Has anyone seen a real chip IBM 45nm Demo?

[udontnome]

Intel had theirs over a year ago.

http://www.itweek.co.uk/personal-computer-world/news/2149173/intel-announces-first-45nm

 

[NMDante]

Intel had theirs over a year ago.

http://www.itweek.co.uk/personal-computer-world/news/2149173/intel-announces-first-45nm

No.... they have not shown anything functioning in a public venue, not even on 65 nm... they have yet to release a 65 nm product.

I think they showed a 45nm SRAM or something, but it was more a joint venture with Samsung and others, along with IBM. Let's see if I can find the article....

*edit*

Well, here's something that might be what I was talking about. Not a full CPU or SRAM, but just a functional silicon circuit.
IBM and friends 45nm thing

 

[LAN_deRf_HA]

I'm curious... could amd skip 45nm and work on 32nm to try and close the gap with intel? or would not having a 45nm product crush their stock...

 

[udontnome]

http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~114948,00.html

JJ,
It seems I remember that immersion is very spendy cuz you have to have all new tools. Since Intel hasn't said anything about it, is it assumed theirs is dry?

 

[NMDante]

True.

I just remember IBM and partners coming out with a working 45nm model, but I couldn't remember if it was SRAM or something else.

@Lan-deRf_HA:

AMD could just bypass 45nm, but they would have to pretty much set the standard and test all 32nm equipment needs, pretty much like Intel and IBM have been doing. It just wouldn't be feasible, since AMD doesn't have the FABs to do such a thing right now. If they had a true R&D FAB, then they could possibly achieve it, but with 2 production FABs, it's not likely they would skip 45nm, unless IBM finds a way to perfect 45nm and goes straight into 32nm experimentation.

 

[ryman554]

Yep, I saw that (or some report similar to this) --- this is vastly different in complexity to a 70 or 80 Mb SRAM chip or an MPU.

Actually, no it isn't.

This sounds like a full testchip (with working sram and various logic building blocks) that will be used to develop / debug their 45nm process. That's the same thing intel showed last year at this time. And, presumably, AMD as well.

A working CPU it is not. But it's certainly far from a trivial announcement.

 

[RandMcnally]

True.

I just remember IBM and partners coming out with a working 45nm model, but I couldn't remember if it was SRAM or something else.

@Lan-deRf_HA:

AMD could just bypass 45nm, but they would have to pretty much set the standard and test all 32nm equipment needs, pretty much like Intel and IBM have been doing. It just wouldn't be feasible, since AMD doesn't have the FABs to do such a thing right now. If they had a true R&D FAB, then they could possibly achieve it, but with 2 production FABs, it's not likely they would skip 45nm, unless IBM finds a way to perfect 45nm and goes straight into 32nm experimentation.

Isn't that what their next expansion in Dresden is? Is an R&D FAB... atleast according to amd's website.

 

[NMDante]

True.

I just remember IBM and partners coming out with a working 45nm model, but I couldn't remember if it was SRAM or something else.

@Lan-deRf_HA:

AMD could just bypass 45nm, but they would have to pretty much set the standard and test all 32nm equipment needs, pretty much like Intel and IBM have been doing. It just wouldn't be feasible, since AMD doesn't have the FABs to do such a thing right now. If they had a true R&D FAB, then they could possibly achieve it, but with 2 production FABs, it's not likely they would skip 45nm, unless IBM finds a way to perfect 45nm and goes straight into 32nm experimentation.

Isn't that what their next expansion in Dresden is? Is an R&D FAB... atleast according to amd's website.

Possibly, but I don't really have that information.

To me, and R&D Fab is mostly just that, for research and development. With Intel, they have D1D, which does most of the R&D work, as far I know. As new technologies are discovered, per se, it is passed off to the other fabs to further spin workable silicon. Of course, this is being done in D1D as well. So, with Intel, they have at least 2 or 3 fabs to help push newer technologies, while AMD only has 1 other fab, minus Charter's foundry, to help excelerate new tech.

Of course, this is just a simplified example, and not to be taken as hard, cold fact. If AMD does make the conversion of Dresden to an R&D fab, that would help AMD a lot.

 

[ryman554]

Yep, I saw that (or some report similar to this) --- this is vastly different in complexity to a 70 or 80 Mb SRAM chip or an MPU.

Actually, no it isn't.

This sounds like a full testchip (with working sram and various logic building blocks) that will be used to develop / debug their 45nm process. That's the same thing intel showed last year at this time. And, presumably, AMD as well.

A working CPU it is not. But it's certainly far from a trivial announcement.

Frankly, I misunderstood the link refering to what he was discussing, I thought he was posting a 65 nm related link....

This is even more trivial than what I described.... it is one thing to simply say 'we have this...", ok I accept that you have it.... now let me see it.... an IO circuit with a memory cell??? Come on.... so you are going to tell me that a UART RS232 chip has the same complexity as an MPU?

Excuse me while I laugh a little.....

Ok, done.

Look....

The component demonstrated a 30 per cent increase in performance over 65nm chips, the group claimed.

Read that article again, show me the metric where it gets 30% ??? Hmmmm.... well at least they claim it....

I am a "show me the data" sorta guy, press announcements like this -- especially from IBM -- while interesting, are not impressive.

Jack,

That announcement was far from trivial. It's a key milestone in their process development.

By your rationale, then intel didn't show squat until last week when they demo'd penryn A0.

What do you think intel used (and "showed off") to develop their 45nm to prepare for Penryn? The thing exactly like that IBM article was describing. A sandbox. Bits of logic building blocks, but nothing even remotely resembling a true MPU. That's not the point of the device.

Here's what I could find about intel's annoucement a year ago:
intel 45nm sram announcement Note the distinct lack of anything that even remotely resembles a MPU. Note, also, intel "claimed" 20% -- but I don't know if they backed that up anywhere (IEDM?)

I make no claims as to how well the thing actually works. That 30% number, to me, as well, appears to be purely conjecture. It's not the point of the article. My point is that it's a key, critical first step toward whatever they're going to make just to start design/process debug. To even get to the point where they *can* get a mask set to print some silicon to get some "data" to extrapolate means they are on their way. And that's far from an easy (or trivial) task.

Of course, on their way in this case means 1-2 years to product.

 

[tamalero]

could be that high-k and metal gates are hard to implement ?
and could slowdown the entire process thus not very convenient for intel's current "crush amd with all we can" production scheme?

 

[CaptRobertApril]

Here is the final retail packaging!!!!

 

[Harrisson]

No..... what Intel showed me was a TEM of the high-K stack, a defect/yield matruity curve, 5 different systems booted and running software. This is DATA, this is IMPRESSIVE.

The actual implementation of the high-K in and of itself is revolutionary, when IBM acutally shows me a working device, then they have implemented a revolutionary material and kudo's to them.

I understand your point Jack, but IMHO you are using double standarts and ryman554 has a point here. Its not just press release from IBM and friends. Yes it may have been just "we have it too" type of press release, so what? They have working silicon and regardless if this was presented "professionally" or not, they moved a long way on this track. Asking video tapes, etc. is superficial IMO. Like you were bashing AMD Barcelona demo, they had it running live so according to your current post "itself is revolutionary, <..>, then they have implemented a revolutionary material and kudo's to them." but no, you bash AMD for not benchmarking in every way you want them, and if they would do it "unprofessionaly", it wouldnt count anyway? :roll:

In my opinion, regardless how companies manage their PR, as long as they deliver newer technology, preferably without long delay, its good enough to me. IBM, Intel, AMD - all have a good record of pushing the latest technologies, and btw IBM is long time leader on new patents, wiping the floor with others every single year.

 

[xpresso]

Oh boy here we go. That press release by IBM and partners is not very credible. IBM is known for paper launches. Question have you seen 65 nm IBM in any mass product yet? IBM known for having massive Phd resources who are very innovative and pile up a lot of patents. But they don't design a high yielding process technology.

Do you remember when IBM released their Cu/low-k process around 1999 only to have to scrap their low-k innovation due to near zero yields as reliability issues?? 8O 8O

BTW, AMD does not innovate, AMD pays IBM tons of money for process R&D. Even AMD chip architectures are not innovative. Their flagship designs were done by buying a microprocessor design company (NexGen) along with ex Intel chip designers

 

[djgandy]

If you have something that's good, you show it off. That's what Intel have done.

 

[Ycon]

No, it would end up in a disastrous process and chips.
Noone can just skip a process generation, you need the results from one generation, see what you could have done better and only then can you proceed to the next one.

Its like trying to jump across the Grand Canyon before having tried jumping across a smaller distance.

 

[LordPope]

Oh boy here we go. That press release by IBM and partners is not very credible. IBM is known for paper launches. Question have you seen 65 nm IBM in any mass product yet? IBM known for having massive Phd resources who are very innovative and pile up a lot of patents. But they don't design a high yielding process technology.

Do you remember when IBM released their Cu/low-k process around 1999 only to have to scrap their low-k innovation due to near zero yields as reliability issues?? 8O 8O

BTW, AMD does not innovate, AMD pays IBM tons of money for process R&D. Even AMD chip architectures are not innovative. Their flagship designs were done by buying a microprocessor design company (NexGen) along with ex Intel chip designers

u know u sound like a dyed in the wool INTEL FANBOI

 

[CaptRobertApril]

Here is the final retail packaging!!!!

You are on a roll....

Kaiser or Cinnamon?

 

[udontnome]

No..... what Intel showed me was a TEM of the high-K stack, a defect/yield matruity curve, 5 different systems booted and running software. This is DATA, this is IMPRESSIVE.

The actual implementation of the high-K in and of itself is revolutionary, when IBM acutally shows me a working device, then they have implemented a revolutionary material and kudo's to them.

I understand your point Jack, but IMHO you are using double standarts and ryman554 has a point here. Its not just press release from IBM and friends. Yes it may have been just "we have it too" type of press release, so what? They have working silicon and regardless if this was presented "professionally" or not, they moved a long way on this track. Asking video tapes, etc. is superficial IMO. Like you were bashing AMD Barcelona demo, they had it running live so according to your current post "itself is revolutionary, <..>, then they have implemented a revolutionary material and kudo's to them." but no, you bash AMD for not benchmarking in every way you want them, and if they would do it "unprofessionaly", it wouldnt count anyway? :roll:

In my opinion, regardless how companies manage their PR, as long as they deliver newer technology, preferably without long delay, its good enough to me. IBM, Intel, AMD - all have a good record of pushing the latest technologies, and btw IBM is long time leader on new patents, wiping the floor with others every single year.

Apples to oranges. Barcelona is 65nm, nothing revolutionary there at all, i.e. transistors built much differently.

 

[BaronMatrix]

Oh boy here we go. That press release by IBM and partners is not very credible. IBM is known for paper launches. Question have you seen 65 nm IBM in any mass product yet? IBM known for having massive Phd resources who are very innovative and pile up a lot of patents. But they don't design a high yielding process technology.

Do you remember when IBM released their Cu/low-k process around 1999 only to have to scrap their low-k innovation due to near zero yields as reliability issues?? 8O 8O

BTW, AMD does not innovate, AMD pays IBM tons of money for process R&D. Even AMD chip architectures are not innovative. Their flagship designs were done by buying a microprocessor design company (NexGen) along with ex Intel chip designers

u know u sound like a dyed in the wool INTEL FANBOI

65nm Cell

 

[xpresso]

hardly I just happen to build my last 2 PC's using AMD thunderbird, yes thunderbird 950 Mhz (you are that old to remember?) & and a 64 bit 3200 CPU was the latest one. Now to educate your sorry little ass, what i have stated are known industry facts, but sure it hurts you IBM/AMD poster boy

 

[xpresso]

nice try, FUD link

 

[Harrisson]

Oh boy here we go. That press release by IBM and partners is not very credible. IBM is known for paper launches. Question have you seen 65 nm IBM in any mass product yet? IBM known for having massive Phd resources who are very innovative and pile up a lot of patents. But they don't design a high yielding process technology.

Do you remember when IBM released their Cu/low-k process around 1999 only to have to scrap their low-k innovation due to near zero yields as reliability issues?? 8O 8O

BTW, AMD does not innovate, AMD pays IBM tons of money for process R&D. Even AMD chip architectures are not innovative. Their flagship designs were done by buying a microprocessor design company (NexGen) along with ex Intel chip designers

Every IT company in the world have done paper launches, including Intel. Sometimes IBM lags in tech. progress, sometimes they are far ahead. Yields - every new tech. usualy have low yields, especially high complex with high transistors count (like Cell), but as fab tech. matures, they do have nice yields as everyone else (Intel, TSMC, you name it) so your point is unbased. Btw, do you know IBM fabs like East Fishkill was named Top fab of the year?

About innovation - AMD indeed pays large sums to co-op with IBM in R&D. This doesnt in any way negate the fact AMD have their own tech. team, which (along with IBM help) does big impact on the industry and even intel follows AMD. Long gone the days when Intel was shining star and everyone else was following, its not the case last 6 years. Major points I remember at this moment:

1) AMDs DDR vision against Intels RDRAM. Intel had no choice but to follow AMD, since market didnt really accepted its "bundles".
2) 64 bits. Intel was follower.
3) Native dual and quad core designs. AMD is first with both, Intel soon to follow with quad in next year or so.
4) Low speed but high efficient cpu. Athlon vs P4 - winner was clear. Intel scraped P4 for similar to AMD approach.
5) On-die memory controler - its only the matter of time when intel will do somthing similar IMO.
6) Hyper transport. Only the matter of time when Intel will offer similar solution, its FSB is outdated for years compared to AMD.

 

[MyStuff]

Some of what you wrote, I disagree with.

1) I dont know much about.
2)Itanium came before k8 and was 64bit.
3)Stupid arguement.
4)Pentium-M
5)Timna
6)I dont know much about either.

 

[Harrisson]

Some of what you wrote, I disagree with.

1) I dont know much about.
2)Itanium came before k8 and was 64bit.
3)Stupid arguement.
4)Pentium-M
5)Timna
6)I dont know much about either.

2) Totally different platform rings the bell?
3) For user native or not doesnt matter much, but from tech. point of view it does. Learn to look from different perspectives :roll:
4) Different market. Intel followed on desktops AMD, not vice versa, forgeting all its songs "GHz is everything".
5) True, forgot that one, but it was a dead duck, never released after i820 chipset fiasco.

 

[bassin]

If you are so genious, what are you doing in this forum, you must be there and showing IBM/AMD the right way to build a 45nm chip...
If you know that, you will be a rich man, becouse AMD and Intel will pay you a lot of money for it.... But if you will just read links and posts it here, so you are like me, a costumer, and not an AMD or Intel engineer.