10.03.2007

The Story Behind Intel's Breakthrough Transistor

Here's an interesting article about the development of the halfnium based high-k plus metal gate transistor that will debut next month on Intel's 45nm processors. Written by those deeply involved with the research, it shows a glimpse of how Intel's R&D diligently work on finding solutions to problems that are predicted to arrive only in a decade or so. The research for the replacement of the SiO2 gate oxide started ten years ago and was completed in 2004. The decision to implement the technology for 45nm in 2005 is a testament to Intel's lead with the rest of the world. Leading the 45nm transition is a feat by itself. But doing so with a key technology advancement makes it even more frightening.

The High-K Solution

The article may also leave you with a conclusion that without this new high-k technology, AMD's 45nm processors will be disadvantaged in terms of power efficiency and performance customisation. As Nehalem arrives at a time when 45nm matures, a strong microarchitechture standing on top of a highly advanced transistor technology is probably the last thing AMD needs.

"The invention of high-k plus metal gate transistors was an important breakthrough. Although we could have continued to shrink transistors to fit the dimensions needed for the 45-nm generation without this breakthrough, those transistors would not have worked much better than their predecessors, and they certainly would have expended more watts. We're confident this new transistor can be scaled further, and development is already well under way on our next-generation 32-nm transistors using an improved version of high-k plus metal gate technology". - MARK T. BOHR

31 comments:

Anonymous said...

funny how intel is breezing through 45nm while AMD is struggling with 65nm. you can only imagine compounding problems every time AMD moves to smaller geometries.

Mo said...

and sci has resorted to personal Attacks.
Mo Said "So in other words, the prediction you had made about AMD breaking even in Q4 is wrong and you would like to adjust it."

Sci said:
Mo, have you stopped beating your wife?

Why is it so difficult for you to repeat things correctly? AMD made a forecast that they felt they could break even by Q4. This was not my prediction as you've tried to claim. I've talked about AMD's statement before and never said that I agreed with it. What I'm saying now is that I doubt that AMD can hit break even in Q4.

Anonymous said...

MO, Break even? They're looking at another 600 million loss for 3Q! Tell that moron 4Q, too. :)

Sparks

Anonymous said...

apparently the "no personal attacks" rule doesn't apply to him.

he's been caught over and over to sensor posts that proves how ridiculously wrong he was. this time around we know he's a hypocrite.

Anonymous said...

high-k vroom vroom


Sci said:
Mo, have you stopped beating your wife?


Ok, that is just low. Spoken like a true wifebeater. That's probably how she got cancer or whatever.

Anonymous said...

Additional story behind HighK/metal gate... the real difficulty is in the integration into a full transistor flow and this is why virtually every roadmap in the industry had this technology in the 32nm or beyond despite the many papers and claims of how people have demonstrated performance. It also uses new equipment (ALD) which has not been used in high volume (though it has been used to some extent in areas like disk drive manufacturing and in DRAM, though with each of these different materials are used). So in addition to Intel proving out the technical solution they also need to ensure that they had a manufacturable one and that meant a lot of significant work being done on equipment that was not quite ready.

There have been papers since early 2000's (probably late 1990's) on capacitor based data showing tremendous leakage improvements or very good EOT (equivalent oxide thickness - basically how thick the material would be if it were SiO2). However what matters is CET - I forget the acronym, but this basically is how the thickness appears in a fully inverted transistor which includes more than just the EOT, it also includes some quantum effects and poly Si depletion if you are not using a metal gate. This is how a true transistor behaves, not a simple capacitor (some call it a CV dot)

The problem (or one of them) has been Vt pinning (that's the threshold voltage to turn on a transistor)... and getting Vt correct for both NMOS and PMOS devices. Obviously as Intel has demo'd this a while ago and they have figured it out - it is not clear yet whether IBM has fully dialed in the Vt on an integrated flow (it can be more easily done on short loops or test vehicles). There announcement was "we can do it and it's feasible" (not we have DONE it and it is manufacturable) - you'll also notice that AMD's stance has been late 45nm or 32nm (which means 32nm) - so my guess is they have the basics done but may not have an fully integrated flow operating at final Vt and other transistor targets or at least one that will yield well, meet performance and have no reliability issues.

Finally you'll notice AMD has claimed an ~20% target performance gain on 45nm, vs the more traditional targets of 30-40%...this means no high K on 45nm and supports the issue with the lack of SiO2 (or more technically SiON) scaling. Simply making the transistor smaller theoretically speeds things up but you run into SCE (short channel effect) issues which start eating away at those theoretical gains.

This is yet another reason why Scientia, Abinstein, Sharikou, et. al are DELUSIONAL when they keep stating AMD is closing the gap on Si technology nodes by using release date and mere feature size as the sole comparison. It is not surprising to me as they all clearly lack a legitimate background in Si process technology and manufacturing, but I am surprised at the multitude of people who simply accept what they say as if they were experts in this area.

AMD is starting to lag FURTHER on performance and this is evident, but is conveniently spun with their PR folks in their CTI (continuous transistor improvement) approach. AMD is still trying to "catch up" on the current node performance 6, 9, 12 months after "release" while Intel is hitting process targets AT RELEASE. The "45nm" process that AMD releases in H2'08, some 9-12 months behind Intel in terms of release date, will be an inferior process that will take at least 9-12 additional months to match performance from a Si process technology perspective.

The continuous improvement process is not taking a "good" 65nm process and making it better - it is taking a glorified 90nm process with 65nm dimensions and slowly implementing things to improve performance (improvements that Intel puts in by the START of the ramp). This is why it should be NO SURPRISE that the initial 65nm parts were not better on the speed end at the initial ramp. (BTW - power is improved due to the lower Vt targets of the new node).

All this said the their is a key benefit in that AMD enjoys the economic advantages of the smaller die earlier then they would if they had waited for a mature process (in terms of performance), but it does add in a whole new set of complexities in terms of how the design teams intercept specific revisions,factory planning if a new reveision means new equipment and a variety of other complexities.

Anonymous said...

To the anonymous essay above:
Nice job.

The problem (or one of them) has been Vt pinning (that's the threshold voltage to turn on a transistor)...

The physics term for this is Fermi pinning so others may google it if you like. Charge traps at the interface create a charge differential driving up Vt voltages (Vt high = bad thing).

This is yet another reason why Scientia, Abinstein, Sharikou, et. al are DELUSIONAL when they keep stating AMD is closing the gap on Si technology nodes by using release date and mere feature size as the sole comparison.

You hit that one squarely on the head.

EOT (equivalent oxide thickness - basically how thick the material would be if it were SiO2).

To expand on this, electrical thickness is the thickness you would calculate electrically based on the geometry/capacitance of said oxide. Electrical thickness and actual thickness are typically not the same (due to gemeotric effects). To others reading.... what this fella is saying is that, for example and just an example, a 200 angs high-k film acts like a 40 angstrom SiO2 film electrically, i.e. at the same dimensions would give the same capacitance. Being able to go thicker turns of the parasitic gate leakage due to quantum tunneling.

Overall nice post.

Anonymous said...

Ok, that is just low. Spoken like a true wifebeater. That's probably how she got cancer or whatever.

While I agree, Scientia's comment was a low blow, in this instance it is not kosher to fight fire with fire, Scientia and family went through a hard time during that period.... we can, and should, extend our sympathies for what he went through.

Unknown said...

That low life tries to defend his statement by telling me about an entertainment icon "Grouch Marx" Born in 1890 who produced films in the early 1900s..

Mo, my apologies for assuming you were familiar with an icon like Grouch Marx. The link will explain it.


He provides a DEAD link....

He later attacks my intellect

Again, I'm sorry you've never heard of Groucho Marx; I'll try to make fewer assumptions about your knowledge in the future.

Can someone explain to me how i'm suppost to take a comment like "mo, have you stopped beating your wife?" and somehow directly associate to a comedian in the early 1900s considering I'm not even an american native?

I replied to him telling him that his link was broken like his theories.....

I think i also said that "atleast i'm not the laughing stock of the enthusiast world, no wonder you only resort to your blog and AMDZone".....

He deleted that post and replied with another link to Groucho marx and behold, another broken link. lol

I love how he pretends he knows shit and posts it on his "blog" and Amdzone but doesn't dare post anywhere else. He knows damn well that people who actually know shit would rip him a new one.
Thus he resorts to his lame blog and that hellhole amdzone.

I also love how he loves talking shit about George Ou on amdzone by dedicating threads but doesn't dare post on George's blog which I beleive is Public lol.

GutterRat said...

mo,

Scientia, abinstein, and sharikou are tools of the AMD borg.

At least with Sharikou, we know the side of the fence he's standing on.

Scientia and abinstein claim to be unbiased but at the end of the day we all know they are AMD fanbois.

What really blows me away is that they are not willing to accept that they are AMD fanbois.

It's easy to see where people stand.

I am sure that Scientia, abinstein, and Sharikou participate in replies to George Ou and in other forums where AMD is challenged.

Anonymous said...

GutterRat and the others,

"Scientia, abinstein, and sharikou are tools of the AMD borg."

And you and the others are tools of the Intel borg.

"At least with Sharikou, we know the side of the fence he's standing on."

And with you, we know the side of the fence you are standing on.

"Scientia and abinstein claim to be unbiased but at the end of the day we all know they are AMD fanbois."

And you and the others also claim to be unbiased but at the end of the day we all know you are Intel fanbois.

"What really blows me away is that they are not willing to accept that they are AMD fanbois."

What really blows me away is that you and the others are not willing to accept that you are Intel fanbois.

"It's easy to see where people stand."

And you know where you stand.

"I am sure that Scientia, abinstein, and Sharikou participate in replies to George Ou and in other forums where AMD is challenged."

I am sure that you and the others participate in replies to any place where Intel is challenged. But I bet you are going to say you never post in Sharikou blog for example.

Anonymous said...

Can someone explain to me how i'm suppost to take a comment like "mo, have you stopped beating your wife?" and somehow directly associate to a comedian in the early 1900s considering I'm not even an american native?

Maybe you can’t, but if you can’t (“obviously”) associate that, how can you be sure he even knows you have a wife?
Even the easy task of opening a link you can’t do it, what do you want?

The funny part is that you all can write bad words to him, but you can’t ear them from him. Funny… and even have the displacement of saying you all want free speech. Funny guys.

Anonymous said...

What really blows me away is that they are not willing to accept that they are AMD fanbois.

They accept it, they just don't intend to admit it openly. It's all part of the trash talk on these sites. It's why you take it with a grain of salt, because all of the fanboi-ism in the world doesn't affect what happens in the market.

It's entertaining to read most of the time, though.

GutterRat said...

Anonymous Coward

Wha, wha, wha!

Show us you you really are

I don't hide under my mommy's skirt, unlike you.

You AMD fanbois are the best. Always giving me plenty of material for my blog.

Cowards.

Anonymous said...

Mo, regarding Scientia ...

I understand your frustration, it is more effective to disregard it and counter his 'facts' with the correct data and logical argument (similar to that chart reading exercise, that was actually funny).

In the end, Scientia post things with such a confident, assertive tone that many are fooled into thinking he really does know his stuff.... just remember, those people who follow it blindly are the same ones who would not want to hear it any other way. :) .... sorta like being unplugged from the Matrix, the mind just cannot handle it :)

In the end, credibility is estabilished when the data is argued and the logical conclusion from such data can be arrived at concisely.

There various aspects of this Intel vs AMD debate that will never be settled because, frankly, the data to settle such debate has never been made public (for example, who has better yields?) ... in the end you cannot logically convince someone that they are wrong if they are so emotionally attached to their thesis such that the attachement robs them of their capacity to reason.

Typically, when someone is losing the debate or cannot provide the compelling argument they resort to asking 'have you stopped beating yoru wife?' or they will spell check your posts :) ....

Jack

Anonymous said...

They accept it, they just don't intend to admit it openly. It's all part of the trash talk on these sites. It's why you take it with a grain of salt, because all of the fanboi-ism in the world doesn't affect what happens in the market.

It's entertaining to read most of the time, though.


:) Well said.

Best to ignore it really.

Anonymous said...

“I am sure that you and the others participate in replies to any place where Intel is challenged. But I bet you are going to say you never post in Sharikou blog for example.”

Perfect Example! I never have and never will. All the ranting, arguing, and groundless factoids aren’t going to change one simple fact. Abeinstein, Shar-a-koo-koo, and the other pinhead have been consistently WRONG about everything they’ve posted. They’ve blown rays of sunshine up every poor bastard’s ass who would believe them.

Fact, AMD has failed miserably. Yields, production, execution, target dates; all of it, has been absolutely the worst. They’ve had heavy losses in 3 consecutive quarters, and are looking at more losses. Key executives are bailing, and they can’t turn a profit.

What I do know, is that I’ve made over 8 dollars a share on INTC. Their chips are dominating the market and they are ramping to 45nM flawlessly. New super fast chipsets (X48) are being built as I write and Yorkfield has tremendous headroom.

The guys on this site have EXTENSIVE, hands on experience in chip fabrication and they’ve been RIGHT all along! READ THE HISTORY!!

I don’t know what the color of the sun is on your planet, but on this planet, AMD’s horrible performance and execution as apposed to INTC stellar performance, is in diametric opposition to one another.

Are AMD’s shareholders happy? Are their employees happy? Are AMD fanboi’s happy. Is Wrecktor Ruinz happy? ARE YOU HAPPY WITH WHAT THEY HAVE TO OFFER? Are you still waiting for a chip from AMD that will blow away INTC?

Keep waiting. It ain’t gonna happen.

Does this constitute me as an Intel fanboi? At eight bucks a share, Intel’s performance, and with the PhD’s on this site, you bet your sweet ass I it does. What guys like you in denial need is a good reality check. Open your minds, and look at the fundamentals. Things aren’t looking good for AMD. If you still think AMD looks good, put your money where mouth is and buy a couple of hundred shares. If not, go back to dream world and keep reading that other rubbish you’ve become addicted to.

AMD’s days in the sunshine are over. They have no money, and Barcelona has fallen flat on it ass. This is gospel.

SPARKS

Unknown said...

At least with Sharikou, we know the side of the fence he's standing on.

Sharikou might be delusional, but at least he doesn't claim to be non biased like Scientia and Abinstein. I've also never known Sharikou to censor posts like on Scientia's blog.

If someone proves Scientia wrong - the posts will be removed. In an older post from a few weeks ago Scientia was comparing performance in LINPACK between Dual Xeon 5150 and Dual Barcelona 2347. The site he was getting the information from didn't list the number of cores the CPUs had, but Scientia ignored this and claimed it as a fair comparison because the AMD setup won. But he was comparing 4 cores to 8 cores so several people pointed it out to him, and he deleted all those posts.

Anonymous said...

But you do the same he does Jumping jack. That’s the funny part!
You filled the Toms forums with complete FUD. Some times you even explain the FUD with (invalid/misleading) data like he does! That’s really funny!!!!


Typically, when someone is losing the debate or cannot provide the compelling argument they resort to asking 'have you stopped beating yoru wife?' or they will spell check your posts :) ....

Or resort to name calling and trolling that is basically the same, Jacky!

Anonymous said...

Gutterrat moron!

You are a complete idiot! And you don’t hide like me moron?
Tell me what’s the difference in an anonymous post and a gutterrat post?
I didn’t know I could go to your address with just the Gutterrat name! Gutterrat is very explicit indeed! Maybe it is possible; I may go there and make you my bitch.
Moron Gutterrat is just an Anonymous Coward with a Name.

Anonymous said...

Scientia's blatant denial of bias and censorship can lead to only one explanation. This is where AMD's marketing is putting its resources.

Shame shame.

Unknown said...

Anonymous is no doubt Baron Matrix. JumpingJack was a very respected member of the Toms Hardware Forumz until his departure.

Please come back JumpingJack. We miss you. :-(

Anonymous said...

But you do the same he does Jumping jack. That’s the funny part!
You filled the Toms forums with complete FUD. Some times you even explain the FUD with (invalid/misleading) data like he does! That’s really funny!!!!


Hey Baron how are ya'

If you would point out a post for me where I am guilty of this accusation, I will either back it up with the data or retract it :) ... just because you don't like what you hear doesn't mean it is FUD.

Anonymous said...

Anonymous is no doubt Baron Matrix. JumpingJack was a very respected member of the Toms Hardware Forumz until his departure.

Please come back JumpingJack. We miss you. :-(


No .. not possible anyway.

If you want to chat .. www.xcpus.com

Jack

Anonymous said...

Reposted, I hate anonymous posts:

But you do the same he does Jumping jack. That’s the funny part!
You filled the Toms forums with complete FUD. Some times you even explain the FUD with (invalid/misleading) data like he does! That’s really funny!!!!


Hey Baron how are ya'

If you would point out a post for me where I am guilty of this accusation, I will either back it up with the data or retract it :) ... just because you don't like what you hear doesn't mean it is FUD.

Anonymous said...

jumpingjack, where do you post now?

Anonymous said...

www.xcpus.com

BTW -- the mass exodus has earned that URL a filter in Tom's HW forum filter... too funny. We catch their IP lurking now and again.

GutterRat said...

giant, wrote

Anonymous is no doubt Baron Matrix

Yes, it's evident :)

Dressed like a silver phallus, topped off with a schmuck.

If the shoe, umm, costume fits, well...

Anonymous said...

Baron Matrix is a closet homosexual.

He hangs out with Perez Hilton.

What a biatch!

Anonymous said...

Baron Matrix,,

How come you don't return my phone calls?

Call me ASAP please. I am sorry for the hurtful things I said to you the other night in the jacuzzi.

What you and I have is special. Let's not throw it all away.

XOXOXOXOXO
abinstein

PS You have great stroke

Anonymous said...

Baron Matrix and abinstein?

Sci will be jealous. I wonder if he knows he's being two-timed?

Pwnd all of them AMD fanboiz