Intel says 32 nm on track for late 2009

[taltamir]

Originally posted by: geoffry
Given that mainstream nehalem CPUs wont be seen till late 2009, you think it can be assumed that if INTC does indeed release 32nm CPUs in late 2009 it would probably be server only, perhaps high end desktop?

intels SSD plans has had 32nm MLC chips slated for 2009 for quite some time...

mmm, but shouldn't those be different manufacturing types?

 

[jones377]

Originally posted by: taltamir

Originally posted by: geoffry
Given that mainstream nehalem CPUs wont be seen till late 2009, you think it can be assumed that if INTC does indeed release 32nm CPUs in late 2009 it would probably be server only, perhaps high end desktop?

intels SSD plans has had 32nm MLC chips slated for 2009 for quite some time...

mmm, but shouldn't those be different manufacturing types?

Those will be built on a 34nm process. And no, that's not a typo.

 

[Nemesis 1]

Exactly. If this wasn't A differant set of circumstances I would agree. But We don't know what Intel has been doing the Last few years. Gulfstream the cancelled processor was 32 core ringbus . Its possible intel could release a Platiform system at 32nm. Something like AMD Dragon. Called Nemesis. LOL!

But lets been honest here. We have 3 companies heading for 1 time line. Intel if their going to win . HAS TO take a chance. AMD isn't just setting with there Tumbs up there butts. They have 2010 in sight and I believe they will hit it. I think NV is screwed. Unless Intel screws up and lays an egg. Dx11 was hugh to AMD. I lol evertime I here how DX11 removes the burden from the CPU . Which is true if you want it to be (NV) But AMD makes CPUs and good ones. ATO/AMD will use the DX11 programmability to get everthing they can from the CPU. Must remember guys AMD/ATI and intel are going down the exact same path. I believe AMDs GPU is better for raytracing than larrabee. But I believe intel will have a stronger platiform. Mainly because of Intels compilers. Intel will use all the resources onboard to power next generation games(project offset) I do believe AMD /ATI have a game being developed also for raytracing. I differant approach but still good results.

 

[Idontcare]

Originally posted by: jones377

Originally posted by: taltamir

Originally posted by: geoffry
Given that mainstream nehalem CPUs wont be seen till late 2009, you think it can be assumed that if INTC does indeed release 32nm CPUs in late 2009 it would probably be server only, perhaps high end desktop?

intels SSD plans has had 32nm MLC chips slated for 2009 for quite some time...

mmm, but shouldn't those be different manufacturing types?

Those will be built on a 34nm process. And no, that's not a typo.

Yeah trying to use the process node label as a means to communicate something significant about the underlying process technology loses a lot of its relevance when comparing MPU (logic) to Memory process nodes.

What makes the memory IDMs label a memory process node as 34nm has very little to do with what makes compels the logic IDMs to label a logic process node as 32nm. Long long long ago the two were aligned, both in the number used for node designation as well as what it was in the underlying process technology that merited the node being labeled as such. But all that has long since changed a decade ago.

Nowadays comparing the process tech in a 34nm memory node to the process tech in a 32nm logic node is like comparing the drive-train and engines mechanics of a 2009 Mazda RX-8 (car) to that of a 2009 Yamaha SX230 (boat)...they are both labeled as 2009 models after all

And yes the 2010 versions of both products are expected to be improvements over their older 2009 siblings but that doesn't mean they are any closer to being comparable to one another just because they transition from a 2009 label to a 2010 label.

 

[Nemesis 1]

Originally posted by: IntelUser2000

Originally posted by: Nemesis 1
Westmere sounds cool 8 cores 16 threads. But can ya here the ranting. Look whats going on with duals Vs. Quads now.

The only differance I see between now and 9 months from now is . Larrabee. So you guys must be thinking larrabbe will use all those cores and threads Westmere brings.

Or is it the 4 core version ya want. I could careless about Westmere. I want to see larrabee and play project offset . Using raytracing. Maybe its LarrabeeX86 will see first @ 32nm 1st. That would make the most sense. Beings intel needs Die size and heat reduction. And power to deal with what AMD is preparing . I believe Larrabee will be the first chip released at 32nm. That solves alot of intels problems and it still keeps them ontime.

1. Westmere is 6 core/12 threads. Have you been reading?
2. Larrabbee is likely 45nm. Because entirely new architecture + new process technology creates too much unpredictability. That's the reason Intel went with the Tick-Tock strategy.
3. You never make sense.

Well I won't take time to find were desktop 32nm was to have an eight core cpu. But Icould find one. I haven't looked for sometime and All I found were 6 core westmeres. No biggy we all heard the 8 core stuff also.

I don't know what your talking about. There are more than a few larrabbee 45nm already in use. Just not commerical quanities. Did you watch the super bowl half time.

That was Intel / Dreamworks. The 3d glasses intel put out. All of dreamworks future stuff is 3d thanks to intel and larrabbee. Tru3D is what I recall intel calling it. Read what dream works is saying about Larrabee.

Actually making sense isn't what this industry is about. Or X86 and MS would have passed on along time ago. All because of lazy programmers . or just not enough talent.

Intel all along said That metal gates @ 45nm. Fact is they until 2006 1/2 said 3D gates @ 45nm . But that was changed second half of 06. 3D gates was moved to 32nm. Now some say . Intel has said no such thing . Thats true . But they didn't tell anyone about metal gates @ 45nm did they . I was the only one saying it. I was saying 3D gates @ 45 until intel article in 06 said HighK/metal gates at 45nm. It was discussed at 65nm.

The thing about the 3D gates Intel has not published an article saying 3D gates will come later than 32nm. There for I say it still happens at 32nm. The info is at your finger tips . Its just weather or not ya believe intell can stay with its time frame. I do they been hitting on all eight and Intel is looking strong.


Why in Gods name do you think NV is marketing Cuda. Stero 3D . Why do you think AMD bought ATI. You guys act like Larrabee is something Intel dreamed up after AMD/ATI announced fusion. LOL. ATI /Intel were working close at the time. Larrabee was already in development along with Gulfstream(Kiefer) Spelling. And one other project the name escapes me. But . Long story short Larrabee Won . Dave of ATI was working close with Intel They had more than one crosslicence agreement in place. Why was it ATI went away from the ring bus right after AMD bought them ???//// ATI Stock holders were the winners. Dave acted according to what was best for his share holders. A good man.

 

[Idontcare]

Originally posted by: Nemesis 1
All because of lazy programmers . or just not enough talent.

You and I both know the consumer-side output of programmers has very little to do with their personal motivation and talent when it comes to being hired-hands.

The scope/scale of all projects are determined by marketing/sales/commercialization departments, the budget is determined by accounting, and the timeline is determined by management.

Whether Johny programmer can make a program with 10x more features or produce the same project features with 1/10 the time or 1/10 the budget has very little impact on what the end product can be.

This is not true for entrepreneurs in the small-scale apps world. iphone apps for instance really allows an individuals work ethic and talent come to fruition in the capability of an app. Not true for a large scale project like say Photoshop CS5 or MS Windows 7.

There your rate-limiting bottleneck in innovation is determined by an amalgam of people and job functions that have never seen a line of code let alone used the very product line they are producing.

Having been an engineer operating in an environment of extreme restriction on one's ability to contribute to improving a product above and beyond its targeted/stated feature-set really changed my mindset when it comes to attributing product feature failure. P4 was not a failure of Intel's IC and process engineers, they built exactly what they were instructed to build. Having also recently become (~3yrs now) essentially a self-employed programmer I also see now firsthand how my products are truly limited by my capabilities and work ethic.

Two sides of the coin.

 

[ilkhan]

I'm not into the deep info stuff that you gentlemen are talking about, but I do have a couple of questions.
Are we expecting a quad core 32nm chip, or just duals and 6-cores?
And
Is Arandale supposed to fill both mobile and desktop dual core needs (with separate socket connections), or do do we think Intel has a separate name we haven't heard yet?

 

[Idontcare]

Originally posted by: ilkhan
I'm not into the deep info stuff that you gentlemen are talking about, but I do have a couple of questions.
Are we expecting a quad core 32nm chip, or just duals and 6-cores?
And
Is Arandale supposed to fill both mobile and desktop dual core needs (with separate socket connections), or do do we think Intel has a separate name we haven't heard yet?

Kind of you to equate us with gentlemen. I assure you Nemesis is undeserving of that honor Me on the other hand, thanks for noticing...

This is the closest thing I have seen to a roadmap of Intel's upcoming processors, and it is hardly anything official.

Intel Desktop Roadmap

Intel Desktop CPU

I fully expect Intel to capitalize on their inherently module design with the Nehalem/Westmere architecture and exploit the fact there is minimal design/validation effort to invest into creating 2-core/4-core/6-core tiered SKU's.

I suspect that whether they do it or not is entirely dependent on what AMD is doing at the time.

As for Arrandale, again the same PC.watch.impress site says this is 32nm mobile 2-core/4-thread and has the same specs as desktop Clarkdale 32nm 2-core/4-thread but whether they are actually same stepping/mask-set is anyone's guess at this point.

Intel Mobile Roadmap

Intel Mobile CPU

These roadmap links are from this latest article on the topic at pc.watch.impress, google translated here:
http://66.102.1.101/translate_...p9L2XGgt9oI9Jm7dUtONAQ

 

[ilkhan]

Thanks Idontcare, Id seen snips of those roadmaps but not the whole things. Very useful.

 

[Idontcare]

Originally posted by: ilkhan
Thanks Idontcare, Id seen snips of those roadmaps but not the whole things. Very useful.

They are kind of overwhelming to me. The volume of information captured in them is seemingly comprehensive.

 

[ilkhan]

Originally posted by: Idontcare

Originally posted by: ilkhan
Thanks Idontcare, Id seen snips of those roadmaps but not the whole things. Very useful.

They are kind of overwhelming to me. The volume of information captured in them is seemingly comprehensive.

Overwhelming at first, but if you focus on specific parts they have a ton of info laid out.

Any idea what "Cougar point" is?

 

[Idontcare]

Originally posted by: ilkhan

Originally posted by: Idontcare

Originally posted by: ilkhan
Thanks Idontcare, Id seen snips of those roadmaps but not the whole things. Very useful.

They are kind of overwhelming to me. The volume of information captured in them is seemingly comprehensive.

Overwhelming at first, but if you focus on specific parts they have a ton of info laid out.

Any idea what "Cougar point" is?

Cougar Point is Sandy Bridge core architecture (the 32nm successor to Westmere) combined with integrated GPU (monolithic) versus the expected MCM GPU/CPU for Clarkdale.

This integrated GPU is NOT expected to be a Larrabee derivative, but that is just public opinion, no educated opinions are available on the topic to my knowledge. (people know which it is, but those people aren't talking yet)

 

[ilkhan]

well that makes sense. Thanks.

 

[IntelUser2000]

Originally posted by: Nemesis 1

Originally posted by: IntelUser2000

Originally posted by: Nemesis 1
Westmere sounds cool 8 cores 16 threads. But can ya here the ranting. Look whats going on with duals Vs. Quads now.

The only differance I see between now and 9 months from now is . Larrabee. So you guys must be thinking larrabbe will use all those cores and threads Westmere brings.

Or is it the 4 core version ya want. I could careless about Westmere. I want to see larrabee and play project offset . Using raytracing. Maybe its LarrabeeX86 will see first @ 32nm 1st. That would make the most sense. Beings intel needs Die size and heat reduction. And power to deal with what AMD is preparing . I believe Larrabee will be the first chip released at 32nm. That solves alot of intels problems and it still keeps them ontime.

1. Westmere is 6 core/12 threads. Have you been reading?
2. Larrabbee is likely 45nm. Because entirely new architecture + new process technology creates too much unpredictability. That's the reason Intel went with the Tick-Tock strategy.
3. You never make sense.

Well I won't take time to find were desktop 32nm was to have an eight core cpu. But Icould find one. I haven't looked for sometime and All I found were 6 core westmeres. No biggy we all heard the 8 core stuff also.

I don't know what your talking about. There are more than a few larrabbee 45nm already in use. Just not commerical quanities. Did you watch the super bowl half time.

That was Intel / Dreamworks. The 3d glasses intel put out. All of dreamworks future stuff is 3d thanks to intel and larrabbee. Tru3D is what I recall intel calling it. Read what dream works is saying about Larrabee.

Actually making sense isn't what this industry is about. Or X86 and MS would have passed on along time ago. All because of lazy programmers . or just not enough talent.

Intel all along said That metal gates @ 45nm. Fact is they until 2006 1/2 said 3D gates @ 45nm . But that was changed second half of 06. 3D gates was moved to 32nm. Now some say . Intel has said no such thing . Thats true . But they didn't tell anyone about metal gates @ 45nm did they . I was the only one saying it. I was saying 3D gates @ 45 until intel article in 06 said HighK/metal gates at 45nm. It was discussed at 65nm.

The thing about the 3D gates Intel has not published an article saying 3D gates will come later than 32nm. There for I say it still happens at 32nm. The info is at your finger tips . Its just weather or not ya believe intell can stay with its time frame. I do they been hitting on all eight and Intel is looking strong.


Why in Gods name do you think NV is marketing Cuda. Stero 3D . Why do you think AMD bought ATI. You guys act like Larrabee is something Intel dreamed up after AMD/ATI announced fusion. LOL. ATI /Intel were working close at the time. Larrabee was already in development along with Gulfstream(Kiefer) Spelling. And one other project the name escapes me. But . Long story short Larrabee Won . Dave of ATI was working close with Intel They had more than one crosslicence agreement in place. Why was it ATI went away from the ring bus right after AMD bought them ???//// ATI Stock holders were the winners. Dave acted according to what was best for his share holders. A good man.

It's likely 3D gates at 32nm. It was also hinted that High K was 45nm initially but then they denied it after and general population have short memory. They never said 3D gates at 45nm. Tri-gate is even more speculative than High K.

Whether Larrabbee will be 32nm or 45nm is entirely dependent on the timeline. But I can tell you, they are not going to release Larrabbe as the first incarnation of the 32nm technology. It's way too risky.

 

[Nemesis 1]

Originally posted by: IntelUser2000

Originally posted by: Nemesis 1

Originally posted by: IntelUser2000

Originally posted by: Nemesis 1
Westmere sounds cool 8 cores 16 threads. But can ya here the ranting. Look whats going on with duals Vs. Quads now.

The only differance I see between now and 9 months from now is . Larrabee. So you guys must be thinking larrabbe will use all those cores and threads Westmere brings.

Or is it the 4 core version ya want. I could careless about Westmere. I want to see larrabee and play project offset . Using raytracing. Maybe its LarrabeeX86 will see first @ 32nm 1st. That would make the most sense. Beings intel needs Die size and heat reduction. And power to deal with what AMD is preparing . I believe Larrabee will be the first chip released at 32nm. That solves alot of intels problems and it still keeps them ontime.

1. Westmere is 6 core/12 threads. Have you been reading?
2. Larrabbee is likely 45nm. Because entirely new architecture + new process technology creates too much unpredictability. That's the reason Intel went with the Tick-Tock strategy.
3. You never make sense.

Well I won't take time to find were desktop 32nm was to have an eight core cpu. But Icould find one. I haven't looked for sometime and All I found were 6 core westmeres. No biggy we all heard the 8 core stuff also.

I don't know what your talking about. There are more than a few larrabbee 45nm already in use. Just not commerical quanities. Did you watch the super bowl half time.

That was Intel / Dreamworks. The 3d glasses intel put out. All of dreamworks future stuff is 3d thanks to intel and larrabbee. Tru3D is what I recall intel calling it. Read what dream works is saying about Larrabee.

Actually making sense isn't what this industry is about. Or X86 and MS would have passed on along time ago. All because of lazy programmers . or just not enough talent.

Intel all along said That metal gates @ 45nm. Fact is they until 2006 1/2 said 3D gates @ 45nm . But that was changed second half of 06. 3D gates was moved to 32nm. Now some say . Intel has said no such thing . Thats true . But they didn't tell anyone about metal gates @ 45nm did they . I was the only one saying it. I was saying 3D gates @ 45 until intel article in 06 said HighK/metal gates at 45nm. It was discussed at 65nm.

The thing about the 3D gates Intel has not published an article saying 3D gates will come later than 32nm. There for I say it still happens at 32nm. The info is at your finger tips . Its just weather or not ya believe intell can stay with its time frame. I do they been hitting on all eight and Intel is looking strong.


Why in Gods name do you think NV is marketing Cuda. Stero 3D . Why do you think AMD bought ATI. You guys act like Larrabee is something Intel dreamed up after AMD/ATI announced fusion. LOL. ATI /Intel were working close at the time. Larrabee was already in development along with Gulfstream(Kiefer) Spelling. And one other project the name escapes me. But . Long story short Larrabee Won . Dave of ATI was working close with Intel They had more than one crosslicence agreement in place. Why was it ATI went away from the ring bus right after AMD bought them ???//// ATI Stock holders were the winners. Dave acted according to what was best for his share holders. A good man.

It's likely 3D gates at 32nm. It was also hinted that High K was 45nm initially but then they denied it after and general population have short memory. They never said 3D gates at 45nm. Tri-gate is even more speculative than High K.

Whether Larrabbee will be 32nm or 45nm is entirely dependent on the timeline. But I can tell you, they are not going to release Larrabbe as the first incarnation of the 32nm technology. It's way too risky.

Well the thing is. When amd went to 45nm process I was reading about it. The Jump to 32 is suppose to be pain free with the same equipment. So 32nm AMD isn't all that far out. Now intel is scrambling. But Intel gets the same break going from 32nm down to 22nm.

But than ya have the New york fab coming Online. Its going to be interesting. Intel wrote alot of papers on HighK and 3D gates from 2002 out to 2006 1/2 . I haven't really looked for more recent info.

http://www.tomshardware.com/ne...-transistors,2927.html

 

[taltamir]

biggest difference i can see, is that intel actually pays for the new and necessary hardware to improve their process tech... AMD skimps on it and it hurts them in the long run. you have to spend money to make money.

 

[Idontcare]

Originally posted by: taltamir
biggest difference i can see, is that intel actually pays for the new and necessary hardware to improve their process tech... AMD skimps on it and it hurts them in the long run. you have to spend money to make money.

You have to have money to spend money.

Don't be Donald Trump and tell everyone the easiest way to become a billionaire is to start out being a millionaire. When solving partial differential equations we call that the "trivial solution".

Don't point out the trivial solution and feel like you've uncovered a solution that was completely overlooked by the combined intelligence of AMD's executive team. It's low, but its not that low.

Thus far the problem for AMD has been "how do we make a billion dollars when starting with none"...the answer so far has been "use other people's money, and lots of it!". Enter ATIC.

 

[taltamir]

ah, but when those companies HAD money they didn't spend it right... AMD could have bought some intense manufacturing equipment for less than what they bought ATI for, and sign some sort of contract with ATI.
Besides it wasn't just an AMD vs Intel thing... its every single competitor intel ever had vs intel... intel crushes company after company over the years. Even when their design was vastly inferior (pentium 4 vs athlon XP) their superior manufacturing kept them afloat.
And then AMD had the x86_64bit compared to intels original 64bit implementation (incapable of running 32bit code, and has to be written AND compiled differently).
And then AMD had the dual core x2 vs intels pentium D....

They had the advantage and they didn't press it. Intel on the other hand, even with inferior designs, could enjoy the benefits of improved manufacturing.

Intel is also ramping up to be the best MLC and SLC manufacturer

 

[Idontcare]

I think anytime you talk about long timelines like that in comparing two companies like Intel and AMD (and I like that you are) you really are speaking to the longstanding corporate leadership and culture...in other words you are really highlighting the differences between how Andy Grove ran his company (trickle down work philosophy) versus how Jerry Sanders ran his company (trickle down work philosophy).

They both had x86 and flash. One business resulted in both efforts diving into the ground, the other thrived in both accounts.

We can blame the predecessors of both Grove and Sanders for whatever our pet peeves are of the companies, but the bottom line is Barrett was chosen to succeed Grove and Ruiz was chosen to succeed Sanders by the very men they were replacing.

By many accounts Sanders treated AMD like his own personal checking account and private company, despite it being a publicly held business. Given Ruiz's compensation records it would appear this philosophy was held intact and survived the transition of power from Sanders to Ruiz.

Likewise by many accounts Grove was a ruthless (as in non-emotional, not barbaric) businessman who viewed business life as fight or flight, life versus death, the hunted and the hunter. This too appears to have survived the transition of power from Grove to Barrett to Otellini.

Say what we want about the P4, it was a business success. It made more billions of dollars of profit for Intel than any cpu has ever made for AMD. We geeks may feel in our nerdy definitions of what is "technically the best answer would be" world that the P4 netburst architecture was a netbust, but Intel's shareholders were certainly not harmed by the creation of P4.

Back to AMD's corporate philosophy...this is what gets me excited about Dirk, he's an outsider not selected by Sanders and came on-board for technical merits, not executive legacy merits. He represents a chance for AMD to become a new AMD, a non-Sanders AMD. Mind you he is but one person still surrounded by people who are legacy holdouts of the Sanders era...so we shouldn't expect miracles.

 

[taltamir]

true, the p4, when released, was clock for clock slower than a p3... it just ramped those clocks to previously unheard of hights. Perfect from a marketing perspective.