Sign in with
Sign up | Sign in
Your question

IBM and partners including AMD first to 22nm

Last response: in CPUs
Share
August 18, 2008 10:22:59 PM

IBM and AMD first at 22 nm, challenge Intel’s manufacturing lead

And:

Quote:
IBM said that it is on track with its 32 nm process and promises that it will use a “leading 32 nm high-K metal gate technology that no other company or consortium can match.”
a b à CPUs
August 18, 2008 11:23:52 PM

We'll just have to see - it is definitely an interesting development though. I hope that they are first to 22nm - AMD really needs that kind of a boost. Plus, faster stuff sooner is never a bad thing :D 
August 18, 2008 11:57:47 PM

Nice I hope this comes out soon... If Phenom can at least reach 3+ ghz at less than 100 watts on that technology, they could be competetive again.
a b à CPUs
August 19, 2008 12:01:36 AM

Phenom should reach 3GHz at 100W or less on 45nm without too much difficulty - I'd expect a hell of a lot more than just that on 22nm (considering how far out it is).

Or, to put it another way, AMD is still at 65nm. 22nm is 3 die shrinks away.
1) 45nm
2) 32nm
3) 22nm

I'd hope they'd have a bit more than a 3GHz phenom at 100W by then.

August 19, 2008 12:13:08 AM

piesquared said:
IBM and AMD first at 22 nm, challenge Intel’s manufacturing lead

And:

Quote:
IBM said that it is on track with its 32 nm process and promises that it will use a “leading 32 nm high-K metal gate technology that no other company or consortium can match.”


They said that for 45nm 65nm 90nm and I do beleive 130nm, now dont get me wrong it be cool if they gave Intel a run for their money on nodes but they havent got 45nm rolled out yet.

Word, Playa.
August 19, 2008 1:02:23 AM

spud said:
They said that for 45nm 65nm 90nm and I do beleive 130nm, now dont get me wrong it be cool if they gave Intel a run for their money on nodes but they havent got 45nm rolled out yet.

Word, Playa.



Are you sure about that?
And i'm curious when Intel will announce 22nm SRAM. Didn't they announce 32nm SRAM less than a year ago? 22nm cells at this stage is a very substantial milestone, and 22nm is believed to be at the limit of CMOS scaliing.
August 19, 2008 1:04:48 AM

They should just skip a node.. look what it has done for ATI with the recent 3xxx and 4xxx seiries. So instead of going for 45nm they should consentrate their resources (whatever they have left lol) on 32nm technology, seeing as how intel is close to developing it. Plus, I don't want to see what nehalem err.. Core i7 running on 32nm would do to AMD.
a b à CPUs
August 19, 2008 2:04:23 AM

piesquared said:
Are you sure about that?
And i'm curious when Intel will announce 22nm SRAM. Didn't they announce 32nm SRAM less than a year ago? 22nm cells at this stage is a very substantial milestone, and 22nm is believed to be at the limit of CMOS scaliing.

Yep - Intel announced 32nm SRAM on Sept. 18, 2007.
August 19, 2008 2:14:11 AM

cjl said:
Yep - Intel announced 32nm SRAM on Sept. 18, 2007.


Thanks, that's what I thought. So I dont't think we can expect a 22nm announcement any time soon. Never know though.
a c 108 à CPUs
a b À AMD
August 19, 2008 2:45:34 AM

terror112 said:
They should just skip a node.. look what it has done for ATI with the recent 3xxx and 4xxx seiries. So instead of going for 45nm they should consentrate their resources (whatever they have left lol) on 32nm technology, seeing as how intel is close to developing it. Plus, I don't want to see what nehalem err.. Core i7 running on 32nm would do to AMD.


Way too late for that ...

they would have to skip 32nm for 22nm (and it may well be too late for that).
August 19, 2008 2:55:28 AM

terror112 said:
Nice I hope this comes out soon... If Phenom can at least reach 3+ ghz at less than 100 watts on that technology, they could be competetive again.


Well, the first Deneb previews show that core hitting 3.2 ghz fairly easily.... and for less than half the power of it's current Phenom counterparts. Others have been able to push the chip to 4 Ghz. That's with a 10+% boost in speed per clock as well. All this with just a shift to 45nm. 22nm, at less than half that will be a huge leap.
August 19, 2008 3:12:26 AM

Malovane said:
Well, the first Deneb previews show that core hitting 3.2 ghz fairly easily.... and for less than half the power of it's current Phenom counterparts. Others have been able to push the chip to 4 Ghz. That's with a 10+% boost in speed per clock as well. All this with just a shift to 45nm. 22nm, at less than half that will be a huge leap.


That is very impressive. I think that If AMD can pull Deneb off with high clocks, low power, and a 10% increase in IPC, we may have a winner on our hands! It is all a matter of time when AMD becomes competative again. If AMD can achieve all those things with deneb, I would expect AMD to be able to outsell intel in the mainstream market due to the extremely cheap prices of ddr2 memory.
August 19, 2008 3:56:46 AM

terror112 said:
They should just skip a node.. look what it has done for ATI with the recent 3xxx and 4xxx seiries. So instead of going for 45nm they should consentrate their resources (whatever they have left lol) on 32nm technology, seeing as how intel is close to developing it. Plus, I don't want to see what nehalem err.. Core i7 running on 32nm would do to AMD.


don't forget AMD will catch Intel this year in 45nm becuase Intel will not have 32nm until next year.
a b à CPUs
August 19, 2008 4:19:26 AM

jaguarskx said:
Kudos to them...

...as long as they don't screw up like Intel did when they released the first 65nm CPU; Prescott (a.k.a. Pres-HOTT!!!).



Prescott was 90nm.
August 19, 2008 4:58:55 AM

blackpanther26 said:
don't forget AMD will catch Intel this year in 45nm becuase Intel will not have 32nm until next year.


Keep in mind that when AMD releases 45nm later this year, it will be built a relatively new, immature 45nm SOI process. It will be later transitioned to another relatively new, immature 45nm HK/MG process. Intel's 45nm will be fully matured by the time Nehalem has launched. When AMD finally hits 45nm HK/MG, Intel is already at 32nm. So in short, no, AMD is still one full year behind Intel in terms of process technology.
a c 126 à CPUs
a b À AMD
August 19, 2008 5:02:43 AM

One full year behind plus it took 11 companies to put their greatest minds together to get to 32nm yet Intel is at it on their own.

Plus we don't know what Intel is doing and they could have working 22nm SRAM chips just not have announced it yet.

Either way its all good for the industry.
August 19, 2008 5:17:32 AM

jimmysmitty said:
One full year behind plus it took 11 companies to put their greatest minds together to get to 32nm yet Intel is at it on their own.


They wouldn't have it any other way.

Quote:
Plus we don't know what Intel is doing and they could have working 22nm SRAM chips just not have announced it yet.


True, they very well might have. If they do, I would expect they would announce it at IDF. They absolutely don't want the world thinking they are behind on R&D.

a c 126 à CPUs
a b À AMD
August 19, 2008 5:20:48 AM

^But you still have to give Intel credit for going so far without any major help and leading the industry for over 20 years.
August 19, 2008 5:33:29 AM

No doubt about it, they have excellent engineers. As does AMD. It's just that Intel has 10x more of them.
a b à CPUs
August 19, 2008 5:38:55 AM

yomamafor1 said:
Keep in mind that when AMD releases 45nm later this year, it will be built a relatively new, immature 45nm SOI process. It will be later transitioned to another relatively new, immature 45nm HK/MG process. Intel's 45nm will be fully matured by the time Nehalem has launched. When AMD finally hits 45nm HK/MG, Intel is already at 32nm. So in short, no, AMD is still one full year behind Intel in terms of process technology.

I wouldn't even put the maturing of Intel's 45nm in the future tense - I'd say it's pretty much fully matured now, as shown by the across the board 45nm products that are showing up. The Q6600 is really the last remaining 65nm CPU that is still in full production, and with the Q9300 and 9400 dropping in price fast, it won't last long either.
August 19, 2008 7:15:32 AM

jimmysmitty said:
One full year behind plus it took 11 companies to put their greatest minds together to get to 32nm yet Intel is at it on their own.

Plus we don't know what Intel is doing and they could have working 22nm SRAM chips just not have announced it yet.

Either way its all good for the industry.


Read between the lines.

The announcement:
Quote:
IBM and its joint development partners, AMD, Freescale, STMicroelectronics, Toshiba and the College of Nanoscale Science and Engineering (CNSE), announced the first working static random access memory (SRAM) for the 22nm technology node, the world's first reported working cell built at its 300mm research facility in Albany, NY.


The 32nm SRAM IBM announced last year:
Quote:
IBM Corp. (Armonk, N.Y.) said today it has developed a test SRAM array using 32 nm high-k/metal gate (HKMG) process technology, an achievement that puts IBM and its Fishkill alliance development partners on track to introduce 32 nm technology in the second half of 2009.


What's the difference between a "cell" and an "array"? Night and day. A "cell" is a research project built in the labs -- this one happens to be at CNSE (a university). An "array" is a full wafer with an SRAM shuttle as the process matures and gears up for production. A "cell" is what's also filling intel's research labs -- you just don't hear about it. A 32nm sram "array" is what D1D has been processing (and perfecting) for the past year as intel readies for Westmere. (the Nehalem tick).

A working cell? An achievement to be sure. But don't for a second think that you're going to be seeing IBM's 22nm come off the production lines in two years time. They're not even out with 45nm yet.

Consider this -- the IBM consortium "beat" intel to the "hi-K, metal gate" announcement/demonstration about a year and a half ago. Remind me again where I can pick up one of IBM's hiK chips?
August 19, 2008 8:13:13 AM

TSMC is doing 40nm, IBM announces this, people are catching up, thats for sure. The smallest, widest used nodes wont be from Intel next year, itll be TSMCs.
August 19, 2008 8:27:10 AM

Given that 32nm process node has already passed validation at D1D, Intel might hit 32nm before TSMC hits 40nm. But of course, its all speculation without official announcements.
August 19, 2008 8:33:18 AM

40nm arrives in 1H 09
August 19, 2008 8:51:47 AM

Quote:
But of course, its all speculation without official announcements.
August 19, 2008 8:56:00 AM

For Intel, not TSMC. My point is, like I said, more and more are catching up. All the comments about IBM either theyre going to sit on it, or deliver is just that, speculation. This is TSMCs, ATIs and nVidias plans.
August 19, 2008 9:18:58 AM

Quote:
interesting but i do not think intel will allow them to be the first to debut actual cpu's on the market. we shall see although what significance it will have at that time in the future is uncertain at best.


>interesting but i do not think intel will allow them to be the first to debut actual cpu's on the market.

Why? Intel allowed AMD to be the first to debut native quad core processors on the market.

IBM is much larger then Intel is btw.

August 19, 2008 9:33:08 AM

piesquared said:
IBM and AMD first at 22 nm, challenge Intel’s manufacturing lead

And:

Quote:
IBM said that it is on track with its 32 nm process and promises that it will use a “leading 32 nm high-K metal gate technology that no other company or consortium can match.”



IBM and AMD first at 22 nm, challenge Intel’s manufacturing lead

Hardware

By Wolfgang Gruener

Monday, August 18, 2008 16:49

Yorktown Heights (NY) – IBM and its chip development partners made a stunning announcement today, apparently beating Intel in the successful production of the first functional 22 nm SRAM cell. 22 nm processors are still three years out in the future, but IBM’s news is a good sign that chip manufacturer will be able to easily scale to this new level by the end of 2011. It appears that, for the first time in several decades, Intel may have to put some extra time into its research and development efforts to make sure it can keep its manufacturing lead at 22 nm and beyond.

SRAM chips are typically the first semiconductor devices to test a new manufacturing process as a precursor to actual microprocessors. The devices developed and manufactured by AMD, Freescale, IBM STMicroelectronics, Toshiba and the College of Nanoscale Science and Engineering (CNSE) were built in a traditional six-transistor design on a 300 mm wafer and had a memory cell size of just 0.1 μm2, which compares to Intel’s 45 nm SRAM cell size (the test chip that was used for today’s 45 nm processors) of 0.346 μm2.

A 22 nm chip is two generations out in the future and AMD even has to catch up with Intel’s 45 nm. Intel presented the first 32 nm SRAM cell wafer in September of last year and in fact is not expected to show 22 nm SRAM cells for at least another year, while first 32 nm CPU prototypes could be shown at IDF this week.

IBM said that it is on track with its 32 nm process and promises that it will use a “leading 32 nm high-K metal gate technology that no other company or consortium can match.” IBM did not provide further details to substantiate this claim, however, Intel has been using its high-K metal gate technology since the introduction of the 45 nm Penryn processors in late 2007.

While we are far from actual 22 nm and 32 nm products, it is clear that IBM and its partners are turning up the heat on Intel. For the first time in decades, there could actually be an interesting race towards a new production node.

http://www.tgdaily.com/content/view/38941/135/
a c 471 à CPUs
a c 118 À AMD
August 19, 2008 1:45:07 PM

BadTrip said:
Prescott was 90nm.


D'oh !!!!!!!!!!

Guess I'm forgetting my CPU generation history.
a c 126 à CPUs
a b À AMD
August 19, 2008 1:52:08 PM

^IBM is right that currently no other "Consortium" Can match them in their 32nm HiK process but Intel is not a consortium they are a company.

TSMC may hit 40nm but they are using a different process and how long will that process hold out? Will it be able to hit 40nm while not leaking more power and causing more heat?

GPUs already are very hot running in the 60-80c under load so how well will this process node hold?
August 19, 2008 1:58:26 PM

Currently, the threshold or highest Ive seen is over 100c!!!. Itll hold, but theres always cooling solutions out there. Id add that even tho youll see tham at 40nm, it doesnt mean that theyll be less hot or use less power, not in terms of performance diminishing. Gpus havnt even reached that level yet. As far as server related gpus, the performance is sooo much higher per power usage, it isnt a question either. Who knows? Everyone says 22nm could be the end, but, gpus havnt incorporated HKMG yet, and may never need to. Time will tell
August 19, 2008 6:20:42 PM

jimmysmitty said:
^But you still have to give Intel credit for going so far without any major help and leading the industry for over 20 years.



They just don't play well with others. As soon as they got the money they began to crap on everyone. Unfortunately, IBM couldn't just stop X86 based boxes. The world we be a better place if Intel wasn't such a ..a.. I can't even describe them. It takes a lot of money to make CPUs and Intel takes full advantage of that fact. I wonder why the company that created the LCD or the TV tube or the EFI for cars isn't like Intel.

All of those businesses enjoy healthy competition.
August 19, 2008 6:26:01 PM

jimmysmitty said:
^IBM is right that currently no other "Consortium" Can match them in their 32nm HiK process but Intel is not a consortium they are a company.

TSMC may hit 40nm but they are using a different process and how long will that process hold out? Will it be able to hit 40nm while not leaking more power and causing more heat?

GPUs already are very hot running in the 60-80c under load so how well will this process node hold?




The process has very little to do with heat production. That's MHz and amount of active logic. The process can be tuned for better transistor perf but any 40nm process will be approximately the same for the same circuits. As HiK shows, it's more materials than process (the same companies sell everyone their litho stuff). It just sounds like you can't stand anyone but Intel making any kind of ICs, especially if they're good at it.
a c 126 à CPUs
a b À AMD
August 19, 2008 6:38:14 PM

^Um but the owner of the patent gets royalties. Like Blu-Ray. Sure a lot of companies actually produce the players but Sony is the main company and gets royalties for each unit/movie sold.

Same goes for the other items you listed. And in the CPU industry its no different. Only problem is it cost more to produce a CPU and research it than it does to just make a new TV. Most of those use technologies developed by another company, the DLP HDTVs for example as the DLP is made by TI for that.

But hey if you want to try and break into a market where there are only 2 dominant players and having to keep up with ever changing technology go for it. But think of this, there are only 2 GPU makers ATI and nVidia....essentially its the same as CPUs qith AMD and ATI.....
August 19, 2008 6:48:59 PM

jimmysmitty said:
^Um but the owner of the patent gets royalties. Like Blu-Ray. Sure a lot of companies actually produce the players but Sony is the main company and gets royalties for each unit/movie sold.

Same goes for the other items you listed. And in the CPU industry its no different. Only problem is it cost more to produce a CPU and research it than it does to just make a new TV. Most of those use technologies developed by another company, the DLP HDTVs for example as the DLP is made by TI for that.

But hey if you want to try and break into a market where there are only 2 dominant players and having to keep up with ever changing technology go for it. But think of this, there are only 2 GPU makers ATI and nVidia....essentially its the same as CPUs qith AMD and ATI.....



So you mean I missed the big battle between EFI, LCD and TV tube makers. Crap. That would have been good. Especially since there still seems to be multiple players with no price wars designed to hurt competitors while doing nothing for consumers.

Business is business. There were more than two GPU makers and there still are. They just are in different markets. there used to be like five x86 makers, but Slot 1 killed most of them. Intel purposefully left the VX chipset just to see who could afford to make a chipset. Hell, ATi used to be on EVERY Intel mobo. (Rage Pro anyone) Then Intel decided they didn't want to share that either and graphics have been the worst ever since.
August 19, 2008 7:04:13 PM

BaronMatrix said:
So you mean I missed the big battle between EFI, LCD and TV tube makers. Crap. That would have been good. Especially since there still seems to be multiple players with no price wars designed to hurt competitors while doing nothing for consumers.

Business is business. There were more than two GPU makers and there still are. They just are in different markets. there used to be like five x86 makers, but Slot 1 killed most of them. Intel purposefully left the VX chipset just to see who could afford to make a chipset. Hell, ATi used to be on EVERY Intel mobo. (Rage Pro anyone) Then Intel decided they didn't want to share that either and graphics have been the worst ever since.


But like you said business is business, Intel is still makeing gobs of money and meeting their markets requirements, think if their GPCPU even does 10% over their current IGP they will have a winner in respects to the market demands/needs. All and all the first to any node is more bragging rights than anything, its the player that can leverage their production capacity and IC design to that node that will win. Frankly IBM and their console chips doesnt spell leverage just cost reduction on their part.

Word, Playa.
August 19, 2008 8:04:06 PM

spud said:
But like you said business is business, Intel is still makeing gobs of money and meeting their markets requirements, think if their GPCPU even does 10% over their current IGP they will have a winner in respects to the market demands/needs. All and all the first to any node is more bragging rights than anything, its the player that can leverage their production capacity and IC design to that node that will win. Frankly IBM and their console chips doesnt spell leverage just cost reduction on their part.

Word, Playa.



I'm not really sure what point you're trying to make. Aren't you tired of hearing horror stories about treatment of OEMs and the general contempt Intel seems to have for every company in the PC\Server business? Only MS has been as bad in terms of product tying, FUD and abuse of monopoly.

I know I am. I just want fair and open competition. If the other guy is better you don't threaten chipset shipments to keep them down. You don't jon a non-profit consortium and crap on the product extolling your own after you turned down the request of the makers.

That's too much for me. I would support a boycott of Intel just for the XO debacle. I can't even express how low I think that was.
a c 126 à CPUs
a b À AMD
August 19, 2008 9:28:07 PM

^No but I get sick of people talking about Intel and MS that way and trying to make other companies such as AMD out to be angel companies when its not true.

Was it not wrong for AMD to say no to the little OEMs when they tried to buy some of AMDs X2 CPUs? Yes it was wrong. Do you talk about that? No you only talk about Intel.

BM its a freaking company out for one thing and one thing only. Your money. AMD wants it too. Don't bring up BS when you know AMD would have done the same damn thing as every other company out there and thats do whatever gets you more money to expand and create more and then sell more for what? More money.
August 19, 2008 11:44:02 PM

BaronMatrix said:
I'm not really sure what point you're trying to make. Aren't you tired of hearing horror stories about treatment of OEMs and the general contempt Intel seems to have for every company in the PC\Server business? Only MS has been as bad in terms of product tying, FUD and abuse of monopoly.

I know I am. I just want fair and open competition. If the other guy is better you don't threaten chipset shipments to keep them down. You don't jon a non-profit consortium and crap on the product extolling your own after you turned down the request of the makers.

That's too much for me. I would support a boycott of Intel just for the XO debacle. I can't even express how low I think that was.


*none angry tone*

K let me get this out in the open, I frankly don’t care anymore about the politics in the industry, it’s all the same and been like this since before I could read. Fact is Baron all I want to do is talk about the technical aspects, the in's and outs of the technology's why x works better than y type deal. That’s all that interests me now hence me generally never even saying boo to you or anyone for that matter; simple because the thread content doesn’t interested me.

Hell if we could have a civil conversation about coding I would be happy, but the only fellow that ever engaged me in such topics is at XCPU's and their forums are still buggy so I don’t post there.

So this is how it is Baron, you want to talk the politics of the industry I am all for it, but I won't actively try and engage in a conversation. When the conversation gets into the ins and outs of the IC hardware we all adore by golly son ill be there with my blue hat on and my green shirt (yes I am aware I have nothing to cover my shame, and darn it I likes it that way).

*smile*
Word, Playa.
a c 126 à CPUs
a b À AMD
August 20, 2008 3:37:37 AM

^well said spud. Well said.

Oh don't forget your red shoes (ATI) and your light green pants (nVidia)...... :p 
August 20, 2008 6:08:24 AM

BaronMatrix said:
The process has very little to do with heat production. That's MHz and amount of active logic. The process can be tuned for better transistor perf but any 40nm process will be approximately the same for the same circuits. As HiK shows, it's more materials than process (the same companies sell everyone their litho stuff). It just sounds like you can't stand anyone but Intel making any kind of ICs, especially if they're good at it.


Baron, I have to call you on this. The process means literally everything about heat production. (Well, aside from what kind of design you are going for). While you are absolutely correct that heat grows (supra linearly) with switching speed, it is most certianly NOT true that a set of transistors from different companies will give you the same thermals.

How, then, do you explain the greater thermals on the AMD chips, clocked at the same speed as intel chips? At 65nm? Where "materials" don't come into play?

I hate to break it to you, but *materials* ARE a fundamental part of the process. You'll also find (if you ever became a FSE and worked on the litho tools) that the tools are only *part* of the story -- it's how the company sets them up -- and the recipies are a VERY CLOSELY GUARDED SECRET. You don't know how many times they tell the engineers NOT to leave screens unattended, or let the FSEs go wandering about. The tools are almost secondary, it's the setup that counts.

And then, you're discarding all of the strain/stressors that go into improving performance at a lower voltage (to give you better thermals). Or the changes in gate oxide thickness (modulated thermals). Or dopant levels. Or channel lengths. Or contact resistance. I could go on and on. Each and every one of these is a finely tuned piece of the overall process we call "45nm".

To say that frequency is the sole driver of heat production is a complete fallacy.

And if you're not convinced -- go check out the IDEM data. And how you see some transistors with a higher switching speed, but lower leakage.
!