775 dual cores wont work on current boards

[saaya]

http://www.xbitlabs.com/news/cpu/display/20050128032121.html

:down:

[caater]

mby it's like with prescott.. some boards support them and some boards wont.
and some boards have only experimental support which results in mosfets melting down in long term use :P

[Xassius]

maybe a bios hack will enable the current 915/925 chipsets to work with dual cores.

[Tim]

That sucks bigtime, the AMD ones ones work on 939 right?

[thegreek]

That sucks bigtime, the AMD ones ones work on 939 right?
nope.... it'll be like socket 1000 something.

[Sniper]

Wrong, AMD dual cores will work in exsisting mb check (http://www2.amd.com/us-en/protected/Weblets/1,,7832_11175_12416_12380,00.html) Login required.

[Tim]

yeah I also found out....939 is also for AMD Dual cores....

I bet this is gonna kill Intel....and AMD to climb even further in terms of sales...thank god... :)

[jjcom]

yeah, S939 and S940 will suport dual core, current chipsets also should support it as well. So no upgrades required on the MB end of things. Should be nice. yeah those dual cores will need so much power to work...those things will be able to heat homes

jjcom

[enzoR]

whats wrong with intel. if AMD can do it why cant they. wait a second. i bet they can. cept they wanna sell more chipsets again. bastards!

[Tim]

whats wrong with intel. if AMD can do it why cant they. wait a second. i bet they can. cept they wanna sell more chipsets again. bastards!

probably :(

[JNav89GT]

roflmao!
well to be honest, we all tend to upgrade our mobos every few months anyway, but funny Intel is making the upgrade necessary!

[jjcom]

yeah, its that whole thing where we do it anyway, but when someone tells us to...

jjcom

[JNav89GT]

exactely :O
oh well, what are we going to do except complain

[Donnie27]

Wrong, AMD dual cores will work in exsisting mb check (http://www2.amd.com/us-en/protected/Weblets/1,,7832_11175_12416_12380,00.html) Login required.

By the end of the year, AMD will be completely switched over to Socket "M2" or did you guys miss that part?

You'd think after 940, 754, 939, and 5 chipsets, complaining about Intel Platforms would be kind of silly? I didn't want to use my i850 with my Northwood "C", just as I wanted someing better than my old KT-133A for my Palomino. That was before I stopped using AMD processors because of their crappy platforms, I looked forward to something better than the last nForce 1 I used. Finnally I gave up LOL!

Donnie

[lutjens]

This soundsl like typical Intel behavior...after all, Intel does sell chipsets too...:rolleyes:

It's simply a not so subtle way of giving theur chipset division a boost when the processors are released.

It'll hurt them in the short term, though...I don't think people will be very eager to run out and get a new motherboard now, knowing it'll be useless with Smithfield.

On a more hopeful note though, hopefully the new (or refreshed ;)) chipset will allow for more PCI-E lanes (no more watering down the second PCI-E slot to 4x).:)

It's largely moot point though...AMD's dual core solution will knock the socks off of Smithfield. I think that AMD's HTT technology will really begin to show its muscle with dual core and the dual core Opterons will indeed be forces to be reckoned with...especially with NUMA support.:slobber:

The old shared FSB design of the Xeon will begin to painfully show it's limitations...especially when four cores are fighting for a slice of the shared FSB. Intel does have plans for multiple FSB's, but they will come with the Blackford and Greecreek chipsets, which are slated to be released substantially later than the Smithfield. It appears that upon its launch Smithfield will be hampered by the shared FSB.

My $0.02...;)

[jjcom]

yeah, I don't understand Intel here...they know their chips preform best with a high FSB...and it wouldn't be hard for them to go with a 1066mhz FSB stock...

jjcom

[OneyedK]

If I were Intel, I'd first release dual-core 800Mhz fsb cpu's three months before my archrival can release any...

Then, I'd wait to see how well the products of my rival perform...
Up my sleeve, I'll keep the 1066Mhz fsb, if it's necessary, I'll enable it to hit my rival with some 33% performance gain :D

And yes, in the meantime, the OC community will already have taken the fsb of my beloved 800Mhz fsb far over my silly 1066Mhz fsb.

Rumours, rumours, rumours...
btw, the i845 chipset never officially supported 800Mhz Northwoods either...
so, let's just wait...

I'm curious how the new AMD's will perform... I came back to Intel because I got used to HT. And if you don't have it, you miss it.
With dual-cores and HT disabled on Intel cpu's, we will finally have a fair comparison...

[jjcom]

Its not a 33% preformance gain with 800 to 1066...maybe if we're talking 1200 but even then I'd say it would be unlikely. 800mhz dual core shouldn't be too bad...I just think Intel will have more yield problems than AMD simply because of how hot these chips will run.

jjcom

[matt9669]

Personally the only reason I can think that Intel would make their dual cores 800MHz FSB is so that the "EE" line stands out, though the EE chips are dead as far as I'm concerned . . . bus contention with two cores/processors is bad enough on Xeons, but ramping the bus on single proc/dual core is much easier than ramping the bus on dual proc/single core (the clock signal doesn't have to go as many places - it's that simple!).

I guess Smithfield is already encroaching on Xeon space, so that'd be another good reason to keep the bus at 800 . . .

Why a new chipset is required makes no sense - take the PC-DL, which runs off the i875 (and runs very well, I might add). New motherboards, that I can understand . . . the power circuits might not be able to handle 130W and/or the VID requirements of Smithfield, even on the S775 Prescott boards.


If I were Intel, I'd first release dual-core 800Mhz fsb cpu's three months before my archrival can release any...
IMHO :devil: the only reason Smithfield exists at all . . .

The new Socket M2 for DDR2-based A64's won't come into play until 2006, and probably more like 2-3Q 2006 at that (read my news post!). Intel will have a good few months lead on dual core chips, but the AMD ones will work in existing 939/940 boards and on a much better power budget, plus they won't be on a shared bus like Smithfield.

[i found nemo]

gotta give it to intel, they sure know how to get cash flow goin'

[matt9669]

From Wired News' Top 10 Vaporware list:


"Intel flopped on this one, and now in this category -- just like any other -- Intel is following AMD's lead," said Chad Meirose. "It is about time that AMD gets the public credit that they deserve, and Intel (who also makes great products) is shown for what they really are, a PR behemoth."
It's just a big marketing game over there - but, the financial bottom line is all that matters in the end, how you get it done is just details ;)

[FUGGER]

I will have one soon.

I am told it fits and works but you need the 945/955 to fully support dual core. Kinda like supporting 1066FSB

CPUID string additions (like early prescott) may prevent boards from firing up without bios updating. CPUZ 1.26 recognizes it as a Prescott .09, speed and bus are correct as well as showing dual processor.

[Xerxes]

A guy at one of the intel campuses here has a dual core and he claimed it would work on a 925 board which is what he bought at the shop i used to work at.

[JNav89GT]

I will have one soon.

I am told it fits and works but you need the 945/955 to fully support dual core. Kinda like supporting 1066FSB

CPUID string additions (like early prescott) may prevent boards from firing up without bios updating. CPUZ 1.26 recognizes it as a Prescott .09, speed and bus are correct as well as showing dual processor.
well that's cool Fugger!
probably will be under NDA, but would be nice to notice real world gains.

I know AMD/Intel are really pushing dual core, but I really question what this gives consumer. Sure multi-tasking will be better, but to be honest, the current P4 w/HT gives us good multi-tasking now. If your not big into multi-tasking and want gaming more-so then AMD is better choice.

I think CPU makers are just realizing they can't keep scaling w/out costing alot of $$$ and want to sell us on new angle?
But what happens in 1-2 yrs when dual cores are topped out? Then what? Burp! Need another beer! :toast:

[Donnie27]

Its not a 33% preformance gain with 800 to 1066...maybe if we're talking 1200 but even then I'd say it would be unlikely. 800mhz dual core shouldn't be too bad...I just think Intel will have more yield problems than AMD simply because of how hot these chips will run.

jjcom

In almost two years, AMD at about the 2 million Hammers built mark. In that same amount of time Intel has built about 50 Million Hyperthreaded processors. AMD has had yield problems out the Ying Yang. Let's see, their yields of Hammers one at a time sucks, but they'll like magic get it right? The same folks who can't get high yeilds of Hammers one at a time is supposed to all of a sudden have two at a time worked out nicely. WOW! Building two processors will be twice as hard to get right. There's twice as much to go wrong and so on.

In a market where 150 to 160 million PCs are sold, not counting servers, MIPS and etc., just PC's, Hammers (ALL OF THEM UP TO NOW) represent Peanuts.

For the record, AMD has talked more about Server Dual Core Opteronss with HP and IBM footing the bill, than Desktop. Intel threw out a price of $240 for Dual Core 2.8GHz Smitties, seen AMD's prices?!

Donnie

[FUGGER]

Looking at mobile and desktop

Both are moving towards dual core

Dothan is behind in features, it will soon have HT and a longer pipeline.

P4 core already is capable of HT, VT, x86-64, sse3, NBX, ...

Dothan is planned to get more features, its very strong and with a bus at full speed it will surely impress.

If we get whats perposed then I can see dedicating a core to an application while other core runs os functions. Down the line I think its for VT where you can run two different os's on each core or other functions.

Its coming quick and I can see use of the dual core technology to be more applicable to us and consumers than x86-64.

I will see if I can get confirmation on HT having higher latency than arbitration on Sonoma.

[matt9669]

Its coming quick and I can see use of the dual core technology to be more applicable to us and consumers than x86-64.
Depends, x86-64 could do wonders for games with >4GB memory spaces - would be the defining factor between consoles with small, high speed shared memory blocks and PC's with somewhat slower, but larger main memory. Think huge, seamless game worlds with no zoning/loads between game areas. RAMdrives, MRAM, 64bit addressing, all points to less HD use and that means way faster execution.

Dual core is the industry's version of slapping on a band-aid, but in a few years it could make multitasking extremely efficient with VT etc. Main problem with dual core is multithreading takes a lot of work on the programmer's end - 64bit addressing does not.

Another thing to think about - the A64 has 16 local registers in x86-64 mode. Compilers that use these additional registers could see speed gains without recoding software. (This is something school has hammered into my head, and seeing how Itanium lives in the market proves it true - software engineering is expensive, and most feel if it ain't broke, don't fix it.)

Note I'm leaving out 64bit integer processing, since only database/scientific apps will have large gains from this.

The concept of dual core is fantastic, not bashing anyone there. Just thought I'd add my 5/2 of a nickel . . . :D

[jjcom]

I don't mean to the point where there is a shortage, I'm just saying that heat I think would cause a few yeild issues, nothing major. yeah, AMD will have abit of troulbe I think as well, I just won't be as heat related. Thats my opinion

jjcom

[uclajd]

But what happens in 1-2 yrs when dual cores are topped out? Then what? Burp! Need another beer! :toast:Quad cores! :lol:

[saaya]

yeah, its that whole thing where we do it anyway, but when someone tells us to...

jjcom

its called freedom :D


By the end of the year, AMD will be completely switched over to Socket "M2" or did you guys miss that part?

You'd think after 940, 754, 939, and 5 chipsets, complaining about Intel Platforms would be kind of silly? I didn't want to use my i850 with my Northwood "C", just as I wanted someing better than my old KT-133A for my Palomino. That was before I stopped using AMD processors because of their crappy platforms, I looked forward to something better than the last nForce 1 I used. Finnally I gave up LOL!

Donnie

yeah, there are new chipsets for amd as well, quite frequently... :

939 nf3 150
939 nf3 250
939 nf4

the difference however is that you can still plug in the latest winchester core and run it in an nf3 150 board...
the diff


i really dont understand intel here, the amd dual cores will work on almost all 939 boards if not all. BUT only the first dual cores, the lower clocked ones will work on those boards alright, the higher clocked ones will probably ned more power and need a new board.

why doesnt intel do the same? why cant the slow 3ghz dual core chip be made compatible with exisiting boards? power consumption is about the same if not less than a 3.8ghz prescott, and there are enough pins as well...

thats what i thougt too, but no, dual cores will still only have an 800qpb! :confused:


Looking at mobile and desktop

Both are moving towards dual core

Dothan is behind in features, it will soon have HT and a longer pipeline.

P4 core already is capable of HT, VT, x86-64, sse3, NBX, ...

Dothan is planned to get more features, its very strong and with a bus at full speed it will surely impress.

If we get whats perposed then I can see dedicating a core to an application while other core runs os functions. Down the line I think its for VT where you can run two different os's on each core or other functions.

Its coming quick and I can see use of the dual core technology to be more applicable to us and consumers than x86-64.

I will see if I can get confirmation on HT having higher latency than arbitration on Sonoma.

you think dothan will get ht? but that would be a mayor reorganization of the core, no? and if they have two cores why add ht? i thougt dothans will just get two cores, 64bit and XD?

[saaya]

hmmm can anybddy explain to me why dual cores from intel wil go desktop first and server later? a 3.8ghz single core prescott will kill a dual core 3.2ghz in games, so whats the idear of selling those chips in the desktop segment? will be great for media en-de coding, but thats about all i can think of the average joe could do to take advantage fo the two cores...

[jjcom]

hmmm can anybddy explain to me why dual cores from intel wil go desktop first and server later? a 3.8ghz single core prescott will kill a dual core 3.2ghz in games, so whats the idear of selling those chips in the desktop segment? will be great for media en-de coding, but thats about all i can think of the average joe could do to take advantage fo the two cores...

Yeah, that does seem odd...Most home users won't need dual core. Server Market would sell more. Just had an idea. intel wants to have dual core first, but my guess its expecting lower yeilds. Desktop users won't need dual core, therfore they won't be buying many.

jjcom

[Marlowe]

Unlike AMD’s plans, Intel has a different strategy how to bring these processors into the market. The Micro-processor giant is not going to start with the workstation and server markets, but with the desktop solutions. In fact, this is a pretty evident and logical plan. It is exactly in the desktop segment that Intel is falling most behind its major competitor – AMD. While AMD has successfully managed to significantly speed up its desktop solutions this year, Intel’s processors got just a little bit faster in terms of clock frequency growth because of multiple technological troubles. That is why Intel is hurrying so much with the development of dual-core desktop solutions: the first CPUs based around new architecture should appear in the market in Q3 2005 already. Today these CPUs are known under Smithfield codename, and they will be targeted as competitors to top desktop AMD CPUs, which performance ratings should reach 4200+ by mid next year. http://www.xbitlabs.com/articles/cpu/display/dual-core_6.html

I don't think we'll see any good performance from Intel cpu's in 2005. But Intel has got deep pockets and they can afford AMD getting a bit larger market share. I'm shure Intel's heavy machinery will come up with fast desktop cpu's again in some time, eventually retaking the performance throne.

[matt9669]

IMHO :devil: it's just typical Intel behavior - they want the industry to focus on multithreading/multitasking, they want the software to be coded that way, so they push dual core to the mass market first.

[Kunaak]

Considering how much a disaster the 915 and 925 chipsets are, I consider this a good move. hopefully intel will make a decent motherboard without such flakey features next time on the 945 and 955 (is that right?) chipsets, with DDR2 that doesn't go nuts when the motherboard heats up and gives you 1000000 mb bandwith, or 5,000 bandwidth next time...

and a board with no OC limits and such.

basically a Canterwood equivalant for the 775.
cause so far, these 915 and 925 chipsets are a joke.

so they have a new motherboard coming out soon, to me, thats good news.

[matt9669]

... the 945 and 955 (is that right?) ...
Yep!

I'd have to agree, I think even Intel has recognized the poor performance of the current i9xx, and VIA may have a good market opportunity with their new PT chipsets. If I remember, it was originally in the cards for all P4's (not just the Expensive Edition) to be on 1066FSB . . . Intel had to make a revision just to support that and yields have not been good to my knowledge.

So if the new chipsets are up to i875 standards, Intel's dual core ovens may be worth it. I'd hope to see an improved DDR2 controller as well, which would reduce infighting between two cores on a shared bus.

[Donnie27]

Considering how much a disaster the 915 and 925 chipsets are, I consider this a good move. hopefully intel will make a decent motherboard without such flakey features next time on the 945 and 955 (is that right?) chipsets, with DDR2 that doesn't go nuts when the motherboard heats up and gives you 1000000 mb bandwith, or 5,000 bandwidth next time...

and a board with no OC limits and such.

basically a Canterwood equivalant for the 775.
cause so far, these 915 and 925 chipsets are a joke.

so they have a new motherboard coming out soon, to me, thats good news.

Yes but that sounds more like excuses to me. There are great choices that can be made if you want an Intel processor. The key is "want". An Intel User is not "stuck" with any one type of Platform. You don't need to mess with the "joke" i9xx anyway.

There are Canterwood Socket 775 mother boards. If you don't like DDR2, there are i915s with DDR1. And they still have Better I/O Controllers, glitchless USB and Firewire, and the only real joke is just how much money Intel made while AMD got red figures for the 4th quarter. Some guy joke that Intel made more money in 12 days than AMD made all year.

Donnie

[Donnie27]

its called freedom :D

yeah, there are new chipsets for amd as well, quite frequently... :

939 nf3 150
939 nf3 250
939 nf4

the difference however is that you can still plug in the latest winchester core and run it in an nf3 150 board...
the diff

you think dothan will get ht? but that would be a mayor reorganization of the core, no? and if they have two cores why add ht? i thougt dothans will just get two cores, 64bit and XD?

Please try again?

No, that's not what was expressed! The poster posted it as if Socket 939 CPUs fit into Socket 754 or Socket 940 and will still be compat with M2. If not the folks need to stop using terms like, "Always stuck to one socket longer", Antone from Xbits labs or the posts here. So this is just plain wrong, not an opinion in this case. Too many sockets is too many sockets, doesn't matter if it's Intel or AMD shipping them. Sc-754 didn't live any longer than the 1 year old Sc-423 lived:mad: This eventhough AMD implied it would live longer. Hammer has seen 3 sockets in less than two years.

Here what I said!



You'd think after 940, 754, 939, and 5 chipsets, complaining about Intel Platforms would be kind of silly? I didn't want to use my i850 with my Northwood "C", just as I wanted someing better than my old KT-133A for my Palomino. That was before I stopped using AMD processors because of their crappy platforms, I looked forward to something better than the last nForce 1 I used. Finnally I gave up LOL!

First all, Socket 939 is NOT the first Hammer socket, came out almost a year after the first Hammer shipped and no amount of spin can change that. Most of the folks whinning act as if 939 launched with the first hammers. That was my point. Yes and chipset added that went from 800MHz to 1GHz Hypertransport Lanes as well.

Donnie

[Donnie27]

Yep!
So if the new chipsets are up to i875 standards, Intel's dual core ovens may be worth it. I'd hope to see an improved DDR2 controller as well, which would reduce infighting between two cores on a shared bus.

AMD will be doing the say thing with Desktop Dual Core on 939.

Donnie

[Donnie27]

I don't mean to the point where there is a shortage, I'm just saying that heat I think would cause a few yeild issues, nothing major. yeah, AMD will have abit of troulbe I think as well, I just won't be as heat related. Thats my opinion

jjcom

"AMD will have abit of troulbe" that's an understatement if there ever was one. I respectfully disagree. AMD and HP can bare the brunt of low yields for Servers but this would kill them for the desktop shipments. If AMD weren't having problems they'd have certainly built more Hammers.

Donnie

[Donnie27]

Yes and I'd love to have a Intel's problems while they cleared 2.7 billion in profits during the 4th quarter.

Donnie

[saaya]

Yes but that sounds more like excuses to me. There are great choices that can be made if you want an Intel processor. The key is "want". An Intel User is not "stuck" with any one type of Platform. You don't need to mess with the "joke" i9xx anyway.

There are Canterwood Socket 775 mother boards. If you don't like DDR2, there are i915s with DDR1. And they still have Better I/O Controllers, glitchless USB and Firewire, and the only real joke is just how much money Intel made while AMD got red figures for the 4th quarter. Some guy joke that Intel made more money in 12 days than AMD made all year.

Donnie
wow, more usb slots and faster firewire slots! omg! :D

you might stick with intel even though its slower and more expensive just because you WANT intel, but i think kunaak and most other guys dont care that much about the brand but about the price and performence.


Please try again?

No, that's not what was expressed! The poster posted it as if Socket 939 CPUs fit into Socket 754 or Socket 940 and will still be compat with M2. If not the folks need to stop using terms like, "Always stuck to one socket longer", Antone from Xbits labs or the posts here. So this is just plain wrong, not an opinion in this case. Too many sockets is too many sockets, doesn't matter if it's Intel or AMD shipping them. Sc-754 didn't live any longer than the 1 year old Sc-423 lived:mad: This eventhough AMD implied it would live longer. Hammer has seen 3 sockets in less than two years.

socket 754 is still alive and will stay alive for quite some time afaik.
when 940 came out everybody know there would be 939 boards later on and that 940 would end up as the server plattform.
and you can still buy 940 cpus and there will be more 940 cpus in future as well, even dual core cpus...
so amd has 3 sockets wich is confusing, but they are all alive and you will be able to get new cpus for them for quite a while afaik :)


AMD will be doing the say thing with Desktop Dual Core on 939.

Donnie
what do you mean?


Yes and I'd love to have a Intel's problems while they cleared 2.7 billion in profits during the 4th quarter.

Donniewhy dio you keep comming up with that? :confused:

hey, nvidia gained more money than ati this quarter! imma buy an nvidia card even though the latest ati card costs less and performs slightly better!
:stick:

and donnie, please stop making a single post for each reply and add it all to one post to save space, thx

[JNav89GT]

and donnie, please stop making a single post for each reply and add it all to one post to save space, thx
He's HyperThreading! Don't bag on the man :cool:

[jjcom]

"AMD will have abit of troulbe" that's an understatement if there ever was one. I respectfully disagree. AMD and HP can bare the brunt of low yields for Servers but this would kill them for the desktop shipments. If AMD weren't having problems they'd have certainly built more Hammers.

Donnie

lol Agreed...it was an understatement. AMD will have problems. Thats garenteed to be a fact lol

jjcom

[Donnie27]

wow, more usb slots and faster firewire slots! omg! :D

you might stick with intel even though its slower and more expensive just because you WANT intel, but i think kunaak and most other guys dont care that much about the brand but about the price and performence.


OMG! Now you pretent to not be able to count LOL! The AMD processors I'd want cost more than the Intel processors I looked at. We've been through that already. You might want to stick to pinning Bogus complaints on me, like reposts or anything else that doesn't go along to get along. Slower, yea rightttttt!

I'll only say you left out the part about going from 266MB I/O to 2GB, improved buffers, newer DMI, doing away with having to put Gigabit on the north bridge since the the south bridge was choked out, and up dates that include better built in sound that even nVidia will use, hehehehe and much more but hey, that might get in the way of your AMD biased spin uh? WOW! :stick:

The rest is really not worth replying to. Like, 754 is dead and has been since 939 launched. Nforce 3 150 is dead since it is slower than VIA. :slobber:

Donnie

[Donnie27]

lol Agreed...it was an understatement. AMD will have problems. Thats garenteed to be a fact lol

jjcom

http://www.theinquirer.net/?article=21073


Intel Centrino brand weak, AMD thinks
AMD appears to have considered - hey it still may be considering - launching a series E low power chip that it will brand Ivanhoe. As for the name of a processor - well AMD toyed with this idea quite a bit before finally plumping for the [Cen]Turion name. It had considered launching chips called Movara and Travera, but the focus groups didn't seem to love those names very much.

Centrino, maybe folks with think Centurion is Intel as well? Oh Brother!

Donnie

[matt9669]

The name of the mobile proc is actually "Turion", the [Cen] was added in the above quote as a reference only. But, still seems cheesy ;)

[perkam]

The dual core Intel's may not be on the 775 platform (all the more reason not go 775, hence the absence of dual-core info on Intel's roadmap) but the dual-core AMDs will not be on the 939 either. Only the s940 server chips will be able to accomodate new dual-core platform without any changes.

Perkam

[matt9669]

The dual core Intel's may not be on the 775 platform (all the more reason not go 775, hence the absence of dual-core info on Intel's roadmap) but the dual-core AMDs will not be on the 939 either. Only the s940 server chips will be able to accomodate new dual-core platform without any changes.

PerkamAMD has publicly stated that current S939 platforms will support dual cores - I don't remember the details but I can look them up if you like. With 775 as new as it is, and built to handle higher current loads, I highly doubt Smithfield will NOT be on 775.

[Donnie27]

The name of the mobile proc is actually "Turion", the [Cen] was added in the above quote as a reference only. But, still seems cheesy ;)

Yes, just the INQ joking around like you said.

Donnie

[saratoga]

Why a new chipset is required makes no sense - take the PC-DL, which runs off the i875 (and runs very well, I might add). New motherboards, that I can understand . . . the power circuits might not be able to handle 130W and/or the VID requirements of Smithfield, even on the S775 Prescott boards.

Smithfields a pretty big hack from what I gather, so they probably had to fudge the way the bus works to accomidate having 2 cores talking through the same socket pins.


"AMD will have abit of troulbe" that's an understatement if there ever was one. I respectfully disagree. AMD and HP can bare the brunt of low yields for Servers but this would kill them for the desktop shipments. If AMD weren't having problems they'd have certainly built more Hammers.

Lets be realistic here. Everyone is getting killed on yeilds at 90nm, not just AMD. Remember the Fall 03 Prescott launch that never happened ? Or the delay for Dothan ? Then theres IBM's endless problems with the G5, and AMD's slow rollout of Winchester.

The difference between IBM/AMD though is Intel is so huge, and so profitable, no one will notice so long as their present problems do not extend into the long term. And I don't think they will be. Prescott has them in a bit of a hole, and Smithfield isn't going to get them out, but in the long run they've got more then enough options to recover.

[Donnie27]

Smithfields a pretty big hack from what I gather, so they probably had to fudge the way the bus works to accomidate having 2 cores talking through the same socket pins.

Lets be realistic here. Everyone is getting killed on yeilds at 90nm, not just AMD. Remember the Fall 03 Prescott launch that never happened ? Or the delay for Dothan ? Then theres IBM's endless problems with the G5, and AMD's slow rollout of Winchester.

The difference between IBM/AMD though is Intel is so huge, and so profitable, no one will notice so long as their present problems do not extend into the long term. And I don't think they will be. Prescott has them in a bit of a hole, and Smithfield isn't going to get them out, but in the long run they've got more then enough options to recover.

You're always the voice of reason!

Two power hungrey Presscotts on one die will not only require a New Motherboard, but a heafty PSU and Giant Proc Coolers that come with BTX as well. I think I'd rather wait on 65nm Dothans or the Desktop version of Yonah (that's what I meant by waiting from Intel).

Donnie

[lutjens]

The difference between AMD and Intel is that AMD knows how to do 90nm correctly. Their Winchesters show a marked decrease in operating temperature. I think they recognize the value in not reinventing the wheel here...as Intel does not. Intel would have been far better off to simply transplant the Northwood to 90 nm and be done with it. No screwing with the pipeline. Intel thought they could kill a number of birds with the same stone by adding all the features to Prescott (SSE3, EIST, EMT64) at once and by adding a longer pipeline. I guess they figured this move would decrease the revisions they would have to make in the future and allow them to activate the features at will (or marketing's discretion) :rolleyes:.

The extra L2 cache that Prescott has does little to help it's performance and is simply a bunch of heat-producing transistors that add to the chip's thermal problems. The other "features" are little more than gimmicks, IMHO. They do little for performance. Dual cores will only compound the thermal problems and doubtlessly require some exotic cooling solutions.

AMD, on the other hand, simply took the Hammer core and stuck it on a 90nm process (with very few changes). This simplistic approach could very well be the reason why AMD is having considerably more success with 90nm than Intel.



Smithfields a pretty big hack from what I gather, so they probably had to fudge the way the bus works to accomidate having 2 cores talking through the same socket pins.

A very interesting comment, indeed. I agree completely. AMD has always planned to move to dual core CPU's and designed the architecture of the Athlon 64 with that move in mind. Intel appears to be playing catch up here, trying to shoehorn their current architecture to work. Much like they did with EMT64...:rolleyes:

Bottom line...Intel is feeling the pinch and seems to be rushing their products out the door to compete with AMD, trying to force their current line of products to do things they weren't initially designed to do. This has been shown clearly with their rush to include EMT64 into Prescott and now with their large push toward dual cores.

Personally, I think AMD's current architecture is more robust and as it was designed with the dual core and 64 bit from the get go, it will likely perform much better than Intel's ad-hoc solutions. Intel will doubtlessly get their dual cores working and eventually get the chipset support snafus sorted out...but the end result will still likely not be as good as AMD's purpose-built solutions.

My $0.02...;)

[Donnie27]

The difference between AMD and Intel is that AMD knows how to do 90nm correctly. Their Winchesters show a marked decrease in operating temperature. I think they recognize the value in not reinventing the wheel here...as Intel does not. Intel would have been far better off to simply transplant the Northwood to 90 nm and be done with it. No screwing with the pipeline. Intel thought they could kill a number of birds with the same stone by adding all the features to Prescott (SSE3, EIST, EMT64) at once and by adding a longer pipeline. I guess they figured this move would decrease the revisions they would have to make in the future and allow them to activate the features at will (or marketing's discretion) :rolleyes:.My $0.02...;)

The difference between AMD and Intel is Intel did its own and AMD depended on IBM for its process. Not an opionion but fact. Only here can something so out in left field could go unopposed.

http://www.hardwareanalysis.com/content/article/1686.2/

""The move to 90nm technology has allowed Intel to cram a comparatively huge amount of cache memory onto the Pentium 4 die. Prescott improves on the previous Northwood processor by boasting a huge 1MB L2 cache. Despite the larger cache, which helps to drive Prescott’s transistor count to over 125 million, the processor’s physical size remains manageable at only 112 square millimeters – roughly 50% smaller than Intel’s first Willamette Pentium 4 with its tiny 256kB L2 cache"""

90nm works for Intel and Dothans prove that. Yes Intel screwed up, but 125 mil trannies, in a area (112mm) smaller than a 131mm Northwood with 55 mil trannies. AMD added what? I loved to see what temps a 90nm Northwood 55mil would get. How big would it be? Then we could make valid comparos. :slobber:

Donnie

[lutjens]

Intel added some excessive stuff (requiring the large amount of transistors) to the Prescott. AMD didn't add anything to the 90nm chips, as they saw that doing so would negate the benefits of increased production capacity (due to 90nm) and the lower power comsumption that the 90nm chips offer. A larger die doesn't add profitability. AMD is already selling the Athlon 64 as fast as they can produce them. The move to 90nm assists them in this regard, as they can produce more dies per wafer. It also helps offset their limited production capacity (as compared to Intel).

The Athlon 64 has 106 million transistors and the Winchester (512kb L2) has 68.5 million transistors. As you said, Prescott is well over 125 million. Add 128kb of L1 to Prescott and see where the count ends up...the Athlon 64 is therefore much more efficient than the Prescott is terms of performance per transistor (my $0.02).

Dothan works very well...due in no small part to it's extremely efficient design. It is an excellent CPU that runs VERY cool, even though it has way more transistors than even the Prescott (144 million). Even the Extreme Edition has 178 million transistors and is still cooler running that Prescott at the same clock...this tells me that something is either very wrong with Intel's 90nm process when running at high clock speed or the "improvements" added to Prescott has a very detrimental effect on heat production. The only way to be sure would be to see a Prescott core made with the 130nm technology and see how hot it was. One can speculate that Intel's rush to include EMT64 on the Prescott hurt it heat-wise (as the P4 never was intended to run 64 bit).

Odd that a processor (Dothan) with more transistors runs so much cooler, yet performs nearly as well or better (in some cases). The clock of the Prescott doubtlessly is the cause, but it requires a high clock to make up for it's inefficient design...an inefficiency that was compounded by the increase in pipeline length from NW to Prescott. Kind of a catch 22...increase clock to make up for a larger pipeline and make a lot more heat in the process. Why not just keep the pipeline short? The only reason I can think of is that the P4 design was maxing out and Intel didn't see enough of a clock speed gain from Northwood at 90nm to compete with AMD's clock speed bumps (and substantial performance improvements as a result).


The difference between AMD and Intel is Intel did its own and AMD depended on IBM for its process. Not an opionion but fact. Only here can something so out in left field could go unopposed.

IBM did not design the Athlon 64. They did, however, pioneer, the SOI technology that the Athlon 64 depends on for it's ability to scale in clock, yet remain a short pipelined chip. AMD had trouble with it initially and turned to IBM for help in getting SOI to work right. A small change in pipeline length was made, but only from 10 to 12 stages. The pipeline would have had to have been made much longer to scale the Athlon 64 in clock without SOI and slowed the chip significantly. SOI is solely a fabrication process, and has nothing to do with architecture. I don't see how SOI can be considered "so out in left field." It allows clock speed increases while not having to deal with the performance-killing side effect of a long pipeline.

Or do you consider AMD to be cheaters (for lack of a better term:)) for using a new technology developed by another company (with their permission and assistance) in their product? I prefer "innovative"...;)

[jjcom]

AMD knew it needed helped and asked for it. Intel didn't. So from that point of view AMD was "big" enough to know it couldn't do something on its own. It was wise to ask for help. Intel mesed up on the Prescott, the pipeline length the design of the core and what is used to make it work well. It uses Strained Silicon...which, correct me if I'm wrong, Intel came up with. Its not very refined. It increased current leakage, which is part of the reason why the Prescott runs so hot. I remember reading something about 65nm could possiblely make more heat if the SS and other design flaws weren't fixed

jjcom

[Donnie27]

IBM did not design the Athlon 64. They did, however, pioneer, the SOI technology that the Athlon 64 depends on for it's ability to scale in clock, yet remain a short pipelined chip. AMD had trouble with it initially and turned to IBM for help in getting SOI to work right. A small change in pipeline length was made, but only from 10 to 12 stages. The pipeline would have had to have been made much longer to scale the Athlon 64 in clock without SOI and slowed the chip significantly. SOI is solely a fabrication process, and has nothing to do with architecture. I don't see how SOI can be considered "so out in left field." It allows clock speed increases while not having to deal with the performance-killing side effect of a long pipeline.

Or do you consider AMD to be cheaters (for lack of a better term:)) for using a new technology developed by another company (with their permission and assistance) in their product? I prefer "innovative"...;)

Oh, I agree until you got to this part, I never said IBM designed Athlon64, I did say, IBM came up with the process and fought off two other companies over SOI, not AMD, IBM orther inventions and patents include Strained Silicon and the first working processor demoed with CU for what it's worth.

http://www.chip-architect.com/news/2003_09_21_Detailed_Architecture_of_AMDs_64bit_Cor e.html

I've read tons about the Hammer core and yes, short Pipeline Processors have both good/Stong points and not-so-good/weaker points. He also shows Yamhill or Intel's take on AMD's x86-64, as I told jjcom, I'll not call it by Intel's name, just as I call SSE1-2-3 just that and nothing else.

It get's old seeing many folks Hype the good and pretend there are no "not so good things" on these same "Wonder Processors". Short Branchy unoptimized = AMD, Streaming Multi-Threaded Optimized = Intel, it's as simple as that.

NO I do NOT consider AMD cheaters, I've said many times, everything they've used to make their product better is smart move. Same goes for Intel. I don't expect AMD or Intel, IBM, nVidia, ATI or anyone else to come up with every piece of technology they'd need, that's impossible IMHO.

Maybe I'm getting old, but I cheered for AMD in their Court battles against Intel. I cheered for Intergraph as well against Intel. I called VIA crooks when they Warezed off Intel, Sides? Sorry, I don't see $850 and $960 dollar processors, made by anyone, being as great! I paid as most that much for my whole computer two years ago. Should this stuff be getting cheaper? Same goes for the 3400+ and thier Intel Counter parts. I'd much rather overclock or get a Dathan after their prices come down.

Donnie

[Donnie27]

AMD knew it needed helped and asked for it. Intel didn't. So from that point of view AMD was "big" enough to know it couldn't do something on its own. It was wise to ask for help. Intel mesed up on the Prescott, the pipeline length the design of the core and what is used to make it work well. It uses Strained Silicon...which, correct me if I'm wrong, Intel came up with. Its not very refined. It increased current leakage, which is part of the reason why the Prescott runs so hot. I remember reading something about 65nm could possiblely make more heat if the SS and other design flaws weren't fixed

jjcom

IBM came up with Strain Silicon, not Intel. Yes, Intel did mess up, they got greedy and skipped too far ahead. There should have been something like a NorthScott or PresWood prior to Prescott. Problems could have been found and corrected before Prescott went into full production. Real World Tech and other sites have talked about many problems within Prescott's Core.

First of all, AMD only asked for help after many failures and a year's long delay. That's not my opinion but what happened. AMD even tried to hide that fact that they were paying IBM 400 million. I remember AMD Fans blasting Tom's Hardware and calling them Intel Sluts for posting the first reports of this. Even Kyle at HarOCP got slammed for doing the same. I was shocked at how many folks thought AMD came up with SOI!

Donnie