IBM, AMD Announces 45nm Chip High-K Gate Process

[cryogenic]

... Well seems like Intel's advantage in 45 nm is not going to be an advantage anymore:
here's the article

it seems a little odd taking into consideration that nether AMD or IBM mentioned high K at IEDM

opinions ? (marketing or amd's good kept secret?)

 

[carlhungis]

Industrial espionage?

Seems kind of crazy that everyone managed to figure this out at the exact same time.

It does seem that Intel will still have time on their side with 45nm coming online before anyone else. Based on AMD's struggle with 65nm, Intel could have that advantage for a while.

 

[cryogenic]

Found more info, actually IMB says their process is more advanced than Intel's and that they too have been studiing High-K for the past couple of years:
article


and another article here on the same issue

 

[blueocean]

check this link http://www-03.ibm.com/press/us/en/pressrelease/20751.wss this was within few days of intel 45nm claims, in dec 2006. at that time amd/ibm went completely opposite direction instead of High-K, "Ultra LOW-K" for 45nm. i guess they might be doing some research on 45nm but it wasn't or they are no where till date.

 

[ajfink]

Only time will tell what is actually going on...

...if they do have it, I've heard it's pretty simple to put into even current fab processes. Could AMD drop it into its 65nm process to extend its life before going to 45nm?

 

[djgandy]

http://www.marketwatch.com/news/story/intel-ibm-building-smaller-faster/story.aspx?guid=%7BA9D92CB4-5969-4FBF-B31C-2D1529B04BD9%7D

Intel claim to have working 45nm and i believe them. IBM though, they are just saying what people want to hear. They might have something in the works by next year, but i can't see it being in full production.
AMD, haven't even got their 65nm working perfect yet. I can't see them on 45nm within 6 months of Intel at all.

 

[gman01]

Intel Says Chips Will Run Faster, Using Less Power:
http://www.nytimes.com/2007/01/27/technology/27chip.html?pagewanted=1&ei=5088&en=8106bacaf54ed4e0&ex=1327554000&partner=rssnyt&emc=rss

Moore's Law seen extended in chip breakthrough:
http://today.reuters.com/news/articlenews.aspx?type=technologyNews&storyID=2007-01-27T060031Z_01_N26325707_RTRUKOC_0_US-MICROCHIPS.xml&pageNumber=0&imageid=&cap=&sz=13&WTModLoc=NewsArt-C1-ArticlePage3

I just want to point out that 'HAFNIUM' is the name of the substance both intel and IBM/AMD will use for .45nm, .32nm, and .22 nm.... And neither can take credit for it.... It was already being used in the obsorbtion of nuclear material....

IBM also has fully funtioning chips made out 45nm, not just intel.... AMD should have no problem swiching to this process.... They need to cut their .65nm losses/problems and move on....

 

[Parrot]

check this link http://www-03.ibm.com/press/us/en/pressrelease/20751.wss this was within few days of intel 45nm claims, in dec 2006. at that time amd/ibm went completely opposite direction instead of High-K, "Ultra LOW-K" for 45nm. i guess they might be doing some research on 45nm but it wasn't or they are no where till date.

You are getting matters confused. High-K and low-K dielectrics are used differently in chips. It is NOT an "either high-K or low-K" choice as the two types are used for different purposes. The high-K dielectrics are used in the transistor structure and low-K dielectrics to separate the interconnect layers in a chip. IBM have been researching BOTH types for many years so this announcement of their's is no surprise. Furthermore, this technology is already in place at their East Fishkill, NY, Fab and is slated for the 45nm node. Please get your facts straight before you post invective.

 

[ajfink]

Intel Says Chips Will Run Faster, Using Less Power:
http://www.nytimes.com/2007/01/27/technology/27chip.html?pagewanted=1&ei=5088&en=8106bacaf54ed4e0&ex=1327554000&partner=rssnyt&emc=rss

Moore's Law seen extended in chip breakthrough:
http://today.reuters.com/news/articlenews.aspx?type=technologyNews&storyID=2007-01-27T060031Z_01_N26325707_RTRUKOC_0_US-MICROCHIPS.xml&pageNumber=2&imageid=&cap=&sz=13&WTModLoc=NewsArt-C1-ArticlePage2

I just want to pint out that 'HALFNIUM' is the name of the substance both intel and IBM/AMD will use for .45nm, .32nm, and .22 nm.... And neither can take credit for it.... It was already being used in the obsorbtion of nuclear material....

IBM also has fully funtioning chips made out 45nm, not just intel.... AMD should have no problem swiching to this process.... They need to cut their .65nm losses/problems and move on....

Well 65nm is almost done. The first batches were rough, no doubt, but that doesn't mean their whole 65nm process is going to waste. They have a hard time with fab capacity in making uber-fast transitions like Intel. They'd have to be changing over their current 90nm almost immediately to prepare for 45nm six months after Intel.

The industry just moves at such a rapid pace these days. It's what competition does.

 

[mpjesse]

http://www.marketwatch.com/news/story/intel-ibm-building-smaller-faster/story.aspx?guid=%7BA9D92CB4-5969-4FBF-B31C-2D1529B04BD9%7D

Intel claim to have working 45nm and i believe them. IBM though, they are just saying what people want to hear. They might have something in the works by next year, but i can't see it being in full production.
AMD, haven't even got their 65nm working perfect yet. I can't see them on 45nm within 6 months of Intel at all.

Right. There's something to be said for Intel showing off its 45nm processors. Clearly their 45nm process is on a roll. Another cool thing is Intel claims that they were able to drop those 45nm processors into existing UNMODIFIED motherboards w/ the i975 chipset. That's awesome. (Though, they aren't guaranteeing all motherboards will work w/ 45nm)

 

[djgandy]

http://www.marketwatch.com/news/story/intel-ibm-building-smaller-faster/story.aspx?guid=%7BA9D92CB4-5969-4FBF-B31C-2D1529B04BD9%7D

Intel claim to have working 45nm and i believe them. IBM though, they are just saying what people want to hear. They might have something in the works by next year, but i can't see it being in full production.
AMD, haven't even got their 65nm working perfect yet. I can't see them on 45nm within 6 months of Intel at all.

You are getting matters confused. High-K and low-K dielectrics are used differently in chips. It is NOT an "either high-K or low-K" choice as the two types are used for different purposes. The high-K dielectrics are used in the transistor structure and low-K dielectrics to separate the interconnect layers in a chip. IBM have been researching BOTH types for many years so this announcement of their's is no surprise. Furthermore, this technology is already in place at their East Fishkill, NY, Fab and is slated for the 45nm node. Please get your facts straight before you post invective.

I quoted that article for this part.

The chip giant said the new transistor technology will reduce chip-leakage by up to five times, cut power consumption by 30%, and boost computing performance by up to 20% over the company's current micro-architecture.

I'm not sure what facts you are asking me to get straight, i never debated what IBM have been researching.

 

[Bluefinger]

Looks like another tech. "arms race". Who can get the best transistor tech out the quickest, etc. However, the approaches that each side is taking is interesting. Intel wants the easiest option to implement, but AMD/IBM want something more substantial and that can earn them more return if they do it right. AMD's 65nm troubles is a warning if they are forced into something to early, and maybe, their 45nm processors may be much more of a threat to Intel if what I've read is correct. Now, AMD needs to make sure they have the necessary uArch to back the process up.

 

[eregular]

lets see the 32core 22nm chips...

 

[BaronMatrix]

Looks like another tech. "arms race". Who can get the best transistor tech out the quickest, etc. However, the approaches that each side is taking is interesting. Intel wants the easiest option to implement, but AMD/IBM want something more substantial and that can earn them more return if they do it right. AMD's 65nm troubles is a warning if they are forced into something to early, and maybe, their 45nm processors may be much more of a threat to Intel if what I've read is correct. Now, AMD needs to make sure they have the necessary uArch to back the process up.

AMD is not having 65nm troubles. Athlon is a 130nm tech that AMDs engineers managed to shrink down to 65nm. The Barcelona chips are native 65nm which should really show what the process will do.

 

[gman01]

lets see the 32core 22nm chips...

I want 'System On A Chip' by 22nm.... No more north/south bridges....

 

[RandMcnally]

http://www.marketwatch.com/news/story/intel-ibm-building-smaller-faster/story.aspx?guid=%7BA9D92CB4-5969-4FBF-B31C-2D1529B04BD9%7D

Intel claim to have working 45nm and i believe them. IBM though, they are just saying what people want to hear. They might have something in the works by next year, but i can't see it being in full production.
AMD, haven't even got their 65nm working perfect yet. I can't see them on 45nm within 6 months of Intel at all.

Let's see, intel...
Type Public (NASDAQ: INTC)
Industry Semiconductors
Products Microprocessors
Flash memory
Motherboard Chipsets
Network Interface Card
Bluetooth Chipsets
Revenue $35.1 billion USD (2006)
Operating income $6 billion USD (2006)
Net income $6 billion USD (2006)
Employees 94,200


And IBM
Industry Computer hardware
Computer software
Consulting
IT Services
Products See complete products listing
Revenue $US 91.4 billion (2006)[1]
Operating income $US 13.3 billion (2006)[1]
(14.6% operating margin[1])
Net income $US 9.5 billion (2006)[1]
(9.3% profit margin[2])
Employees 329,373 (2005)


Sorry, but if there is a company to believe it is IBM, they build amazing technology everyday. You'd have to be an idiot to say that IBM says what people want to hear. www.ibm.com read their press releases. Every day they create ground breaking technologies. And IBM isn't restricted by bing a semiconductor only company.

Looks like another tech. "arms race". Who can get the best transistor tech out the quickest, etc. However, the approaches that each side is taking is interesting. Intel wants the easiest option to implement, but AMD/IBM want something more substantial and that can earn them more return if they do it right. AMD's 65nm troubles is a warning if they are forced into something to early, and maybe, their 45nm processors may be much more of a threat to Intel if what I've read is correct. Now, AMD needs to make sure they have the necessary uArch to back the process up.

AMD is not having 65nm troubles. Athlon is a 130nm tech that AMDs engineers managed to shrink down to 65nm. The Barcelona chips are native 65nm which should really show what the process will do.


What makes you think that ... ??

Dude --- 31% area shink, increased L2 latency, 8 weeks from launch, launch speeds slower than older technlogy, OC ceilings lower than old technology, and we are just now seeing low bins show up... come on ...

that depends who tested it, the first 65 nm core that was put as a review got to 3.1 ghz with the quad fx demo cooler he had had. So I don't think you'reright on that.

 

[Ranman68k]

Makes no difference, you can state 'My CPU will run at 5.6 GHz' or 'We just made a 500 GHz transistor'.... but until you can produce a functioning device, demonstrated and running then this type of information is worthless.

"Worthless" is a little overstated don't you think? I would sure like to have proof-of-concept before pushing millions (or billions) of dollars into production on just an idea.

AMD has produced a new process technology that underperforms on all levels the prior technology --- this is an industry first, and if you cannot see that in the data then you are blind.

I thought they reduced power consumption on equivalently clocked processors.

 

[SSS_DDK]

IBM has always been a research giant..So i wouldn't be that surprised if AMD, to catch up, start buying their technology...Which is great for us.

 

[tamalero]

that depends who tested it, the first 65 nm core that was put as a review got to 3.1 ghz with the quad fx demo cooler he had had. So I don't think you'reright on that.

3.1 from a Forum of some guy who later said 'don't have time, I am gonna stop doing this', is not what I would say is reliable and good data. Aside form that, 3.1 GHz is still below the 90 nm ceiling....

I am 100% correct on this, from OCability to the fundamental transistor characteristic measurements, it is underpeforming and falls 10-15% short of Intel's 65 nm process. Heck, it is not even truly 65 nm if they could only get 31% area shrink.... it is completely screwed up and in limited supply which suggest 'mature yields' are quite low.

AMD has produced a new process technology that underperforms on all levels the prior technology --- this is an industry first, and if you cannot see that in the data then you are blind.

The short of it is -- the AMD 65 nm process is technically not ready, but they are releasing anyway --- for costs. Builds confidence in your next purchase doesn't it.

This is the problem with blind fanboyism --- people will choked down anything they read and then project that assumption based on limited data and, in this case, accept data from unreliable sources as fact.... does it not piss you off, as an evident AMD fanboy, that AMD touted not more than a year ago that their 65 nm process would produce a 40% (boy they love that round figured number) improvement --- when all we see is garbage.

actually I remember a lot of people angrily yelling and insulting intel cause their first die shrink and DDR2 cpus decreased the speed of the p4 architecture.
but later fixed with the Cedar Mill core if I remember correctly.

 

[ajfink]

IBM has always been a research giant..So i wouldn't be that surprised if AMD, to catch up, start buying their technology...Which is great for us.

They don't have to buy their technology. AMD and IBM do joint research on microprocessor technologies. AMD gets to leverage IBM's massive R&D skills, IBM gets AMD's focused processor-only research and drive to help their own processors. They don't do architectural research together, but fab process and materials research is done together.

 

[NightlySputnik]

This is an easy one.

1: Intel announce product, or in this case process.

2: IBM and AMD don't want to look too much behind Intel and announce the exact same thing before they would have announced it otherwise (probably in 6 months from now).

3: Intel release 45nm high-K process at the end of 2007.

4: IBM and AMD release 45nm High-K process 6-12 months after. No change there since process changes don't come this easily.

The cause for this might be either: 1: industrial spying. 2: that's just where research is now and both companies had to get there now both on a financial and a manufacturing point of view. 3; (most probably) a little bit of both.

But for today, IBM/AMD only had to announce it to look competitive with Intel. That's all.

Disclaimer :!: : This isn't a pro-Intel post Intel would have done the same if the situation would have been reverse.

 

[zornundo]

...and we are just now seeing low bins show up... come on ...

Me thinks AMD is holding the high bin's for this fall's back to school sales!

 

[qurious69ss]

...and we are just now seeing low bins show up... come on ...

Me thinks AMD is holding the high bin's for this fall's back to school sales!

:lol: :lol: Just like they were holding on to the X2 5000 last year??? :lol: :lol:

 

[bixplus]

This is an easy one.

1: Intel announce product, or in this case process.

2: IBM and AMD don't want to look too much behind Intel and announce the exact same thing before they would have announced it otherwise (probably in 6 months from now).

3: Intel release 45nm high-K process at the end of 2007.

4: IBM and AMD release 45nm High-K process 6-12 months after. No change there since process changes don't come this easily.

The cause for this might be either: 1: industrial spying. 2: that's just where research is now and both companies had to get there now both on a financial and a manufacturing point of view. 3; (most probably) a little bit of both.

But for today, IBM/AMD only had to announce it to look competitive with Intel. That's all.

Disclaimer :!: : This isn't a pro-Intel post Intel would have done the same if the situation would have been reverse.

Agree with most of the above...except for this line in your disclaimer :

Intel would have done the same if the situation would have been reverse.

I could be wrong, but I don't remember Intel doing something like this in the last 18 months...do you have a link to such an instance. Not saying they aren't capable of such a thing, but just don't believe they would have to or need to resort to this type of thing in their present condition. Just an opinion though.

 

[piesquared]

This is an easy one.

1: Intel announce product, or in this case process.

2: IBM and AMD don't want to look too much behind Intel and announce the exact same thing before they would have announced it otherwise (probably in 6 months from now).

3: Intel release 45nm high-K process at the end of 2007.

4: IBM and AMD release 45nm High-K process 6-12 months after. No change there since process changes don't come this easily.

The cause for this might be either: 1: industrial spying. 2: that's just where research is now and both companies had to get there now both on a financial and a manufacturing point of view. 3; (most probably) a little bit of both.

But for today, IBM/AMD only had to announce it to look competitive with Intel. That's all.

Disclaimer :!: : This isn't a pro-Intel post Intel would have done the same if the situation would have been reverse.

Agree with most of the above...except for this line in your disclaimer :

Intel would have done the same if the situation would have been reverse.

I could be wrong, but I don't remember Intel doing something like this in the last 18 months...do you have a link to such an instance. Not saying they aren't capable of such a thing, but just don't believe they would have to or need to resort to this type of thing in their present condition. Just an opinion though.

While not %100 related, I think this lends to Intel's credibility (or lack thereof)

The Register:

Google did not boot AMD for Intel