Scarchunk

Distinguished
Mar 9, 2006
328
0
18,780
Found this over at Anand. Chinese site, unverifiable,etc., you know the drill. Anyways looks freakin incredible. 8 cores with hyperthreading=16 threads!

http://www.chiphell.com/?action-viewnews-itemid-178

2_200802021201392.jpg
 

PSYCHoHoLiC

Distinguished
Oct 10, 2006
246
0
18,680
Nice!

I'd like to get my hands on the quad when they come out, but an 8 core? Thats just overkill, I have a quad now and hardly ever do anything that uses it.
 

Kamrooz

Distinguished
Feb 8, 2007
1,002
1
19,280
8 cores with hyper threading = 16 cores.

I think it's bull thought, If memory servers me right, we won't see octa core till 32nm. only dual and quad will be available on 45nm till the 32nm refresh. Correct me if I'm wrong here, but I know for a fact (at least I think) this early for a octal won't happen.
 

Slobogob

Distinguished
Aug 10, 2006
1,431
0
19,280
For those that keep saying it is 16 cores, it is not. It's 8 cores, 16 threads. Stick to the terminology or you will confuse people and in the end confuse things yourself. Still, it is an impressive feat.
 

Scarchunk

Distinguished
Mar 9, 2006
328
0
18,780

Kamrooz

Distinguished
Feb 8, 2007
1,002
1
19,280


Typo, I meant to say threads.




Lol..Indeed it is, I am 99.9 percent sure, so I just wanted to leave an out for me =P. Regarding the screenshot, that's is indeed right, But I remember hearing we won't see octal till they go 32nm, so the Westmere process (which is the "tick"), will be octal.

But yes, if that is indeed a dual socket mobo, then it makes sense, but we won't see an octal single unit till 32nm, 99.9 percent sure on that.
 

Scarchunk

Distinguished
Mar 9, 2006
328
0
18,780


You may be right (I'm hoping your wrong) but if you scroll down to the Nehalem section of the Intel fact sheet I linked to you will see the following:

Intel.jpg


It states that Nehalem will launch in 08 with up to 8 cores and 16 threads!
 

NightlySputnik

Distinguished
Mar 3, 2006
638
0
18,980


Well, that's exactly what came to my mind when I saw the screenshot. CPU-Z most probably doesn't support NEHALEM cpu yet. Still, an ES at 2.13GHZ make sense.

One last thing. Wasn't Nehalem suppose to have on-die memory controler? If so, I think the a 533mhz link is kinda slow, unless it's 128 OR 256-bit wide.

Anyway, thanks for the link... but I'll keep my judgement for mid-end of the year when we'll have actual product in hand.
 

cnumartyr

Distinguished
Nov 3, 2007
2,287
0
19,780


What bugs me is how they "hid" part of the CPU-Z.

Nehalem has an on-board memory controller, but no FSB. It uses QPI. It should read more like an AMD CPU-Z. 0.975v seems low unless C1E is active.

And yes this is a dual socket motherboard. Read the entire CPU-Z and it answers your questions. This also means it should have QPI.
 

Kamrooz

Distinguished
Feb 8, 2007
1,002
1
19,280


Nowhere in that image does it say 08, but we all know Nehalem will see a LATE 08 release, but expect it ONLY to be extreme editions as they always do. Then followed by the mainstream/midrange lineup around 2-3 months after.

Also, it states 1 to 16 threads, and 1 to 8 cores. That does not mean you will get them all at launch, the 32nm shrink will still indeed be a Nehalem architecture, just shrunken down to 32nm in their "tick" process. So that will be the octal core cpu with 16 threads, just because it says 1 to 8 cores or 1 to 16 threads doesn't mean we will get it fast, and it still is under the Nehalem architecture. I do hope I'm wrong though, it would be nice to have 8 cores, but I don't think it'll happen till 32nm, even so, a new architecture that is said to contain a even larger gap in performance compared to the transition of netburst to core is quite a feat, even if it is just 4 cores with 8 threads, it'll be a huge jump in performance....So waiting for the octal won't be that hard considering the claims they are making about Nehalem.

But yea, don't expect to see an octal till 32nm, 99.9 percent sure on that.
 

Scarchunk

Distinguished
Mar 9, 2006
328
0
18,780


Yeah the 08 part was cut off b/c I couldn't fit it all in one shot. It's in the link. Anyways, I'm sure you're probably right about the dates and I'm being overly optimistic, but I'm not so sure that the 8 core variant of Nehalem will be on 32nm and not 45nm. Everything I've linked to including Intel's IDF coverage supports the 45nm 8-core Nehalem theory.
 
Too bad we do not have any real idea about Nehalem. We can say this is real or its fake. You can tell its a dual socket mobo as the CPU-Z processor selection part is useable unlike mine which is kinda greyed out.

I am pretty sure that Nehalem will scale from 1-8 cores. 8 cores will probable come out a bit later but I am sure they are trying for 45nm. I will be waiting until Mid 2009 so that they have a nice stepping and 32nm.
 

Grimmy

Splendid
Feb 20, 2006
4,431
0
22,780
:oops:. o O (heh.. I just counted what was on the taskmanager)

BTW... wouldn't that be a totally different socket MB too?

And.. how would you be able to tell the hyper threaded cores from the real ones?

Gawd.. I can see the problems now..

from a distance... o O <(hey.. my taskmanager only shows 8 cores... where's the rest of 'em?)
 

Grimmy

Splendid
Feb 20, 2006
4,431
0
22,780
Well.. I still have my P4 3.0 HT running. The only logical guess on how to tell the hyper threaded ones would be:
G#-Graph number
G1----G2----------G3-----G4-------G5-----G6---------G7------G8-----G9----G10
[Core0-HTcore0] [Core1-HTCore1][Core2-HTCore2][Core3-HTCor3][ECT-ECT]

I think that would be it, since the 1st graph should have been the real core, the 2nd HT off my old P4 3.0 HT CPU.
 

Slobogob

Distinguished
Aug 10, 2006
1,431
0
19,280

On the P4 there was no difference between real and fake cores. Differentiating between the two on a single socked dual-threaded system was and still is totally pointless too. In addition to the HT hardware there is certain software required which will make the differentiation between real and fake but it can switch between them. One moment Thread 1 is the core, the next moment Thread 2 is the core. In theory both could be the core too, but that would require a multi-core and HT aware application running that uses both threads wich are both synchron. As a matter of fact the CPU saves the architectual states of both threads.
On a dual socket system or a single socket system employing more than 1 core it will matter which threads belong to which CPU core though. I'm quite eager to see how intel implemented it - ie. how the load balancing works with different tasks, power management and performance in mind. This is where it cang get quite complicated. If there are 2 cores with a total of 4 threads running but only 2 programms. Will each core get a programm for maximum performance or will the programms run on a single core with both threads and the other core goes into power saving? Will the user have any control over it or will the software (NUMA) handle it all? I can't wait to see it tested.
 

xnamerxx

Distinguished
Aug 22, 2006
60
0
18,630
whats bugging me about this picture is that windows xp cannot support that many threads and should be unable to indentify it as such. As well as if you look at the cpu its being identified as a 4 core system with 8 threads but somehow windows is seeing 16 threads unless this is a dual cpu system. Also the rated fsb is too low and with the onboard memory controller is should be identifiying itself like a amd cpu. It also isnt listing a correct amount of l2 cache by only seeing 256kb. Im sorry but I dont believe this is a real cpu pic
 

cnumartyr

Distinguished
Nov 3, 2007
2,287
0
19,780


I agree with you that it's a fake.. But:

It is a dual CPU system, notice the select processor isn't grayed out.

133 MHz Bus is logical for an ES and odds are if CPU-Z did actually recognize it it wouldn't know to put it up as a QPI Processor and is just using 4xBus to still show a FSB.
 

yomamafor1

Distinguished
Jun 17, 2007
2,462
1
19,790
I'm not pretty sure its fake...

1. CPU-Z has not implemented ID string into their latest edition.
2. Its even very difficult to get a peek in Intel plant.