backtop


Print 42 comment(s) - last by verndewd.. on Jan 31 at 8:16 PM

IBM to enlist high-k metal gate technology with its 45nm chips

Alongside Intel’s news of 45nm process technology, IBM today announced its own 45nm technological advancements that apply to products manufactured in its East Fishkill, NY plant starting in 2008.

Working with AMD and its other development partners including Sony and Toshiba, IBM has found a way to construct a critical part of the transistor with high-k metal gates, that substitutes a new material into a critical portion of the transistor that controls its primary on/off switching function. The material provides superior electrical properties compared to its predecessor, silicon dioxide, enhancing the transistor's function while also reducing leakage.

As important as the new material is the method for introducing it into current manufacturing techniques. The creation of this transistor component with the new material was accomplished by the IBM team without requiring major tooling or process changes in manufacturing - an essential element if the technology is to be economically viable.

“Until now, the chip industry was facing a major roadblock in terms of how far we could push current technology,” said Dr. T.C. Chen, vice president of Science and Technology, IBM Research. “After more than ten years of effort, we now have a way forward. With chip technology so pervasive in our everyday lives, this work will benefit people in many ways.”



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

45nm
By IntelUser2000 on 1/28/2007 5:31:08 AM , Rating: 3
If some people thought Intel was the only one knowing marketing tactics, they are wrong. Announcing their 45nm process the same day Intel announces is to calm the IBM/AMD investors/fans for the fact they don't really have a near term solution to Intel's 45nm. It's pretty good imo.

Funny, they were also dismissing Intel's High K soon as Intel announced it. AMD also said they'll catch up at 65nm node, if anyone actually bothers to check the closest AMD had to catching up Intel was the 180nm node, which was 5 months. Intel sure did increase the lead.




RE: 45nm
By Viditor on 1/28/2007 7:49:09 AM , Rating: 2
quote:
Funny, they were also dismissing Intel's High K soon as Intel announced it

As you say, marketing tactics...
I seem to recall a great deal of derision from Intel over x86-64 as well.


RE: 45nm
By Lonyo on 1/28/2007 10:40:39 AM , Rating: 3
Yes, but that was justified in some ways.
It's been 3 1/2 years or so, and only now (IMO) with the release of Vista are we going to see more widespread adoption of 64-bit versions of operating systems in the mainstream (server market not withstanding, I don't know what the adoption rate was there), but 45nm we could see in use by the end of the year, and (Intel probably hope) widespread adoption of 45nm processors too.

I've had an x86-64 processor for quite a long while, and I won't get an x86-64 OS until probably June of this year at the earliest. And the CPU to go with it will probably be a Core 2 Duo, which is a relatively new CPU, soon to be built on the 45nm process :P


RE: 45nm
By Smurfer2 on 1/28/2007 10:52:42 AM , Rating: 2
Lonyo,

I feel the same way. Vista will be my first 64bit OS too. I will first get a new processor. (64bit netburst CPU's are only so great, lol) Then at a later point I will get Vista, just not sure when yet.


RE: 45nm
By shady28 on 1/28/2007 11:45:16 AM , Rating: 2
There is something to be said for the Athlon 64 in terms of longevitiy compared to Intels previous products. ie, as 64 bit OS and applications hit the mainstream over the next few years, those who have a 64 bit A64 will see performance boosts (see link below showing 5-20% boosts in performnace on 64 bit os & software benchmarks).

Naturally core 2 is still faster in 64 bit mode, no question. However, Intel really hasn't had anything of value in the 64 bit world until the last 7 or 8 months. Keep in mind the original 'core duo' was not truly 64 bit.

None of this helps AMD looking forward, except to say that prior to Core 2 AMD was demonstrating a lot of vision in looking to the 64 bit world. It also means the A64 will have much longer legs than its Pentium 4 class rivals :

http://www.devhardware.com/c/a/Computer-Processors...

" This is why the newest Star Wars was rendered on AMD A64's and XP x64. They even used the beta version of XP x64. "


RE: 45nm
By IntelUser2000 on 1/28/2007 11:42:52 PM , Rating: 1
quote:
Keep in mind the original 'core duo' was not truly 64 bit.


Core Duo doesn't have 64 bit at all.

If you meant the current Core 2 Duo, yes it does have true 64-bit support. If you are talking about Core 2 Duo having lack of Macro Op Fusion at 64 bit, that's strictly a performance feature. It can run full "AMD64" regardless. The Pentium 4/Pentium D's before 2006 has partial implementation of the 64-bit.

Even then they are still full 64-bit. They are just not full AMD64.


RE: 45nm
By visaris on 1/28/2007 11:56:58 AM , Rating: 3
I disagree. I have been using a 64-bit OS for years. Further, all my apps are compiled for a 64-bit machine. Linux (Gentoo in particular) were supporting AMD64 in a meaningful way looong before Microsoft, and I've been there getting the benefit the whole time.

To say that Intel was justified in dismissing AMD64 is just silly. Intel tried to downplay the major advancement as long a possible, because it was effecting their precious Itanium line.

I see real world speed boosts of around 5%-10% when running 64-bit, and that is no small change when it comes for free. AMD64 will become the standard (already is), and from now on, instead of saying x86, what people will really mean is AMD64.


RE: 45nm
By Hyperlite on 1/28/2007 2:41:49 PM , Rating: 2
true, but what percent of the user base do you represent?


RE: 45nm
By Ringold on 1/28/2007 5:28:30 PM , Rating: 3
According to 2004 surveys, 2.8% of desktop computers ran a flavor of Linux.

10% of those desktop Linux boxes ran Gentoo (2006 survey).

Which means he represents .28% of the desktop market. :)

Though to be fair, several major distro's have 64bit version available, but not all. So. Maybe he represents 1% or 1.5% of the desktop computer install base. Linux hasn't exploded in growth since 2004, so probably still less than 4% or so of the current market.

In other words, not much. :)


RE: 45nm
By Viditor on 1/28/2007 11:23:56 PM , Rating: 1
quote:
and only now (IMO) with the release of Vista are we going to see more widespread adoption of 64-bit versions of operating systems in the mainstream (server market not withstanding, I don't know what the adoption rate was there),


That is indeed the key...remember that:
1. The server market is the most lucrative
2. The adoption of 64bit in servers has been VERY high
3. Windows 64bit, Linux, and even Solaris edition has been taking huge amounts of marketshare away from traditional Risc based servers.

You're right that consumer level is only just now happening, but x86-64 has been revolutionary in the server/commercial area for years now.

quote:
And the CPU to go with it will probably be a Core 2 Duo, which is a relatively new CPU, soon to be built on the 45nm process

Fair enough, and I don't blame you. But if you plan on 64bit, AMD still appears to have the edge...and 45nm won't be available to you for ~another year.
That said, if I were getting a consumer PC in June, it would also be a C2D...


RE: 45nm
By RockfordFile on 1/29/2007 12:16:26 PM , Rating: 2
What you're forgetting though is that it wasn't just "64-bitness" that AMD's strategy was all about. Intel spent 10+ years developing a proprietary 64-bit solution that required not only special motherboards and slots (to say nothing of a terribly humungous processor unit that ran very hot, soaked all kinds of power, and was very expensive), but a completely new platform. This forced any company/department thinking about upgrading to a supposedly better technology to have to consider the huge cost of switching EVERYTHING, hardware AND software.

AMD called Intel's bluff, spent far fewer years in development, delivered on promises (mostly) and ended up with an inexpensive, viable, market-tapping solution that did not require major hardware or software costs. AMD's strategy was about recognizing the economics of the customers Intel was aiming Itanium at, and acting intelligently to predict their concerns and address the overriding problem in Intel's strategy.

If you ask me, AMD made Intel out to be a fool for missing such a compartively simple and ingenius solution to the problem. 64-bit may still not be here, but it will be someday, and by then the younger generation of software engineers will already have cut their teeth developing on ancient AMD64 CPUs.


RE: 45nm
By JackPack on 1/29/2007 3:31:04 PM , Rating: 2
quote:
If you ask me, AMD made Intel out to be a fool for missing such a compartively simple and ingenius solution to the problem.

Nobody missed the obvious x86-64 solution. Intel's Yamhill x86-64 project began even before AMD's.

The point of IA-64 was to eliminate the Technology Exchange Agreement between Intel and AMD. IBM in the 1980s stipulated that Intel license its x86 chips to someone else (AMD) so that IBM would not be stuck with single source of supply. Intel wanted IA-64 to replace x86 so that it would be game over for AMD. IA-64 has been successful with commercial workloads and in the 16+ socket space due to its massive FP performance advantage. But general computing needs integer performance, an area IA-64 is not strong in.


RE: 45nm
By RockfordFile on 1/29/2007 4:38:11 PM , Rating: 2
That may be, but Yamhill or not, it wasn't Intel's core strategy until AMD released the Opteron and started outselling Intel and gaining ground with other vendors that previously wrote AMD off. It was a PR win for AMD for sure.

Hindsight may be 20/20, but Intel wanting to replace x86 with a new platform was a plan that had huge hurdles to overcome when IA-64 began, and those problems have only increased exponentially by this point. I think if anyone ends up moving away from x86, it'll be AMD, because I'm sure they don't want to be yoked to Intel to approve use of the technology, as much as Intel doesn't want to be yoked to AMD as a silent validation of AMD's competing business.

At some point they will diverge technologically, and AMD will have to work to differentiate and prove itself again, to remain Intel's most viable threat when running the same operating system in average households. They need to have another 5 years of homeruns on their roadmap to compete with the very strong 1999-2004 they had.


RE: 45nm
By Locutus465 on 1/29/2007 8:04:47 PM , Rating: 2
quote:
Yes, but that was justified in some ways.


Except Intel wasn't putting down x64 becaues they didn't want people to switch to 64bit, it's because they wanted people to switch to IA64 so they could be the only game in town.... Remember the old road maps that included cosumer level IA64 chips?


Great news
By verndewd on 1/28/2007 6:21:50 PM , Rating: 2
At this point a person has got to admire how Intel is pushing the industry,or seems to.I thought High K would be late,from all my reading at applied and other lithography news sources.To come out and do the High K so soon just makes me applaud intels intellectual capacity.

I have been an strict AMD user for years,But seeing how Intel can solve the HF0 issues by itself and ahead of schedule,is really impressive.Hf is not an easy thing to purify,according to the periodic tables its among the most difficult to seperate from impurity.

What Intel did alone IBM did with several partners;amazing.
Perhaps these partners will stay together in developement and keep such drastic leads in performance down to a minimum in the future.Either way its good news for all enthusiasts to have lower power and higher perf.So regardles of who what when and why we all win.




RE: Great news
By flipsu5 on 1/29/2007 3:22:35 AM , Rating: 2
I know some of the guys who were behind the high-k announcements. It's really very good (and hard) work.

I noted in the other thread for Life with Penryn that the Intel CPU TDP did not go down dramatically with the high-k metal gate. In fact, for some important chips, it went up. Imagine if there were no high-k there. It could be even worse (although circuit techniques can also help reduce leakage significantly, independent of high-k).

It leads to think that everybody has to use this for 45 nm CPU. One of the big concerns was this would be expensive, since there would be few users, but it may become established technology by the time AMD starts 45 nm.


RE: Great news
By Master Kenobi (blog) on 1/29/2007 8:36:58 AM , Rating: 1
Which will likely be around the same time frame that Intel is polishing its 32nm process.


RE: Great news
By flipsu5 on 1/29/2007 10:18:45 AM , Rating: 2
quote:
Which will likely be around the same time frame that Intel is polishing its 32nm process.


...assuming no surprises...


RE: Great news
By Viditor on 1/29/2007 10:28:19 AM , Rating: 2
quote:
Which will likely be around the same time frame that Intel is polishing its 32nm process


Cmon guys...at least put a smiley face there.
You just know that some newb will read this and tell everyone that he read on DT that Intel is pushing their 32nm process ahead by 1.5 years...


RE: Great news
By jak3676 on 1/29/2007 10:51:23 AM , Rating: 2
let me go post that on the inquirer - I'll say I saw it here ;)


RE: Great news
By verndewd on 1/29/2007 6:57:33 PM , Rating: 2
found a pdf through a link shared at toms from the register uk,it lists beryllium as the metal gate.
http://www.theregister.co.uk/2007/01/28/intel_ibm_...

and heres a periodic table description.
http://periodic.lanl.gov/elements/4.html


RE: Great news
By verndewd on 1/29/2007 6:58:57 PM , Rating: 2
RE: Great news
By flipsu5 on 1/29/2007 7:20:55 PM , Rating: 2
Thanks for the link. The metal gate was still not identified though.


Will AMD use high-K?
By flipsu5 on 1/28/2007 5:54:53 AM , Rating: 2
After reading the IBM release, not sure what to think. High-k wasn't mentioned at IEDM, but they claim to have a solution. I am sure IBM can have a process ready for 45 nm, but AMD would need to plan this for their 45 nm fab if they still want to do it.




RE: Will AMD use high-K?
By Viditor on 1/28/2007 7:46:26 AM , Rating: 2
quote:
High-k wasn't mentioned at IEDM, but they claim to have a solution

Well to be fair, Intel didn't mention it at IEDM either...


RE: Will AMD use high-K?
By ChipDude on 1/28/2007 6:41:21 PM , Rating: 2
Minor detail is that INTEL did NOT annouce their 45nm at IEDM while AMD/IBM did.

By design or not INTEL choose not to say much at IEDM. I can only suspect the delay was by design. I can ask why wouldn't IBM annouce HighK if it was indeed planned on their 45nm x86 line. I'm sure INTEL wanted to annouce functional Penrym which is what they have now. I can only concluded that perhaps they didn't have functional CPUs in hand at the conference deadline.

I'm not clear, but is sure looks like IBM has little to annouce but a reactionary annoucment timed to coincide with INTEL's with little/NO details, but a me too. If you look at INTEL's annoucement it seems much more prepared, pictures of wafers, fancy diagrams, and lastly a fully functional x86 64 bit CPU too.

I only refer to this interesting quote from an IBM executive. Does this sound like someone ahead or someone really caught with his pants down and reacting?

"It’s the difference between can openers and Ferraris"
...I.B.M. had simply chosen to deploy its new process in chips that are part of high-performance systems aimed at the high end of the computer industry. Its good to know that IBM executives see x86 CPUs as canopeners. I sure AMD is glad to know that is what the IBM guys think of them and their Dresden fab. Just a canopener. IBM is doing much more important things then bothering with HighK for x86.

I'm sorry what is more important then deploying the latest silicon technology on the most visible product, x86 CPUs.

Oh.. I know IBM was saving it for power, maybe that is the secret sauce that is going to get POWER to 5 GHz. Or maybe they are saving it to get higher yields on Cell. What a crock response.




RE: Will AMD use high-K?
By Goty on 1/28/2007 1:23:32 PM , Rating: 2
IBM's and AMD's production processes have been closely linked for some time now, so assuming that Intel and IBM see some tangible gains from using this technology, AMD will assuredly use it.


45nm vs. 90nm
By livelouddiefast on 1/28/2007 2:24:12 PM , Rating: 1
i was just thinking about this:
Excluding R&D, or rather counting it as a sunk cost, do 45nm chips theoretically cost half as much to produce (i understand different materials are now being used, so this won't be exact...)?

if so intel has the potential to keep their price wars going, and that's fantastic... now we just need a solid state drive price war/ technology race. delicious




RE: 45nm vs. 90nm
By Dactyl on 1/28/2007 4:18:14 PM , Rating: 3
If you have two chips, both 65nm, one with twice as many transistors as the other (so the smaller one takes up half as much die space), the smaller one will be less than half the cost to produce.

First, you can fit twice as many of the smaller chips on a wafer. Maybe slightly more than twice as many (because you're trying to fit square chips onto a circular wafer)

Second, you have to consider the defect rate. If on average X big chips per wafer have defects, on average more than X but fewer than 2X small chips will have defects.

Since the costs are basically the same either way, and you end up getting more than twice as many chips per wafer, yes, it costs less than half to produce a half-size chip.

---

But there are big differences when you move to a new process.

First, I don't know if the cost of the wafers and other materials changes when you're using 65nm or 45nm (I have no idea why it would or wouldn't).

Second, if you're using a different process, the defect rate will change. In fact, defect rates can drop during a process's lifetime as it gets tweaked. A newish 45nm process will almost certainly have a much higher defect rate than a tried-and-true 65nm process.

The exact figures are (probably) closely held secrets of IBM, AMD, Intel, etc. because this is really important information.

Eventually, the 45nm Conroe-equivalents should cost Intel about half or even less than half as much to manufacture as the 65nm Conroes.

But there's more to price than the cost per unit to manufacture, because manufacturing capacity is limited. Intel can't just snap its fingers and turn a 65nm fab on Monday into a 45nm fab on Tuesday. Even if the per-unit costs are half as much, if Intel can only produce 10,000 45nm CPUs a month, Intel will have no reason to sell them at half price. In fact, if their 45nm chips are real screamers, Intel can charge more for them (and would be stupid not to).

So in the near term, I wouldn't expect much difference in prices. Maybe when the first 45nm enthusiast chips come out, the 65nm enthusiast chips will go down in price a little--but remember how expensive the 65nm extreme edition Pentium IVs were even after Conroe came out?


RE: 45nm vs. 90nm
By Dactyl on 1/28/2007 4:21:01 PM , Rating: 2
(the "square chips" are actually rectangular, of course... sorry)


RE: 45nm vs. 90nm
By Viditor on 1/28/2007 11:33:09 PM , Rating: 2
Nice post Dactyl.

The one thing you forgot to mention though is that NOBODY keeps their design the same...
By that I mean that they add a lot more things into the chip which reduces those savings (in the case of Penryn, they are vastly increasing the cache and other things).
If the defect yields are equivalent, then all that really matters is the footprint of the CPU (how big it is). You can either make it smaller (and save a lot of money but the same performance), or add things to it (make it more competitive and more expensive), or even a little of both...

The bottom line is that we won't know if Intel's 45nm are cheaper to produce until we know it's size and defect rate (and we NEVER know the defect rate...).


IBM, AMD Announces 45nm Chip High-K Gate Process
By Nemesis 1 on 1/28/2007 12:18:49 PM , Rating: 2
I think A little reserch is called for here. This development is part of the Fab club whom AMD is not a part of. AMD is a partner of IBM but they are not in the fab club. I doubt the other fab club partners would be happy with IBM if they just gave tech they spent millions on to develop was just given to AMD.




By koss on 1/28/2007 1:44:20 PM , Rating: 2
Do not worry
As they are development partners, thus shall be applied to the patient of their discovery. Therefore implanting the technology into IBM's solution line would require payment towards AMD (basicly for the help). Furthermore both of the companies are not really in a competition, having in MIND who rules the current CPU market, both in TERMS of PDA/Laptops/Desktops/Servers (maybe rockets as well lol).
In my opinion, they will license the technology for each other, implanting into own products:
I see their partners using more or less in mobile devices (or pocket ones).
AMD is seeking nemesis from Intel in the FRONT line with desktop/server CPU's, while Cell is going 45nm with High-K.
I would bet things will get pretty messy this year, as the corporating business will tighten the things around.
AMD is major player and a complete solution developer as of ATI aquisition. This presented in terms of both CPU and home GFX market. I would see co-operation with IBM anyhow.

Otherwise plainly Intel got it all. Technology, plants power, resources... Whatever it takes to build another 5 or so years lasting empire. And it is ridiculous to speak that future is unpredictable at least at this point.
Besides Intel ALWAYS preserve a 25+% Laboratory GAP. They got C2D @3,8-4 GHz working right now I am sure.


By Viditor on 1/28/2007 11:35:07 PM , Rating: 2
AMD was listed in the credits for the development...


Nextgen CPU or microNuke!
By codeThug on 1/28/2007 1:48:33 PM , Rating: 1
I guess I'm a bit nervous about these new CPU's using Hafnium.

http://www.globalsecurity.org/wmd/intro/hafnium.ht...

Maybe IBM & Intel have finally found a way for CPU's to power themselves!!!




RE: Nextgen CPU or microNuke!
By verndewd on 1/28/2007 6:24:40 PM , Rating: 2
:) thats funny,that is in reference to Hf 128 an isolated isotope,one gram can hold several hundred megajoules of energy.

Hfo is an insulator that cannot absorb energy.


RE: Nextgen CPU or microNuke!
By verndewd on 1/28/2007 6:27:01 PM , Rating: 2
correction Hf 178.


IBM highK whitepaper
By verndewd on 1/29/2007 7:17:29 PM , Rating: 2
http://regmedia.co.uk/2007/01/28/ibmhighk.pdf

pwerbaselx at toms dug this jewel up.




"It seems as though my state-funded math degree has failed me. Let the lashings commence." -- DailyTech Editor-in-Chief Kristopher Kubicki

Related Articles
Life With "Penryn"
January 27, 2007, 12:01 AM













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki