Leaked Intel slides reveal 8-core CPUs, AVX instruction set

We recently learned that Intel would be (officially) calling Nehalem Core i7 and Centrino Atom, um, Atom. Now, however, we've got a few more related details for you to digest thanks to a stack of leaked presentation slides. The Tick Tock Development Model explains that both Westmere and Sandy Bridge (codenames, of course) will be fabricated with 32-nanometer technology in 2009 - 2010. Moving even further into the unknown, geeks can expect Ivy Bridge and Haswell (both doing the whole 22-nanometer thing) to surface between 2011 and 2012. The Sandy Bridge architecture will reportedly "double the number of cores per die to eight," while a new instruction set coined Advanced Vector Extensions (AVX) will likely get most of the attention. For those of you who haven't conked out yet (you know who you are), hit the links below to get a better idea of what your future rigs will likely house.

[Via Electronista]

[avester @ Aug 16th 2008 9:41AM]

El Scandale?

Ay Caramba!

[Zeth Ethereal @ Aug 16th 2008 9:57AM]

Crazy Bunny! killl it!!

[Mike10010100 @ Aug 16th 2008 10:03AM]

The duck is a scandal?

Translation aside, wtf? 22 nm?

[LanFeusT @ Aug 16th 2008 11:10AM]

carnard is slang for newspaper

[Dapman02 @ Aug 16th 2008 11:11AM]

22nm is the size of the transistors on the processors, what this allows is faster clocking, less power consumption, and cooler running.
Most 65nm Core 2's could only do about 3.5Ghz, but most (including mine) wolfdale or yorkfield 45nm's can do 4.0Ghz+

[thedesolate1 @ Aug 16th 2008 10:13AM]

Hay Dios Mio! what will they think of next?

[pe3ko @ Aug 16th 2008 6:54PM]

1nm 128core!!

[Mike @ Aug 16th 2008 10:25AM]

Sandy Bridge huh?

Sounds like a shaky foundation for such a large company.

[Tym @ Aug 16th 2008 11:12AM]

Yeah, but then they fill it in with Ivy! Makes it sound like the first eight core processors will sort of be thrown together just to get it out, then really optimized later.

[eiki @ Aug 16th 2008 11:40AM]

they had to change the name from Gesher, "as the former codename for this core was Gesher (means 'bridge' in Hebrew), but that codename was abandoned on 17 April 2007 because a political party in Israel is also named Gesher" - it has alleged links to Hamas

[kewl @ Aug 16th 2008 10:29AM]

o.O new architecture...128 bit i hope..

[ethana2 @ Aug 16th 2008 1:48PM]

No.

[Alex @ Aug 16th 2008 10:32AM]

You know, they probably know how to make it already. They are just milking the market.

[tekdroid @ Aug 16th 2008 11:33AM]

I hear Bill G is on 666 cores and laughs at the plebs with 4.

[Steven1350 @ Aug 16th 2008 10:33AM]

AVX Instrusctions? Thats pretty gay, since NO versions of Windows (and as far as I know Linux too) can use that instruction set. This will prolly end up being for servers only :(

[Chip @ Aug 16th 2008 10:56AM]

Don't you think there might be a new version of Windows by 2011?

[Jaimi @ Aug 16th 2008 11:17AM]

Completely ignoring the idea that an instruction set can have any sort of sexual preference, AVX is aimed at 3D - it will be utilized in games and other multimedia applications.

[John @ Aug 16th 2008 11:44AM]

vector instruction sets aren't really all that important for most operating system tasks. The cell proc uses 128 bit instructions, and uses vector operations for doing stuff to sets of 4 integers/floats or 2 doubles at a time. It does speed up stuff considerably when your application fits it, but for your everyday computing, it's not necessary.

[Xavian-Anderson Macpherson @ Sep 23rd 2008 11:05PM]

Any new instruction set introduced by Intel will always be available in Linux before it is available in Windows. Intel has long promoted the Open Source community, and will continue to do so, knowing the large amount of free development that is provided by that community.

Which is why all of you running 32-bit versions of Windows can't use more than 3.25GBs of ram, where Linux has been able to do that in 32-bit for some time now. And besides, Linux doesn't include within it the restrictiveness of the attached Windows licensing. As such, there are also no restrictions on the amount of processor cores any Linux user may and employ. So even home users are able to incorporate the Linux advanced server attributes in their daily lives, while sitting in their living rooms.

Shingoshi
Tyan S4980 Four-Socket Quad-Core
(Ju$t try pricing the licen$ing co$t of running that in Window$: Linux=Free!!)

[Extreme Emo @ Aug 16th 2008 10:40AM]

Looks like I'll be sticking with my Q6600 until Sandy Bridge comes out in 2010. Feeling to cheap to buy a new motherboard right now.

[andrei030 @ Aug 16th 2008 10:43AM]

256 cores in same processor size, that is the future.

[ethana2 @ Aug 16th 2008 1:51PM]

graphics chips really are amazing, aren't they?

[tha-don @ Aug 16th 2008 10:56AM]

8 cores! more cores - more fun!

[VeganFreak @ Aug 16th 2008 11:07AM]

LOL! That honestly made me laugh, nice one!

[T.H. @ Aug 16th 2008 1:07PM]

Six Flags amusement park commercial, the "More flags, more fun" one.

[thatrotierkid @ Aug 18th 2008 9:15AM]

thanks T.H.

/sarcasm

[Billy Gun @ Aug 16th 2008 11:09AM]

92nm -> 64nm -> 45nm -> 32nm -> 22nm what's next? 12nm? 7nm? 3nm? 1nm? 0,6nm?

Love in the year 3000 (B-52's)

[Kjetil T. @ Aug 16th 2008 11:17AM]

The factor is the square root of 2, so the next feature size will be 15mm.

[Billy Gun @ Aug 16th 2008 11:50AM]

Kjetil T., Thanks.. but wow! Tech is getting so far... I just whant to know when will I be able to buy an Mini Pen Drive of 1tb? In 2030?

[paul34 @ Aug 16th 2008 12:02PM]

I wonder how they keep getting smaller. Isn't that already very close to the size of an atom? Or the wavelength of light?

[MioTheGreat @ Aug 16th 2008 12:31PM]

Once we start to get to 16nm, various quantum effects dominate, and the more mesoscopic properties we rely upon break down. You're talking gates 30 atoms wide. There is going to be MASSIVE leakage through the transistor.

It'll be interesting to see what direction things head in. MOS technology (What we've been using for the past 40 years) isn't going to scale after 3 or 4 more shrinks.

[david @ Aug 16th 2008 12:58PM]

Damit. and I am still using 130nm CPU :) Still works fine... :) Image 22nm, that's almost 6 times smaller. Damn, technology is really going fast. I just hope they will require less power and will output less heat. Maybe one day we will not need any fans running on top of CPUs.

[ethana2 @ Aug 16th 2008 1:49PM]

16nm, actually, but you're close and the difference really Doesn't Matter.

[tekdroid @ Aug 16th 2008 11:24AM]

8 cores in 2009-2010, eh.

cutting-edgist: "yeah man, you should come see my OctoCore"
average joe: "pfff. CPUs are so old hat and so low-value on the ... u know, the value scale"
cutting-edgist: "you need one now! did you see me get attacked by that octopus in Carmack's new one? Tell me you don't want that"
average joe: "call me when it makes Word faster"
cutting-edgist: "my avchd vids can almost be edited as fast as standard DV"
average joe: "Zzzz"

[Leonard @ Aug 16th 2008 11:32AM]

nehalem i7? aah, so 2009. I think I'll wait till Haswell comes before I change my laptop. and by then computer would likely pack these specs :

1. DirectX® 11 or 12
2. DDR 3 or 4
3. 20GHz Haswell
4. USB 3 or 4
5. Blue Ray writer and 500GB disc
6. terrabytes of super fast SSD.
7. Solar Charging
lots of fun eh.
would you guys rather wait?

[eiki @ Aug 16th 2008 11:40AM]

20GHz? You know little

[Kizorblade @ Aug 16th 2008 11:52AM]

My 3ghz 256 core beast will kick your single processor clocking at 6.66 times higher than mine.

[Itsuru @ Aug 16th 2008 12:44PM]

Why wait? There'll always be something better around the corner.

[Weet @ Aug 16th 2008 12:09PM]

Hey, some of us can use all the power we can get. :) More power = better rendering techniques given a set time frame.

[ethana2 @ Aug 16th 2008 2:02PM]

That is true.
..I never really got into 3d that much, Blender's so powerful it's scary.. it's one of those things where, after you learn it, it's perfect and fast, but before you learn it, it's just intimidating.

--I think I made a liquid monkey head though.

[AlphaTeam @ Aug 16th 2008 12:14PM]

Now we just need more developers to write multi-core applications.

[ethana2 @ Aug 16th 2008 1:57PM]

Yep.
Although notice that graphics API's are scaling up to hundreds of cores quite easily-- I'm thinking maybe we just need an abstraction layer apps can pass things to and have the machine run them as parallel as needed. That's really why things like PhysX are superior to on-CPU physics. You could implement it on an octal core machine or a gpu, and the application developer doesn't have to write all this crazy code with locks and threads and goodness knows what else..
..maybe i'm slightly mistaken on the details, but I really look forward to the age of accelerated physics APIs.

[Justice PID Freedom, III @ Aug 16th 2008 2:31PM]

Not gonna happen, the Cell CPU's are bad enough :)
-Freedom (an actual UNIX Software Engineer)

If you think experts are expensive, wait until you hire an amateur!!!
Katharine Hepburn - "Life is hard. After all, it kills you."

[lamb @ Aug 16th 2008 3:18PM]

AVX sounds like an instruction set they'll use in their Larrabee chip..

[buttabean @ Aug 16th 2008 4:59PM]

hmm 22nm wonder if tin splintering will become a huge problem at that point

[Roofus @ Aug 16th 2008 10:51PM]

The current models only increase efficiently up to 16 cores, then adding cores is no longer cost effective. So as you can see they'll be facing another wall in processor design in the next few years, one they can't just throw a quick fix like more cores at.

P.S. thanks for mentioning that the read link was in french.

[testing4humans @ Aug 17th 2008 7:56AM]

The paradigm of parallel computing is going to have to be developed a lot more in order for there to be an actual use of anything over 16 cores or so anyway.

A very simple way to leverage this technology could be to give each process a core of it's own. And the linked processes can be looked at as similar to other distributed systems like the web without the problem of communicating over long distances.

from ethana2

"..maybe i'm slightly mistaken on the details, but I really look forward to the age of accelerated physics APIs."

Me too! The potential for simulation in this type of computing environment is immense. This may also help us understand how extremely complex systems interact on finer and finer levels and help us get a better grasp on the next model which will harness the properties inherent at the quantum scale. (i.e. this could be a right step on the path to scalable and practical quantum computing.)