Someone in a chinese forum postet that new Roadmap:
http://img323.imageshack.us/img323/8...roadmap3sn.gif
Conroe : :slobber:
Printable View
Someone in a chinese forum postet that new Roadmap:
http://img323.imageshack.us/img323/8...roadmap3sn.gif
Conroe : :slobber:
2.66GHz starting frequency, very impressive.
Note the ditching of SMP (aka HT)
Only 2mb shared? I was expecting 4 and 8 for XE.
great.....more chipsets.
So will i975x support Conroe or not!?
SO when Is Intel going to do something intelligent and Make One CHIPSET that WILL beable to use ALL 775 CPUs?
i985X what the heck! I won't got Conroe XE for sure!
Im thinking that conroe XE will be quad core from seeing how it lacks HT.Quote:
Originally Posted by NiCKE^
im sorry, but i wouldnt trust a roadmap that gets preslers cache size completely wrong...maybe its a bit old and out of date?
it says 2X1, but we know its 2X2
Sucks no HT.
read my postQuote:
Originally Posted by Pinnacle
efficient Pentium M-derived architecture at the same insane clockspeeds as Netburst?
.... :eek:
Too bad they're still using the aging Netburst architecture for the northbridge.... 1333Mhz = 10.6GB/s max theoretical bandwidth between the CPU and chipset. Actual throughput will probably be more like 8GB/s.
Out of that 8GB/s, about 4GB/s is used for PCIe communications in a SLI setup, which leaves a meager 4GB/s for memory bandwidth. Dual channel DDR2 will be able to provide 10.8GB/s of bandwidth EASILY, with 2xPC5400 Micron D9 sticks at 3-2-2 timings.
The only hope is that if one is able to overclock the FSB from 1333Mhz to 2Ghz, which would boost bandwidth from 10.6GB/s to 16GB/s. I highly doubt the northbridge will be able to do that though. There's just too many connections going to it: The CPU, PCIe, memory, AND southbridge.
the conroe xe at stock is equal to an amd x2 @ about 3.6ghz lol
thats total pwnage
but the cache should be 2mb on all models as someone said before, so dont get ur hopes up for those clock speeds.
northbridge isnt used for pci express communications....and there is no sli on intel setups(lol, unless u use the crappy nvidia chipsets which are totally gay)Quote:
Originally Posted by HKPolice
i dont know why you would say that...conroe has yet to be benched. performance can only be guessed atmQuote:
Originally Posted by grimREEFER
wow, you're so knowledgeable!Quote:
Originally Posted by grimREEFER
HAHAAHAHAHA :lol:Quote:
Originally Posted by grimREEFER
Please die.
*drools over conroe XE* 4MB L2 Cache!!! holy mother
Please let Evolution kill you :slap:Quote:
Originally Posted by grimREEFER
all conroe's were supposed to be 4MB L2 (with some 2MB units thrown in)...which is another reason i question this...Quote:
Originally Posted by ingentingmendeg
no:slap:Quote:
Originally Posted by AkXb70
smart....:rolleyes:Quote:
Originally Posted by Pinnacle
EDIT: actually though, initial P8 chips werent supposed to have HT, although ALL later iterations will (aka MMX)
http://img482.imageshack.us/img482/5...diagram6yg.gifQuote:
Originally Posted by grimREEFER
erm.....:stick:
really impresive lets see benchmark to see if they are clock for clock faster than AMD A64
thatl probably be a few months down the road...although my guess is theyll be at least as fast as yonah (due to the very small increase in pipeline length and the tweaked execution engine)Quote:
Originally Posted by leviathan18
my hopes though are for something thatl smash yonah clock for clock (but..thats hopes...)
If it's rated for upto 3.2Ghz STOCK, then I'm sure even the lower binned chips can OC to at least 3.6Ghz on good air (assuming FSB doesn't become a limiting factor). Even if that's equal to ~3.4Ghz A64 in performance, I don't see how AMD's dual core 90nm process can even come close on air.
Don't get you hopes up though, AMD is going 65nm by end of 2006. Assuming some delays, Q1 2007 at the latest.
that if they are as fast as A64 clock for clock if the cache is slower and they have longer pipelines it wont be as fast as yonah or A64 perhaps it will be fast but not that fast im guessing 15~20 more than actual AMD offers until AMD gets the 65nm (or 64nm) down the road DDRII and the new process for the silicon
I think this thread is getting a little gooeey...i've stuck it so that everyone can see :)
WHOA !!! No HT On Conroe's ??? No conroe for me...i'm gonna settle down with a nice presler/cedar mill chip...or if they drop in price enough, a cheap 6xx pressy.
Perkam
Why do you even need it the only comparable chip is Presler Extreme Edition available for 999US bascially shutting out the bulk of people.Quote:
Originally Posted by perkam
Allendale is already better then the Single Core Cedar Mill with Hyper Threading as it is a Dual Core processor with 2MB of cache.
Conroe and Conroe XE will have 4MB LV2 cache and 1066FSB and 1333FSB respectively. Conroe XE is doubtful to be slower then Presler XE in anything short of really heavy multitasking, though I believe the clockspeed should be enough to compensate for Conroe.
Conroe needs no HT to perform as well in Multi-Tasking than an actual Presler EE. ;)
Compare:
Conroe: 14-stage efficient pipeline; Netburst: 31-stage pipeline (Prescott and later)
Conroe: 4-issue out-of-order execution unit; Netburst: only 3-issue out-of-order
Conroe: 4 MB shared cache; Netburst: max 2x2 MB (which is slightly worse than a shared cache)
Conroe(Merom) also will have improvements in Branch Prediction, Target look-a-side Buffers, FP&SSE executions, Direct L1 to L1 Link, L2 bandwidth optimizations and a new chipset.
Netburst needed HT because of the long (inefficient) pipeline, Conroe would not really perform much besser with HT.
So Is HKpolice wrong?Quote:
Originally Posted by 3NZ0
That's with 1066Mhz FSB which is 8.5GB/s max theoretical bandwidth.Quote:
Originally Posted by Nosfer@tu
Current SLI setups don't show much of a improvement going from 8x PCIe (2GB/s each = 4GB/s total) to 16x PCIe (8GB/s total). So I assumed that the PCIe load on the i965 chipset would max out at around 4GB/s
OH MY GOD - NO HT :p:
Actually NetBurst while working could have some blocks unused(even 100MHz FSB CPUs has those). That`s why Intel made HT to use`em (that also explains why with HT on some apps run slower than w/o it). Conroe won`t have such blocks - so no HT for it ;)
hokay, so an i975 motherboard can cover everything but Conroe XE. I can live with that.
When they list the chipset bit though, I'm assuming that i975 can cover anything that needs i965, i955, i925 etc... If not then that's just retarded
Why is everyone worried so much about chipsets ? Third party chipsets like ATI's RC600 series and others should also be supporting Conroe architecture....and that'll be more of the one chipset for all Intel processors kind of solution.
http://img521.imageshack.us/img521/4...telrdma1ll.jpg
More Info here: http://www.hardwaregate.com/intro.php?id=363 ... where it actually predicted the x600 coming as an onboard vid solution in July 2005...a story confirmed by The Inq on 29th of December: http://www.theinquirer.net/?article=28610 .
Though push those release dates at least 6 months ahead as it seems ATI's working to get the RD580 AMD Chipset out right after the R580 launch.
Perkam
I've never had an Intel rig, so I'm really looking forward to what Conroe can offer. My Venice 3000+ really stinks at multi- tasking (even overclocked), and I find the X2s a little expensive. But I fear Conroe cpus will be expensive too, are there any hints whatsoever about pricing? And will Presler and Cedar Mill co- exist along with Conroe, or will they be fased out?
I think Presler and Cedar Mill will fase out, because no one wants longer a Netburst CPU when Conroe is avaible. :D :banana:
I'd definitely say this is fake... If they got presler's info wrong they more than likely have conroe wrong too.
dubious roadmap in my opinion:stick:
Dudes:
Intel admitted themselves: HT sucks ass.
Experts have been saying for years (before HT was available) that HT even at optimal performance would only be a 25% increase in speed.
Intel hasn't come close to optimal, even with software compiled to perform best under HT the speed increase is only about 10-15%.
Also HT was needed to prevent pipeline stalls inherent to the netburst architecture. If one would implement HT on a short pipeline architecture the speed increase would be close to 0% and the netto speed change may even be negative.
In real life HT gives almost no speed increase (more a feeling then a measurable increase) and at dual core systems this feeling is also gone (a dual core system responds so fast to user input the feeling is fast no matter what conditions).
HT was more a futile attempt to solve some of the problems inherent to the netburst architecture and was used by the (very good) marketing department of Intel to praise their CPUs.
Conroe will kick ass, no other Intel CPU before it will come close to the performance.
As for the roadmap: Time will tell all you need to know.
argh, i misread ur post and i mixed up some :banana::banana::banana::banana: lol.....because i misread ur post lol.:slap:Quote:
Originally Posted by HKPolice
Intel said HT sucks ass, hey?? Intel said that??
can you link me
Quote:
Originally Posted by Nosfer@tu
well it looks like the pci-e connection goes to the nothbridge, doesnt it?
on topic: HT =:slapass:
thanks for the info on HT, Thorry.
Offcourse they didn't shout it from the roofs, but various sources have stated this is the idea about HT and Intel has admitted it to be less then a full succes.Quote:
Originally Posted by Pinnacle
Some links (if you Google I'm sure there will be a lot more about HT)
http://news.zdnet.co.uk/0,39020330,39237341,00.htm
http://arstechnica.com/paedia/h/hype...reading-1.html
http://blogs.msdn.com/slavao/archive...12/492119.aspx
http://www.2cpu.com/articles/43_3.html
http://www.dotnet247.com/247referenc...56/284993.aspx
Obviously:rolleyes:Quote:
Originally Posted by Thorry
I just wanted an intel paper that said, hyperthreading has no advantages
If it was so bad, as you say, why then is it on the 955 EE?
Resent benchmarks showed that their WAS an improvement with it enabled, and it did come close to matching the 4800+
Actually Hyperthreading is just the Begining.. Thread Level Parallelization is the next Big thing..
Marketing.... The EE is not that much faster but way more expensive, they throw everything they have onto that just so they can say it's uber.Quote:
Originally Posted by Pinnacle
I can also make a benchmark showing my new blabla feature increases total system performance 50%.
Real life improvement 0, even Microsoft says: Please turn of HT, it slows down your sh*t.
HT is a lame attempt, parallelization however is the future. Just take a look at cell processors, that's the way to go.
Quote:
Originally Posted by Thorry
HT still has some performance increase.
Parallelization is, nobody is arguing that.
But Parallelization will not Hit the Mainstream for atleast another Year.. So until then It IS not a benefit but It is a Future Proof technology that has yet to prove it's worth
Some fresh info to you guys :)
The InqQuote:
Since Intel is sending out Conroe XEs at 3.33GHz to friends and people buried in NDAs until you can't find any trace of their existence, around the end of this month, we think things are in great shape.
If Conroe is based on Yonah architecture...3.33ghz should be nuts :slobber:
Perkam
considering the increased pipeline and higher latency cache compared to dotan, but countered by a four issue wide core with various architectural improvements and a faster front side bus, I believe this would beat even a 3.4ghz a64, which is the max we'd be able to see on a select bin part, even if the tweaks amd and ibm are making to the 90nm process yield a 10% increase in clockspeed. However I find it a little hard to believe that such a chip will be with us anytime soon, after all how much extra headroom is there to compete with the amd 65nm move? My bet is that the initial highest speed will be closer to 3ghz, increasing to 3.33ghz when amd shifts to 65nm, and can offer a competing product. Wonder what the TDp of a 3.33ghz part will be - have they achieved that speed by being more liberal with power and heat requirements? thats an overclockers job!
Hmmm if thats true this will certianly give AMD a run for its money, thats for sure. However I worry about the FSB speeds of these chips. AMD will have them beat on the FSB/HTT end of the equation but if they are really 3ghz capable chips AMD might have a seirous problem with the top end
According to Charlie Demerjian of the Inq (one of their more reliable reporters), Intel already has Conroe working at speeds of up to 3.33ghz!
Thats amazing if true, because if Intel releases Conroe/Merom on schedule, the fastest chip AMD will have will be 2.8ghz (I think). The 3.0ghz X2s will probably make their debut towards the end of the year.
Can you imagine a 3.33ghz Conroe? That would be akin to having a 4.3ghz K8, if the rumours of the 30% higher IPC prove to be true (in some benchmarks I'm sure)..
I doubt they'll release at so high a clockspeed though, because that would be overkill!
Also, it looks like the XE models will be working with 1333 FSB and the mainstream parts will have 1066..
It's been so long since I've upgraded, I may just blow my entire tax return on a Conroe setup :D
Source
WTH are you doing at XS? :slapass: :DQuote:
I doubt they'll release at so high a clockspeed though, because that would be overkill!
Intel is going to kick *ss with this CPU, but just wait what AMD comes up with.
What I meant was, that Intel probably won't want to ramp clockspeeds that fast, because AMD is already at a disadvantage by using 90nm process and they don't NEED to introduce a chip at that speed to whoop their ass..Quote:
Originally Posted by Thorry
Then again, the XE is in a class by itself, so perhaps they will introduce it at 3.33ghz. The XE chip will be competing directly with AMD's FX part, which should be 3ghz or so.
The mainstream parts will probably start at 2.6ghz, and top out at 3ghz though most likely..
The Chips are going to get slower and Slower and we are going to see more and More Cores and Less 4ghz overclocks
Wow :slobber:
lol?Quote:
Originally Posted by Thorry
We have a thread for the Conroe...in which this is old news ;)
--Threads Merged--
Check the first post in this thread ;)
Perkam
ive seen this coming for over a year now... amd bumps clocks every few months to keep us happy, but nothing real significant, intel is going to come out with something great, theyll up the volts and max out thier cpu's, then everyone will :banana::banana::banana::banana::banana: about how hot amd's run.
sounds kinda like what happened to intel eh?
Eh, AMD has their new silicon method and 65nm coming next year... They won't have heat problems or power issues, that's for sure.
Next summer is going to be a definite fight on the CPU front!
I wonder if the new Desire of Amd and Intel For multible cores will Cause a Rebirth to Risc
ya, good questionQuote:
Originally Posted by nn_step
not only intel :DQuote:
Originally Posted by Carfax
and there are conroe samples running that speed for months already :D
saaya knows something we don`t
Share some plz :)
all i can say is that conroe will be DAMN fast... sorry :(
Please for the love of God tell me they aren't still trying too keep the Clock speed high..Quote:
Originally Posted by saaya
I would prefer if they dropped the speed and did more for every Clock
saaya will they be i975X compatible ? Or yr birdy uses different type chipset ?
Why do you care?Quote:
Originally Posted by nn_step
Yes the netburst architecture could only work on very high clockspeeds (Prescott should have been 5-6 ghz) and yes that didn't work for various reasons. But a new architecture could not have these problems.
What if the next generation CPUs run at 25 Ghz and still only be twice as fast as the current architecture, why do you care?
Failing of high clockspeeds on netburst architecture is a seperate issue from high clockspeeds. Actually in a FSB environment higher clockspeeds mean more bandwidth which is good.
There are only two problems with higher clockspeeds:
- High clockspeed circuits put out more EMI
- High clockspeed circuits are more vunerable to EMI
That's why clockspeeds need to be kept lower instead of higher, however a compromise needs to be found where that compromise lays nobody really knows. As circuits get smaller and smaller more parallization is possible which leaves more bandwidth at the same clockspeeds. On the other side smaller circuits can be switched faster which means higher clockspeeds.
People should know by now clockspeed says absolutely nothing...
Here is my take on the roadmap:
http://www.xtremesystems.org/forums/...1&d=1137907655
I am hopeful this is correct, would be nice as Allendale can target some of the lower price points.
New info from HKEPC.com
http://www.hkepc.com/bbs/attachments...yIDq6NYhZF.jpg
http://www.hkepc.com/bbs/attachments...SmaqSIsB9r.jpg
So Conroe should still be sitting on LGA775 - that`s good. Today you can buy mobo and 6xx/9xx and then upgrade to Conroe. I guess this year Intel is on "small investments upgrade" road :D
Hopefully those rumors with i975 rev.2 won`t come true or at least current i975 mobos will support Conroe without any issues.
UPDATE
http://resources.vr-zone.com/newspic...7/Conroe-2.jpg
I guess Asetek will make it possible to run Conroe on curent i975x based mobos with simple BIOS update :D...I hope few of brand makers will make it.Quote:
Originally Posted by Visionary from VR-Zone forums
I`m just really curious what kinda mobos are used by people confirming 3.3GHz spec :rolleyes:
New info on Conroe and i975x compatability
Mobos for Conroe chips will have different VRM (voltage regulator module). Of course they will be compatible with curent cpus. Hopefully some hardware mods will allow todays i975 mobos to run Conroe.
Source - russian website
Big disappointment :(
http://www.theinquirer.net/?article=29504
I think I'll go with Presler untill 3.33Ghz Conroe appears....
not really look at the xeon linup - they have their xx60 chip listed as 3ghz, which would indicate headroom above 3ghz. Maybe the e9 series will scale to 2.7 on launch with a 3ghz ee with decent speedbumps to follow. After all this architecture will have to last intel a good while yet, and the e9 series classification may not last too long (move to a higher fsb?). This news is not dissapointing, instead its rather good news, and indicates good overclocking headroom on these chips as well.
Thats more in line of what I have heard/expected. Numerous sources have told me that Conroe has a top end problem similar to Presler and Winchester. Its a great cool running CPU but most have issues past 2.6ghz and a rare few could scrape 3.Quote:
Originally Posted by maratus
Frankly this is far more optomistic than I thought it would be. I wholly expected the top end Conroe chip to be 2.33ghz and not 2.6
2.6GHz will definitely challenge Intel's architecture efficiency against AMD. AMD might have theirs at 2.8GHz by then.
4 issue wide core vs 3 though, should counter the imc nicely.
That, and the ridiculous amount of L2 cache.
2.67GHZ/1066FSB/4MB @ 530US doesn't sound too bad though.
Also with the 3GHZ/1333FSB/4MB available on the server side for 850US, we could easily have an EE for 999US of this same chip but on the desktop.
It would be interesting to see what AMD Athlon FX Dual Core is like by then 2.8GHZ with DDR2, how competitive to say for certain however is hard to say.
News - Conroe, Woodcrest, Merom
Maybe these are some interessting infos...
http://pics.computerbase.de/news/13053/1.gif
http://pics.computerbase.de/news/13053/2.gif
//Edit
http://img15.imagevenue.com/loc4/th_038db_Unbenannt.JPG
http://img145.imagevenue.com/loc113/...Unbenannt2.JPG
http://img144.imagevenue.com/loc31/t...Unbenannt3.JPG
Source computerbase.de (GER)
Wow...the revolution cometh...
Perkam
Looks like E6600 gonna be very popular among Intel users. Hopefully those will be [/]mostly[/i] high OCable.
Also Very good prices
$1300??? They can forget that.
Where did you saw that price ?
E6600 only 315$
if you look on the left of the roadmaps they give estimated full system prices not cpu prices.
shame there is no socket / chipset information that would be usefull.
Quote:
Originally Posted by Cybercat
WoD > Welcome to XS! :toast:
I see no Woodcrest info though?
Will there be no extra incentive except clockspeed to go EE?
Since E6600 will be fairly cheap and lowest bin, it shouldn't OC the most, but maybe good enough?Quote:
Originally Posted by Cooper
So, on a side note, if you assemble a $1500 system, Intel thinks at least $900 should go into the Intel CPU... RIGHT.
Oh sorry, I forgot Woodcrest... i'm gonna fix it... ^^
oh, didn't see that.Quote:
Originally Posted by leejsmith
I mean it`s gonna be like 2.4C Northy ;)Quote:
Since E6600 will be fairly cheap and lowest bin, it shouldn't OC the most, but maybe good enough?
Those numbers look more in line of what Ive heard. Frankly I still think 2.66 is high for Conroe, most of what I know tells me that it will cap off at 2.33 not 66
From VR-ZoneQuote:
Originally Posted by vrzone.com
Hope that comes soon
:slobber:Quote:
Originally Posted by Cooper
As for EE, it should be ~3 Ghz (since highest clocked Woodcrest is 3 Ghz). Which should go up against 2.8 Ghz FX-62. Intel may have a winner there, we shall see.
I'm thinking that XE would be 3.33. Correct me if I'm wrong, but it seems that Intel always releases the desktop higher clocked chips before the server chips. I say we'll see 2.66 and 3.33 at launch and upto 3.0GHz on the server side, then a couple months later we will get the 2.93GHz desktop and a 3.33GHz server.
Thats what seems logical to me.
i guess intel will have successfully made transition to pentium netburst to M tech with conroe.. Still can play "More MHZ the better" game.
edit: Also...Since the core seems to vary from quite low clock speeds to high (EE). What do we reckon the max OC is gonna be on these things? will it go beyond 5 ghz? I apologise if i'm a little behind :/
If you think of the lower clocked chips as being like the low voltage xeons that were clocked at half the speed of the normal chips, intel categorising the chips into different market segments isn't a new phenomena. However when you consider the chips design it will not clock anything like a netburst chip does under extreme cooling. The extreme cooling helps the long netburst pipeline achieve its potential, wheras conroe eliminates the problems of leakage and as such their should be much less wasted potential.
I'm having visions of 4.2Ghz+ Conroe under phase change :slobber:
you will need very good board for to go over 400 mhz fsb :slobber:
Thats not gonna happen. Word on the street was that intel was petrified about how badly Conroe and Yonah was scaling, almost to the point that they thought of canning it and moving to plan C.Quote:
Originally Posted by vapb400
This timeline is easily followable and you can see where I am getting these facts from. The inital samples wouldnt do over 2ghz hence the 1.8 samples. The newer batches just as Presler was about to be released were doing rougly over 2.6 and now its sitting just below 3ghz. 3.3XE is simply not going to happen at all unless Intel can pull a major rabbit out of the hat.
From what I hear now the problem is semi-fixed but it will take atleast .45nm before we will start seeing over 3 on air. They are still suffering the same leakage issues since they still using "dumb" silicon compared to what AMD is using currentally
Hope Conroe XE will have unlock multi...Quote:
Originally Posted by d@rkn1ko
Hmm didn't realize that. The easy clocking of Yonah to 2.7+ in a crappy (?) motherboard with a crappy heatsink is a positive though.Quote:
Originally Posted by Sentential
We are still looking at a Q3/Julyish release?