonah, Merom, and Cedar Mill, Presler along side the Conroe/Allendales. NetBurst production will still continue with the arrival of Conroe. From what we have seen so far shared cache isn't that densely packed in comparison to the older cache of Dothan on the 90nm node, As with reduced cost processors, this will give Intel the ability to target lower price points or just make more money. You also have to keep in mind Merom will be made with 2MB of shared cache, so some of the work has already been done on a 2MB NGMA derivative.

The 10% figure you just came up with, interesting but I don't see it, if by using your 60mm2 figure for the cache on Yonah then 30mm2 ~ represents the 2 cores on Yonah, so assuming 120mm2 will allow for 4MB cache on Conroe, this assumes the cache density is the same, you will probably also have 25-40mm2 for the 2 cores themsleves depending on how complex and how much more space they take up on Conroe. I would expect the improvement to take more space on the core then less though. You also have to keep in mind transistor budget for EM64T, done well not as an afterthought on like on the NetBurst architecture.

Dothan has <b> micro-ops fusion</b>, macro-ops fusion doens't exist on Dothan, what is new that will be introduced in Merom is something called <b> macro-ops fusion </b>, plus you have to remember also the direct L1 to L1 cache links between the two cores this architecture will introduce. Not to mention the move to a 4 issue wide, the last that time we went to higher issue cores was done was when the Pentium Pro core was introduced which moved from 2 to 3. That's a huge assumption that they copied the cache archtiecture completely of Yonah, it would make more sense to improve the cache for Merom/Conroe as Yonah was a test run.

http://images.anandtech.com/reviews/tradeshows/IDF/2005/Fall/Day1/NewArch/IMG_2637.jpg
http://images.anandtech.com/reviews/tradeshows/IDF/2005/Fall/Day1/NewArch/IMG_2638.jpg
http://images.anandtech.com/reviews/tradeshows/IDF/2005/Fall/Day1/NewArch/IMG_2639.jpg

And once again, I am saying the Conroe will have some elements of NetBurst included in it's designed such as EM64T pioneered by Prescott, as well as FSB technology pioneered by Willamette, as well as the branch prediction work put into the this processor. To my knowledge Itanium 2's have 400 MHZ FSB or higher up to 667 MHz, evne still this was introudced after Pentium 4's. Itanium 1 though only has 100 FSB so not that, but Itanium does have some X86 elements within it's design to retain legacy compatability. From Pentium M architecture (Banias and it's derivatives Dothan and Yonah), we get the power optimizations, shared cache architecture, low pipeline length. The FSB technology was developed for NetBurst to synch with PC800 RDRAM at the time.

I am not saying it's a NetBurst processor at all. Intel isn't about to throw away all the work done to NetBurst, some elements can still be used in this new architecture. There is nothing wrong with saying this processor has some of the best features of NetBurst, it doesn't mean it's going to suck. I beleive the FSB is quite an important part of an architecture.

Yonah EE is most likely going to be 2.5GHZ or beyond as Yonah is slated for an update to 2.33 GHZ in Q2 2006, it also is in the E TDP which is > 50W category so it's beyond 50W, unlike plain Yonah which is on the T 25to49W category with Yonah own TDP being 31W.

Well from what I remember, Merom 35W, Conroe 65W, Woodcrest 80W.

AMD can't really use the Dual Die design on anything due to how their Dual Cores are designed it would hurt performance NetBurst Dual Cores are fine as they communicate through the FSB anyway, only reason Smithfield was made in monolithic die format was to get it to market quickly. Intel might still be able to go the Dual Die route as long as they keep the memory controller in the North Bridge.

I am not going to go by that roadmap religiously as anyone can produce one of those Microsoft Word table charts and the clockspeeds seem a bit high for something that has better performance per clock then even Yonah Pentium M.
3.2 GHZ seems much too enthusiastic to me I was thinking more along the lines of ~ 3 GHZ tops. Not to mention the error on Presler cache size.

Oh yeah FunkyRider here is a link for Conroe in realtion to Presler
http://www.anandtech.com/tradeshows/showdoc.aspx?i=2505

Could you pehaps do some work in photoshop to compare the die sizes if possible, normalizing it with 81mm2 for each of the Presler dual dies. Do it also with the photo poste in this thread as well as you can then get a different perspective and a 2nd estimate.

[saaya]

There is a 2.4Ghz Yonah? Either im understanding you wrong or my head must be under a rock! :slobber:
there will be a 2.4ghz yonah, or was it 2.33?

got a quick question, will conroe be compatible with 955x chipsets?in theory... yes

You talk to AMD? You got some inside connections :p:that chat was almost a year ago at cebit with some amd guys :D

ingentingmendeg, nope, intel still makes p4s "en masse" for dell etc...


dont know if anyone already saw this but its interesting

http://images.google.com/imgres?imgurl=http://www.bit-tech.net/news_images/intel_processors_conroe/article_img.jpg&imgrefurl=http://www.bit-tech.net/news/2005/08/23/intel_processors_conroe/&h=250&w=300&sz=23&tbnid=HWvQMqD9DGkJ:&tbnh=92&tbnw=111&hl=en&start=4&prev=/images%3Fq%3Dintel%2Bconroe%26svnum%3D10%26hl%3Den %26lr%3D%26client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:official_s%26sa%3DG

The bottom line? Next year, Intel should have a kick ass chip that should be fantastic for gaming, whilst keeping power and heat to a minimum. Can I get a 'w00t'?
those graphs are very misleading...
intel seems to like to compare its future and latest products to stuff thats already out for years and then say its 10x faster :P
on that graph dothan is 50% faster than banias per watt... wich is not true afaik. banias and dothan are around the same wattage and dothan is only around 10-30% faster, in some apps theres no difference at all.
they say integer performence, doesnt sandra test integer performence?
i ran sandra on my banias and a dothan and they scored exactly the same... in my exprieince dothan is only faster in apps that need a lot of system bandwidth hence they scale with large l2 cache wich saves system bandwidth...

coldpower, how does slower l2 cache increase yields?
and it reducs power consumption? i dont think that cache has a big impact on power consumption, its just a few watts, 5W for 1mb at most, afaik cache is always the cool part of a cpu.
and considering that l2 cache to l2 cache on smithfield is creating a huge bottleneck for smithfield it would make more sense to go for faster cache i think.
think about it, intel needed a dual core cpu fast, so why would they go for the prescott1m to base it on that if they already had a more advanced prescott 2m chip?
it had more tweaks in the logic and cache than the prescott 1m, the only draw back was larger cache, but they can cut off cache just like that without a redesign, so i think it all points to intel basing smithfield on prescott 2m, all they did for pressler now was to shrink prescott 2m to 65nm and put two of them on one extra substrate to get btter dual core yields.
thats what i think at least.

hahah gotcha! was thinking the same thing about merom beeing the 2mb chip :D
merom will be identical with conroe afaik, so mabybe the merom will be 2mb cache, because in laptops the 5W or 10W extra cache needs is not worth the little performence gans, especially since laptops dont need that much performence.
so maybe there will be 2 chips and the 2mb conroe will be identical with merom?

well about amds dual cres, yes, but those are gigantic cores, and the ones with less cache are almost half in size, amd cant reach as high cache densities as intel, not even close.
for intel conroe with 4mb and conroe with 2mb is just a difference in size of ~110 vs ~140, wich i dont think is worth it that much. thats a die shrink of a mere 20% compared to amds 50%+ for the different dual core chips.

cache density on yonah is worse than on dothan? hmmmm

dothans die size is around 83.6mm&#178;
57% of dothan is l2 cache, thats 47.7mm&#178;
2mb on 47.7mm&#178;

yonah is 96mm&#178;
only around 40% of yonah is l2 cache, thats 38,4mm&#178;
2mb on 38,4mm&#178;

so the cache on yonah is more dense, but only 19.5%, wich is not much considering yonah is made in 65nm vs dothan in 90nm.
65nm should actually result in a 48% more dense cache if the density of the transistors would remain the same....
interesting!
thx for the pointer :toast:

pressler is 81mm&#178;
the only time ive seen a die shot of pessler is at intels idf, and it was tiny...
http://img317.imageshack.us/img317/313/blownuppressler7aa.jpg
but if that pic is right then presslers l2 cache is taking around 40% of the die space... thats 2mb on 32.4mm&#178;

hmmm that mans the l2 cache in pressler is ~15% more dense than in yonah... thats still not very dense... :/
but lets assume conroe has the same density of l2 cache.
lets build conroe with the new parameters :D

yonah is 96mm&#178;
38,4mm&#178; 2mb l2 cache block
57.6mm&#178; logic

yonah with 4mb cache would be:
38,4mm&#178; 2mb l2 cache block
38,4mm&#178; 2mb l2 cache block
57.6mm&#178; logic
=134.4mm&#178;

yonah with 4mb dense cache would be:
32.4mm&#178; 2mb l2 cache block
32.4mm&#178; 2mb l2 cache block
57.6mm&#178; logic
=122.4mm&#178;

conroe is around 145mm&#178; so there are 22.6mm&#178; of new logic, or to say it with other words, there is enough rom to increase the logic by 40%!
now thats more like it! :D
and maybe intel manages to squeene the cache even more dense... thougt i doubt it tbh.
intel has to keep the power consumption low and the core isnt really that large, its not worth to tweak the cache even more and risk higher power consumption and worse yields.
but it IS kinda weird that intel cant sem to keep the cache density high in 65nm...
40% more logic is nice!
plenty of room for 64bit registers, ht (even though it might be disabled and only come to life later on in a new core) and other improvements, bigger L1 cache, longer pipeline, better bus management logic...

sounds promising! :D

whats micro ops fusion then?
if you ask me thats just more advanced macro ops fusion :D
its not a new technology :P
or am i wrong?

and i didnt mean that intel coped the l2 cache block from yonah 1:1, of course not, but the basic design is most likely the same i think.
or do you think they took the pressler l2 cache block?
possible as wel, makes sense sense its more advanced as it came out later i guess... hmmm

ok, i agree that conroe has some elements that were first used in netburst cpus, but its not a netburst cpu as many people think.
intel mans they use the best tech from both plattforms, the pentium m and the netburst and took it to the next level. they never said conroe is a netburst cpu, and it isnt.
thats what i keep trying to say :)

true, intel can go dual die with conroe as well for servers...
that way they can put 4 cores on one cpu package without the exponentially growing yield issues...
and they moved a part of the northbridge on die already, wich is a smart move... now they get good yields AND a more efficient bandiwdht usage of the system bus...


http://img477.imageshack.us/img477/2268/conroesize9ze.jpg
i get the same size again, 140mm&#178;

[saaya]

http://img317.imageshack.us/img317/313/blownuppressler7aa.jpg
hmmmmm or 60% of the core is cache... that could be possible as well...
then we would have 81mm² 48mm² for 2mb cache... but that doesnt make sense, that would be a horrable cache density, even worse than amd! :D

[Pinnacle]

how does slower l2 cache increase yields?

ISnt it cheaper to produce and implement?

[saaya]

cheaper to produce how?
and implement... i guess, you start with slow less dense cache and then tweak it to squeze more cache onto the same space and make it run faster, but if you already have a weaked fast cache, then why would you slow it down again, especially if the cache speed is that crucial for dual cores who share the same system bus.
cache to cache is important, and cache helps to keep the shared system bus botleneck under control, so going for slower cache doesnt make much sense... hence i think they went for the slow cache because they used the prescott 2m, wich has slower l2 cache, and didnt tweak the cache because they didnt have time for it. they were in a hurry to release smithfield asap.
i dont think they went for prescott 1m with fastb l2 cache and then tweaked it to be slower... why would they waste time to make their cpus perform worse if they are in a hurry?
i cant imagine that making it slower increases yields...
if the cache is damaged its damaged, that part of the cache block has to be disabled then, its not that having slower cache saves you from defective sectors :D

[Pinnacle]

implement... i guess, you start with slow less dense cache and then tweak it to squeze more cache onto the same space and make it run faster

Thats what I thought.

[saaya]

k, well intel had prescott 1mb cache (fast) and prescott 2mb cache (slower, but more tweaked) so if you had to build a dual core cpu, wich of them would you pick?
the older but faster 1mb cache or the newer more tweaked 2mb cache?

so far everybody thougt they went for prescott 1m, wich makes most sense at first thougt, but i guess prescot 2m has tweaked cache for better yields and more importantly has some more logic tweaks prescott 1m doesnt have, it consumes less power, runs cooler, and performs slightly better.

the question now is whether intel went for prescott 1m, and changed the cache latency, or went for prescott 2m and cut half the cache off, wich isnt a big problem at all.

imo the latter makes more sense...

[FunkyRider]

Banias/Dothan's L2 seems better than Prescotts'

Since Dothan L2 only has 7 cycles latency, in contrast of > 20 cycles latency in Prescotts :D


More importantly, Dothan's L2 is very power optimized (16 partitions, 4 quadrants...)

[FunkyRider]

Thanks for the link of the photo. Now I've perspective-corrected the photo and scaled it to 1 pixel = 0.1 mm.

Did a best match measurement and got this result:

http://img8.picsplace.to/img8/5/conroe_die_big.jpg

Now processed the back as well, so, disclosed conroe photos! :D

http://img8.picsplace.to/img8/5/conroe_back.jpg

[ingentingmendeg]

lemme get this straight, yonah is more of the pentium M speeds like in the 2ghz range and the conroe is more of the P4 speeds in he 3ghz range?

also, sorry for all the questions, but will yonah be compatible with 955x?

[saaya]

yes, and no, compatible with the 955x chipset, yes, compatible with the boards... we dont know.
there have been rumors about desktop yonah chips but they will be 478 afaik
(not compatible with dothan 479 and not compatible with p4 478 either though :/)

so you will need a totally new plattform for desktop yonah chips... :/

and conroe will hit 3ghz speeds, eventually, the top end cpus, but it wont clock nearly as high as prescott cpus hitting 5ghz+
ad yonah comes with speeds up to 2.33ghz, so yeah, clockspeeds like the pentium m, aka dothan cpus.
but yonah ocs worse than dothan afaik...

DIT: btw, very nice work funky rider! :D

[flakpyro]

That Aopen Yonah board posted somewhere on this forum is S479 isnt it? Its interesting that desktop Yonah might be S478. Better make sure you get the two matched up correctly, hate to have a S479 Yonah and a S478 board haha

[grimREEFER]

dam, y are we talking about yonah in a thread about conroe? lol
i want a low end conroe....especially if it works on my board

[Donnie27]

I don't remember where I saw it but Conroe was to have something like 379 million trannies. More than and 7 X as many as the first 55 million trs P4's.

The amount each transistor cost isn't as important as the Wafer they're stampped out on. Point blank and to the point, the more Intel can stamp on each wafer, the cheaper each cost. If Intel can stamp out more 81mm2 than 134mm2, not matter watch features each has, the 81mm2 procs will be cheaper=P

Slower SRAM is easier to make=P

Controllers for INTEL'S version of Dual Core increases latency. I don't have any friggin' idea why Intel did it with Single core, 17% added to Presott's 2MB L2 after 1MB was already slower than Northwood's 512K.


975 has 1333 support?
thats a guess your taking, 975 does NOT have official 1333 support!
if it does then link me plz =P

afaik the current boards dont, only a nw revision of them has 1333 support, the 975xe is what they will call it i guess... maybe the delay in 975x boards means intel will wait and tries to jump to 975xe right away or cancel 975xe and make all 975x boards 1333 capable? makes sense seeing as conroe comes in just a few months and at least some are 1333

The i975x has a 1333 overrided, that's support. The 1333MHz FSB Conroe XE is the only processor with 1333 support and so far it is planned to ship with i985x, the only board with true official support, NOT any of the others. It's NOT something I'll touch with a 10ft Pole, IMHO, that's Bullspit!!!:mad: Bullspit part two, Conroe only lists i965. So if you but the Top board with an MSRP of $350, that came straight from an Intel rep, you get support of All Intel processors. If you pay less, you get less compatibility and support for fewer processors. So the Idiot at Intel responseable for this should be :slapass::slap:


and according to micron and some tests ive seen and what several mainboard and memory makers told me at cebit almost a year ago, ddr2 does consume MORE than ddr1. but that may vary from chip to chip i guess... samsungs ddr1 and winbonds ddr1 are very power hungry for example...

http://www.micron.com/products/dram/syscalc.html

sdram power consumtion 444.5mw (per chip?)
ddr sdram power consumption 349.2mw (per chip?)
ddr2 sdram power consumption 365.4mw (per chip?)

its all 256mb sticks, ddr 400 and ddr2 400 iirc
since ddr2 clocks much higher than 400 effective speed it consumes even more power at those high speeds

I stand corrected! Thanks for the Info.


sounds like theinq fell for the amd 2000mhz fsb marketing or at least enforces it as they call 1333fsb slow?
its by now means slow LOL

amd = 200fsb up to 266mhz aka ddr1 533 possible with dividers on the latest cpus and boards without ocing the fsb
intel = 266fsb up to 500mhz aka ddr2 1000 possible with dividers without ocing the fsb

amd am2 = (most likely) 200fsb and up to 500 mhz aka ddr2 1000 memory possible with dividers without ocing the fsb
so its more like amd will catch up with intels memory speeds!
they just have the advantage of beeing more efficient with the memory bandwidht with the integrated memory controler, so they didnt need the fast and high bandwidth memory before.

My Asus Probe says my 3500+ is 800MHz FSB LOL! Small correction though.
I do agree that AMD uses what it has much better. All 1333MHz FSB needs is DDR2 667=P That's effective cover bandwidth wise for 1333MHz FSB. All 1066 needs is 533MHz FSB. The BUSes have went from 6.4 to 8.4 to 10.2GB to cover these. The last be 12.6GB when 1600mhz is reached.


intel will move to 333fsb wich means at least 666mhz aka 1333mhz ddr2 memory speed, so this is not slow at all, its just a question of how well intel will use this massive bandwidth, how efficient the new onboard outer can distribute the bandwidth between both cores.

with the shared l2 cache wich reduces the bandwidth needed/makes the bandwidth usage more efficient, and the on die router, wich is half a memory controller on die if you want, PLUS the massive raw bandwidth of 1333 ddr2 i think intel will beat amd bandwidth wise.

so its kinda weird how theinq calls 1333 fsb slow ^^
or maybe it was meant ironic?

Maybe they're confused by bitwidth? Maybe they don't know how Intel came up with 10.2GB of bandwidth? It is very rare that that much bandwidth is ever needed. More times than not, it's how fast something is accessed.

Premiere Pro 7 takes full adavantage of Hyperthreading and SSE2/3. This app can fill the working units and do it without misses and cache flushes. 2.8GHz Prescott runs this app as fast as a 2.6GHz FX 64.


the big question remains though, at what price will intel gain the top spot, at least memory bandwidth wise, back?
ddr2 1333mhz wont be cheap, the 1333fsb cpus wont be cheap either, and neither will the boards be cheap...
__________________

It's already answered for me. 2GHz Yonah for $419 ( pre order deal/offer by a trusted un-named VAR) closes the Intel door for me. I don't like the wayyyyy too many overpriced motherboards, now add to that overpriced Processors as well. The Top Spot becomes MOOT for me personally, more power to those who care about such things though. The Top Spot never made Intel or AMD any money, Intel's profits the last 26 quarters shows that very clearly.

Donnie27

[saaya]

That Aopen Yonah board posted somewhere on this forum is S479 isnt it? Its interesting that desktop Yonah might be S478. Better make sure you get the two matched up correctly, hate to have a S479 Yonah and a S478 board haha
its 478 but its not compatible with 478 p4 or 479 dothan sockets :confused2
pretty lame of intel...


dam, y are we talking about yonah in a thread about conroe? lol
i want a low end conroe....especially if it works on my board
conroe = yonah+x


The i975x has a 1333 overrided, that's support. The 1333MHz FSB Conroe XE is the only processor with 1333 support and so far it is planned to ship with i985x, the only board with true official support, NOT any of the others.your conradicting yourself :D
i know what you mean though... but i wouldnt say because the boards have a 1333 override they support it... on a64s you can also get an fsb of 300 or change the memory divider to get ddr500 or higher with a stock fsb, but none of them are officially guaranteed to work.
if they dont, then they dont, i think itll be the same with intels 975x, they run 1333 but if your board only does 1332... too bad, you wont get a replacement board that does 1333...

well at least indirectly i think intel does make quite some money from the top spot, so does amd.
thats the lsson ati didnt get when they didnt go multi gpu...
crossfire and sli are pretty much bs if you ask me...
its yet another e-schlong you can buy and slap in peoples faces you dont like and feel good about yourself, but who seriously needs that graphics power?

if you have the fastest product out then more people buy your products.
people bought intel celeron for years because the p4 was the fastest cpu around... so having the fastest chip always improves sales quite significantly i think.

and dont forget that the high end cpus are the very same silicon as the low end chips. intel and amd , ati and nvidia make money by selling them as low end chips, everything they make more from selling them as high end chips is a bonus for them! thats why they make the biggest profits from the high end/upper mainstream class of products i think.

[coldpower27]

coldpower, how does slower l2 cache increase yields?
and it reducs power consumption? i dont think that cache has a big impact on power consumption, its just a few watts, 5W for 1mb at most, afaik cache is always the cool part of a cpu.
and considering that l2 cache to l2 cache on smithfield is creating a huge bottleneck for smithfield it would make more sense to go for faster cache i think.
think about it, intel needed a dual core cpu fast, so why would they go for the prescott1m to base it on that if they already had a more advanced prescott 2m chip?
it had more tweaks in the logic and cache than the prescott 1m, the only draw back was larger cache, but they can cut off cache just like that without a redesign, so i think it all points to intel basing smithfield on prescott 2m, all they did for pressler now was to shrink prescott 2m to 65nm and put two of them on one extra substrate to get btter dual core yields.
thats what i think at least.

I don't believe that Intel has the ability to cut off cache easily, they would simply disable it, if parts of it weren't needed, that's what they have done for most of the Pentium 4's life.

The die size of Prescott-2M is still larger then Prescott-1M, using two of those cores, would result in something 250mm2 die with having to slice off 20% of it for each and every core made, not the most efficient method of producing processor cores to me. As well in addition to this defect cores would not be usuable for anything and would have to be thrown out as the cache latency is worse then the original Prescott-1M core used for Celeron D and Pentium 4 5xx. Slowing down cache in a processor is not problem, making something quicker is. Yields would simply be highest on the Prescott-1M cores due ot the fact that Intel had much more time to work with them, this is of utmost importance when the die size of the Dual Core is going to be so large.

Well with the Cedar Mill/Presler cores I think Intel wanted to make the transistion more smoother and not have cases where Smithfield would be outperforming Presler at the same freqency due to the fact that the cache was a bit faster. So they knew what Cedar Mill's cache speed would be and slowed down Smithfield caches to that speed so it will have the same speed as Presler cache's guranteeing performance improvement from Presler at equivalent clock frequencies.

Not to mention slower cache means easier yeilds, if that wasn't the case we would have much faster cache on everything wouldn't we as all processors benefit to some degree from faster caches.



hahah gotcha! was thinking the same thing about merom beeing the 2mb chip :D
merom will be identical with conroe afaik, so mabybe the merom will be 2mb cache, because in laptops the 5W or 10W extra cache needs is not worth the little performence gans, especially since laptops dont need that much performence.
so maybe there will be 2 chips and the 2mb conroe will be identical with merom?

Well from what I have read now Merom is actually a 4MB chip as well, with 2MB versions though Allendale/Conroe are the desktop version but they have different codenames, that to me suggest that instead of disabling cache it acutally isn't there for the desktop version. With Merom 2MB/4MB they might just disable cache instead and use the extra space to decrease thermal density.



well about amds dual cres, yes, but those are gigantic cores, and the ones with less cache are almost half in size, amd cant reach as high cache densities as intel, not even close.
for intel conroe with 4mb and conroe with 2mb is just a difference in size of ~110 vs ~140, wich i dont think is worth it that much. thats a die shrink of a mere 20% compared to amds 50%+ for the different dual core chips.


No not really on the half the size part, AMD Toledo core on 90nm is 199mm2, while Manchester with half the total LV2 cache is 147 mm2. This represents about 26% reduction in die size, your 50%+ figure is completely off. AMD's cache desnity is inferior but not that that bad. :)

If going by 145mm2 vs ~ 110mm2 ish were looking at reduction in die of about 22%. That's like going from 420 to 550 chips per 300mm Wafer not insignifcant to say the least.



cache density on yonah is worse than on dothan? hmmmm

dothans die size is around 83.6mm²
57% of dothan is l2 cache, thats 47.7mm²
2mb on 47.7mm²

yonah is 96mm²
only around 40% of yonah is l2 cache, thats 38,4mm²
2mb on 38,4mm²

so the cache on yonah is more dense, but only 19.5%, wich is not much considering yonah is made in 65nm vs dothan in 90nm.
65nm should actually result in a 48% more dense cache if the density of the transistors would remain the same....
interesting!
thx for the pointer :toast:

pressler is 81mm²
the only time ive seen a die shot of pessler is at intels idf, and it was tiny...
http://img317.imageshack.us/img317/313/blownuppressler7aa.jpg
but if that pic is right then presslers l2 cache is taking around 40% of the die space... thats 2mb on 32.4mm²

hmmm that mans the l2 cache in pressler is ~15% more dense than in yonah... thats still not very dense... :/
but lets assume conroe has the same density of l2 cache.
lets build conroe with the new parameters :D

yonah is 96mm²
38,4mm² 2mb l2 cache block
57.6mm² logic

yonah with 4mb cache would be:
38,4mm² 2mb l2 cache block
38,4mm² 2mb l2 cache block
57.6mm² logic
=134.4mm²

yonah with 4mb dense cache would be:
32.4mm² 2mb l2 cache block
32.4mm² 2mb l2 cache block
57.6mm² logic
=122.4mm²

conroe is around 145mm² so there are 22.6mm² of new logic, or to say it with other words, there is enough rom to increase the logic by 40%!
now thats more like it! :D
and maybe intel manages to squeene the cache even more dense... thougt i doubt it tbh.
intel has to keep the power consumption low and the core isnt really that large, its not worth to tweak the cache even more and risk higher power consumption and worse yields.
but it IS kinda weird that intel cant sem to keep the cache density high in 65nm...
40% more logic is nice!
plenty of room for 64bit registers, ht (even though it might be disabled and only come to life later on in a new core) and other improvements, bigger L1 cache, longer pipeline, better bus management logic...

sounds promising! :D


Interesting comments. :D On another note I have also included a better Cedar Mill picture then the fuzzy thing you posted. :)



whats micro ops fusion then?
if you ask me thats just more advanced macro ops fusion :D
its not a new technology :P
or am i wrong?


No I would assume macro-ops fusion is more difficult to do as it is a new feature to be implemented, while micro-ops fusion already exist on Dothan, no point in implementing something more primitive that what you already have now isn't it.



and i didnt mean that intel coped the l2 cache block from yonah 1:1, of course not, but the basic design is most likely the same i think.
or do you think they took the pressler l2 cache block?
possible as wel, makes sense sense its more advanced as it came out later i guess... hmmm


Well I believe that Conroe/Merom will have taken the lessons learned in Yonah shared cache and improved on it. so I don't expect it to be exactly the same, but similar.



ok, i agree that conroe has some elements that were first used in netburst cpus, but its not a netburst cpu as many people think.
intel mans they use the best tech from both plattforms, the pentium m and the netburst and took it to the next level. they never said conroe is a netburst cpu, and it isnt.
thats what i keep trying to say :)


And then here is where you keep misunderstanding me, I never said Conroe is a NetBurst CPU either, I just said it has some elements of NetBurst architecture, doesn't make it a NetBurst CPU just by that statement. That is what I keep trying to say but each time you still don't get it and think somehow I said Conroe is a NetBurst processor.:confused:



true, intel can go dual die with conroe as well for servers...
that way they can put 4 cores on one cpu package without the exponentially growing yield issues...
and they moved a part of the northbridge on die already, wich is a smart move... now they get good yields AND a more efficient bandiwdht usage of the system bus...


Interesting what have you read on them moving what parts of the NorthBridge to the CPU?

[grimREEFER]

any chance current nvidia boards might support conroe?

[K.I.T.T.]

any chance current nvidia boards might support conroe?
Probably not :(

Anyway, nVidia + Intel = :smoke: :nono:

[grimREEFER]

Probably not :(

Anyway, nVidia + Intel = :smoke: :nono:
good point, but with everyone preparing to buy conroe's, and with intel chipsets supporting crossfire....nvidia better do something

[saaya]

coldpower27, why would intel have a problem cutting off cache from their cpus?
i think thats the easiest change you can do on a cpu design, they probably dont even need a new stencils for the new cpu revision without the full cache.
whether you have the cache and have it disabled, or whether its disabled and the cache isnt even present isnt a difference i think...

prescott 2m is larger than prescott 1m... because it has more cache... or do you mean the logic is larger? thats new to me :confused:

prescott 2m isnt a new design really, its just a new revision of prescott1m, so every tweak the found to increase yields for prescott 1m should work for prescott 2m as well, hence prescott 2m yields should be better (well worse because its larger, but you get what i mean...)

and i think the dies with a bad core were sold as prescott 1m chips or celeron d chips. i doubt anybody will notice that they have a 15% higher latency l2 cache, this doesnt mean the cpus are 15% slower, only around 5% at most id guess.


Well with the Cedar Mill/Presler cores I think Intel wanted to make the transistion more smoother and not have cases where Smithfield would be outperforming Presler at the same freqency due to the fact that the cache was a bit faster. So they knew what Cedar Mill's cache speed would be and slowed down Smithfield caches to that speed so it will have the same speed as Presler cache's guranteeing performance improvement from Presler at equivalent clock frequencies.good point!


Well from what I have read now Merom is actually a 4MB chip as well, with 2MB versions though Allendale/Conroe are the desktop version but they have different codenames, that to me suggest that instead of disabling cache it acutally isn't there for the desktop version. With Merom 2MB/4MB they might just disable cache instead and use the extra space to decrease thermal density.or they make two cores, one with 2mb and one with 4mb cache, and 2mb will mostly be used as merom 2mb, 4mb will be used as merom 4mb and conroe...


No not really on the half the size part, AMD Toledo core on 90nm is 199mm2, while Manchester with half the total LV2 cache is 147 mm2. This represents about 26% reduction in die size, your 50%+ figure is completely off. AMD's cache desnity is inferior but not that that bad. :)
my bad :D


No I would assume macro-ops fusion is more difficult to do as it is a new feature to be implemented, while micro-ops fusion already exist on Dothan, no point in implementing something more primitive that what you already have now isn't it.yeah but we have to keep in mind that those are all just marketing phrases, macro ops fusion isnt a "technology" :D
just like "hyper threadding" is just a marketing phrase, well thats at least my opinion... i think macro ops fusion is just more advanced micro ops fusion, hope there will be more infos about it soon :)


And then here is where you keep misunderstanding me, I never said Conroe is a NetBurst CPU either, I just said it has some elements of NetBurst architecture, doesn't make it a NetBurst CPU just by that statement. That is what I keep trying to say but each time you still don't get it and think somehow I said Conroe is a NetBurst processor. :confused:my bad, sorry :D


Interesting what have you read on them moving what parts of the NorthBridge to the CPU?nothing, i consider the system bus management logic, the bus router as intel calls it i think, as moving a part of the northbridge to the cpu.
as far as i know with smithfield and other older intel cpus this is all done in the chipset, in the northwbridge to be exact.
managing the buses and dataflow to and from the memory, cpu ,videocard and rest of the system is pretty much the main task of the chipset in my opinion... and intel moved a part of this to the cpu.

thx for the pic you attached!
wheres the high res version? :D

[Donnie27]

its 478 but its not compatible with 478 p4 or 479 dothan sockets :confused2
pretty lame of intel...


conroe = yonah+x

your conradicting yourself :D
i know what you mean though... but i wouldnt say because the boards have a 1333 override they support it... on a64s you can also get an fsb of 300 or change the memory divider to get ddr500 or higher with a stock fsb, but none of them are officially guaranteed to work.
if they dont, then they dont, i think itll be the same with intels 975x, they run 1333 but if your board only does 1332... too bad, you wont get a replacement board that does 1333...

I doubt Conroe is just a Yonah+ =P Intel's slides showed Netburst + Dothan and that's what I'll go with until it ships. There's NO-WAY Dothan or Yonah could scale to a 1066 and 1333MHz FSB for starters.

Not at all. It is no different than Asus, Abit and others adding PAT support to i865. If AMD shipped a 250MHz FSB proc instead of 200MHz and the board makers created the needed settings as a Defult settings, who cares?

Even while I was being tagged an Intel Employee, I said Intel sucked for all of the socket Crap-O-La. Add to this mess Yonah's new-old socket, it sucks even more, sheesh!


well at least indirectly i think intel does make quite some money from the top spot, so does amd.
thats the lsson ati didnt get when they didnt go multi gpu...
crossfire and sli are pretty much bs if you ask me...
its yet another e-schlong you can buy and slap in peoples faces you dont like and feel good about yourself, but who seriously needs that graphics power?

Nothing more than bragging rights=P GPUs are different than CPUs.


if you have the fastest product out then more people buy your products. people bought intel celeron for years because the p4 was the fastest cpu around... so having the fastest chip always improves sales quite significantly i think.

Only in this tiny market that doesn't matter in the Grand sceme of things. AMD even admits that all of the AMD DIYers amount to about 2% of its sales. That's 2% of roughly 20% of the market, ouch! Google it if you think I'm kidding:)?

People buy what works. No matter what you, I or anyone else says, folks saw AMD as Buggy. Most this was VIA's fault, some goes to AMD for not having a Chipset dept and even the first buggy nForce boards. Now that that's not so much the case, for about the last year, AMD sales picked up here in the US. Folks see it as not much different than the way they buy cars. Honda nor Toyota make faster "Consumer" cars, but try to get a long time Honda Owner to buy a Ford or Toyota owner to buy Chevy?


and dont forget that the high end cpus are the very same silicon as the low end chips. intel and amd , ati and nvidia make money by selling them as low end chips, everything they make more from selling them as high end chips is a bonus for them! thats why they make the biggest profits from the high end/upper mainstream class of products i think.

I made the same point about low-end Procs rescured from the trash bin for various reasons. Even in that removed stickey thread I said Cedar Mill will prove to be the big money maker. Even as others talked about it being older and yada yada! I told Nasgul on another forum to "Don't look past Cedar Mill so quickly", that was almost 5 months ago. Though it's not an X2, it will kick the crap out of AMD's MAINSTEAM single core procs that AMD will still have to sell. Price is the only thing Intel can screw up IMHO! Waiting for Conroe makes no sense, just as waiting for M2 or etc..

Though I'm getting great gaming performance with my 3500+, it is NO faster than my un-overclocked 2.6C for most multi-media apps and gets crushed when multi-tasking. It's a great gamer rig though, hehehe! Mine also is one of those "Please DON'T overclock me" models. So much for all of the REV-E wonderful overclocking bragging. Honestly though, that's NOT a biggie for me.

Donnie27

[rotjunior]

I didn't read through this entire thread but could anyone tell me what is the earliest boards that support conroe? I am looking for a mobo upgrade and I want to be able to go to conroe when it comes.

[saaya]

I didn't read through this entire thread but could anyone tell me what is the earliest boards that support conroe? I am looking for a mobo upgrade and I want to be able to go to conroe when it comes.

get a 975 :)
955 should work but youll be limited because of the lower fsb it supports.


I doubt Conroe is just a Yonah+ =P Intel's slides showed Netburst + Dothan and that's what I'll go with until it ships.dude!
thats marketing... :slap:
and they do use tech they had in netburst cpus or that is based on netburst technology, so the best way of explaining this to the analysts and customers whop usually cant tell apart a smashed bean from a cpu is to make a slide like they did.


There's NO-WAY Dothan or Yonah could scale to a 1066 and 1333MHz FSB for starters.dothan does 300fsb (1200 quad pumped) just fine with a decent board...


Not at all. It is no different than Asus, Abit and others adding PAT support to i865. If AMD shipped a 250MHz FSB proc instead of 200MHz and the board makers created the needed settings as a Defult settings, who cares?depends... some mfgs officialy supported pat, others didnt...
whether current 975 boards officially support 1333 fsb will be seen in the future, atm they dont. if they would then it would show on all the mainboard boxes of the 975 boards in huge letters "1333fsb support for future cpu upgrades!"
but it doesnt... so how can you say 975 supports 1333fsb?
a64 cpus can be oced to 300fsb but that doesnt mean they support 300fsb... same for the mainboards.

well ok, depends on how you define support, supports feature xxx means its guaranteed from my understanding.
maybe thats where we misunderstand each other :D


Nothing more than bragging rights=P GPUs are different than CPUs. because...? :P

if a brand is established youve won big times.
just look at how people readted to prescott. people bought it even though its one of the worst chips intel ever made and praised it... thats the power of brand loyalty.
and even the people who didnt buy a prescott dont consider intel as bad now, they only say prescott sucks, and still think positive about other intel brands. thats the power of brand loyalty!
same for amd ati and nvidia... remember how people defended the geforce fx and people bought it even though it sucked biiig time :D
brand loyalty...
if somethings good then its good... whoops its not good?
nah, must be a mistake, still not good?
get another one...

the first impression is very important, if its positive then people keep comming back for that brand even though the latest products suck...


Though it's not an X2, it will kick the crap out of AMD's MAINSTEAM single core procs that AMD will still have to sell. Price is the only thing Intel can screw up IMHO! Waiting for Conroe makes no sense, just as waiting for M2 or etc..how does a 3ghz dual core beat the crap out of a a64 3800+? in most apps the a64 still wins because most apps arent multithreadded yet, and even with multithreadded drivers games are still much faster on a single core a64 than on a 3ghz smithfield.
and i doubt this changes a lot when ocing...

plus right now a x2 3800+ costs the same as a smithfield 830 3ghz...
if your smart you can even get a 165 for the same money...

gotta go, hey no offense donnie, getting more and more fun to discuss this stuff with you even though i still disagree in many parts :D :toast:

[grimREEFER]

according to asus, the 955 chipset doesnt even support the new extreme edition(with 1066 fsb), how can it support conroe?

[Donnie27]

get a 975 :)
955 should work but youll be limited because of the lower fsb it supports.

dude!
thats marketing... :slap:
and they do use tech they had in netburst cpus or that is based on netburst technology, so the best way of explaining this to the analysts and customers whop usually cant tell apart a smashed bean from a cpu is to make a slide like they did."

Not only them, but when a 2 min burn takes 3 minutes, it is very easy to tell the difference LOL! When it takes 25 mins to transcode 200 photos instead 14mins, yes there's a difference.Taking one hour to do 37 minutes of work, yes there is a difference:slap: Mostly how you use YOUR computer, how much money you're willing to spend and etc.. means more than AMD vs Intel crap-o-la.


dothan does 300fsb (1200 quad pumped) just fine with a decent board...

Conroe XE 1333MHz FSB at a Speed of 3.33GHz. Conroe's other model is 3.2GHz, not the first talked about 2.93. These are stock settings, no overclocking at all. Overclock estimates means something like 1400MHz for i965 and at least 1600MHz for i985 and higher.


depends... some mfgs officialy supported pat, others didnt...
whether current 975 boards officially support 1333 fsb will be seen in the future, atm they dont. if they would then it would show on all the mainboard boxes of the 975 boards in huge letters "1333fsb support for future cpu upgrades!"
but it doesnt... so how can you say 975 supports 1333fsb?
a64 cpus can be oced to 300fsb but that doesnt mean they support 300fsb... same for the mainboards.

I said 1333MHz FSB, I don't know if that means Conroe or even which Conroe. Conroe = 1066MHz FSB. Conroe Extreme Edition = 1333MHz FSB. Conroe = i965/i963 for its mobil version. Conroe XE = i985. Even the Intel guy said i975 shipped to introduce8GB RAM support, ICH8, Improved DME, Flixible Memory slots, and SLI/CF PCI-E. I was told last nite by the same folks that i955X and i975 will NOTsupport either Conroes except for a Single core Celeron type version. That's what created the seemingly Contradition.


well ok, depends on how you define support, supports feature xxx means its guaranteed from my understanding. maybe thats where we misunderstand each other :D

Maybe.


because...? :P

if a brand is established youve won big times.
just look at how people readted to prescott. people bought it even though its one of the worst chips intel ever made and praised it... thats the power of brand loyalty.
and even the people who didnt buy a prescott dont consider intel as bad now, they only say prescott sucks, and still think positive about other intel brands. thats the power of brand loyalty!
same for amd ati and nvidia... remember how people defended the geforce fx and people bought it even though it sucked biiig time :D
brand loyalty...
if somethings good then its good... whoops its not good?
nah, must be a mistake, still not good?
get another one...

People also bought the first TBread that were little easybake ovens, what's your point?

I've always went back and fourth between AMD and Intel. I liked my old 66MHz Intel chip, as did an AMD 120. The first PR sucked big time. 166MMx killed the PR-150 that was slower an Intel 133MHz. I like my for K6-300, thought the 380 sucked a big one, wouldn't run at the needed 95MHz FSB. Or the VIA motherboards that didn't get the correct 4 in1's until March 99. Let's see my KX-133 sucked, Abit KT 133 REV2 and 133A REV1 and Asus 266M. It was a long time before I went back to anything to do with AMD inside. ALL of MY Intel systems have performed like Swiss Clocks so it is hard to move away from them.

That doesn't give Intel a free Pass on all of the Different Chipsets and Sockets, that's absolute BS IMHO! So when you have a new chipset for each new processor, compatibility SHOULDN'T be a frackin' problem. It's not like running 3 generations on 400 BX or i875/65.


the first impression is very important, if its positive then people keep comming back for that brand even though the latest products suck...

how does a 3ghz dual core beat the crap out of a a64 3800+? in most apps the a64 still wins because most apps arent multithreadded yet, and even with multithreadded drivers games are still much faster on a single core a64 than on a 3ghz smithfield.
and i doubt this changes a lot when ocing...

That's just it. To many of these folks the Products doesn't suck. They don't use their systems like many of us do. Intel users thought it was crazy to not multi-task with these expensive systems. I'm not talking about AMD vs. Intel in general. I'm talking 1. MY AMD system and 2. MY Intel system. When purchased they were the same price.


plus right now a x2 3800+ costs the same as a smithfield 830 3ghz...
if your smart you can even get a 165 for the same money...

gotta go, hey no offense donnie, getting more and more fun to discuss this stuff with you even though i still disagree in many parts :D :toast:

I built the AMD rig to tide me over until the Fall. My HTPC will do for now and there's NO *real pressing need to replace it.

Hey, if everyone agreed on everything, this would be one boring place. I just started back to posting here because there seemed to be fewer AMD Crusaders soiling up the Intel side of the forum.:)

Donnie27

[Sentential]

Im reading this info and it doesnt feel right. Intel is good but they arent magicians. How in the world did they get Conoe @ 14 to clock to almost as high as Presler @ 31 with the same amount of cache and a significantally reduced TDP? I mean.. doesnt that defy some law of physics?

It doesnt make sense..... either netburst was really that f-e-d-up or those specs arent correct.

The last time that I had 100% confirmation of Conroe chips were ESs that were clocked at 1.8ghz, not 3.33. I dont understand how they can jump the span of 2 fold in a matter of 4 months when clearly Presler is having yield issues.

[onewingedangel]

I don't believe pressler is having yield issues, and if you think about it 4.2ghz+ is easy with pressler, and the 31 pipeline stages is a design for speeds of approx 5ghz on a 90nm node. Heat and power consumption meant that this wasn't achievable, and the design filtered down to 65nm. Tejas was likely going to have even more stages and be clocked even higher for the 65nm node. The netburst plan didn't work, developers didn't code for it, which meant that the design vastly underperforms compared to its potential, and problems with ever increasing gate leakage on smaller processes meant that it was unfeasable for intel to plug ahead and hope that developers came around to their way of thinking. (I still maintain that if a game was coded for netburst a p4 would likely perform 50% greater than it currently does in cpu limited scenarios).

For 14 pipeplines in a 65nm node 3.33ghz sounds about right, remember that the 18 stage p4 did north of 2ghz stock and could best 3ghz at 180nm, and the 21 stage p4 was able to clock north of 4ghz at 130nm.

I do however wonder about how much beyond that speed intel can scale - andremember that this microarchitecture looks to remain for a long time, and until 45nm, other than clock increases there doesn't seem to be any core improvements, so its only clock and bus speeds that will improve, and if they launch at 3.33ghz and the design can only scale to ~3.5ghz there is virtually no room for improvement to compete with amd's 65nm offerings other than offering a mcm version with 4 cores. I'm hoping that the design has headroom above this, maybe encroaching on 4ghz, but this seems unlikely. A 3.33ghz EE would seem to offset the problem with transitionng from p4 to the next generation architecture, as a 3.33ghz chip is not clocked much lower than the 3.46ghz EE we currently have, so even joe six pack will be willing to make the transition.

[Donnie27]

Im reading this info and it doesnt feel right. Intel is good but they arent magicians. How in the world did they get Conoe @ 14 to clock to almost as high as Presler @ 31 with the same amount of cache and a significantally reduced TDP? I mean.. doesnt that defy some law of physics?

It doesnt make sense..... either netburst was really that f-e-d-up or those specs arent correct.

The last time that I had 100% confirmation of Conroe chips were ESs that were clocked at 1.8ghz, not 3.33. I dont understand how they can jump the span of 2 fold in a matter of 4 months when clearly Presler is having yield issues.

Don't you think AMD could do FX-62 at 3.2GHz if they added two pipes and were on a 65nm Process?

From what other *experts said, the pipeline lenght might be changed from Dothan and Yonah.

What if Conroe used something more like, a moded version of the Northwood's 20 stages or pipes? Northwood at 65nm would probaly do 4GHz easy so 3.3GHz wouldn't be much of a strain at all. Even for 14 pipes.

Donnie27

[Donnie27]

I don't believe pressler is having yield issues, and if you think about it 4.2ghz+ is easy with pressler, and the 31 pipeline stages is a design for speeds of approx 5ghz on a 90nm node. Heat and power consumption meant that this wasn't achievable, and the design filtered down to 65nm. ...........
......... I'm hoping that the design has headroom above this, maybe encroaching on 4ghz, but this seems unlikely. A 3.33ghz EE would seem to offset the problem with transitionng from p4 to the next generation architecture, as a 3.33ghz chip is not clocked much lower than the 3.46ghz EE we currently have, so even joe six pack will be willing to make the transition.

Aweeeee But Intel estimates it will be at least 40% faster. On the other hand, MAC users will not have a choice LOL!

Oh, what the other guy said about Hyperthreading? I was told of a Single Version with Hyperthreading, VT and all of the other features.

There are Roadmaps starting to show up with the 3.2 and 3.3XE on them as well. Unless these are a Hoax, not impossible for that to happen for sure.

Donnie27

[Nasgul]

* As for price, the price when it hits the shelf will be expensive, cuz this would be recoginzed to as a "new generation processor" from Intel.
id guess, yeah, very expensive, especially since its probbably going to beat amd, so they can ask for a lot


Dead F:nono: ing:

W-R-O-N-G (http://www.tgdaily.com/2006/02/11/intel_conroe_to_launch_in_q3/)

E6600 for $316 :toast:

Intel Rules!

[coldpower27]

Now that we have a price point to gather from we can see what the competition will likely be.

Intel Third Quarter 2006

Core Processor E6700 2.67GHZ/1066FSB/4MB 530US Conroe
Core Processor E6600 2.40GHZ/1066FSB/4MB 316US Conroe

Core Processor E6400 2.13GHZ/1066FSB/2MB 241US Allendale
Core Processor E6300 1.86GHZ/1066FSB/2MB 209US Allendale

Intel Third Quarter 2006

Pentium D 960 3.6GHZ/800FSB/2x2MB 316US Presler Engages E6600, 2.40GHZ/1066FSB
Pentium D 950 3.4GHZ/800FSB/2x2MB 241US Presler Engages E6400, 2.13GHZ/1066FSB
Pentium D 940 3.2GHZ/800FSB/2x2MB 209US Presler Engages E6300, 1.83GHZ/1066FSB
Pentium D 930 3.0GHZ/800FSB/2x2MB 209US Presler
Pentium D 920 2.8GHZ/800FSB/2x2MB 178US Presler


Intel Dual Core launch lineup for NGMA is 4 processors which matches the Athlon 64x2 launch amount as well, not to mention the Presler (Mainstream) launch.

[Salvador]

Where did you get those prices from? 316 US fr that E6600 > Pure love perhaps.