Interesting post from intel engineer on 100W cpu's

[Guest]

I just ran accross <A HREF="http://forums.anandtech.com/messageview.cfm?catid=27&threadid=1101996" target="_new">this post </A> on Anand's forum from apparently an intel engineer. Here a few relevant quotes:

In the past power has increased because the industry crammed more and more transistors into a smaller and smaller space. The smaller transistors burned less power as they were shrunk but designers kept adding more of them. So power gradually started to creep up. But more recently power has been increasing at an even faster rate because transistors have gotten so small that they are acting less like switches and more like leaky faucets; essentially they leak power. With each successive generation going forward (0.09um, 0.06um, etc.) it is going to get worse. A specific type of leakage called "gate leakage" increases by 10x with each successive generation. Another type called "subthreshold leakage" increases more slowly with each process shift, but it has a head start over gate leakage and so it's the bigger problem currently (pun unintended).

<b>At 0.25um, leakage wasn't really an issue. At 0.18um, it was a minor annoyance. At 0.13um, it is a serious problem. Going forward at 0.09um, it becomes a dominant contributor to the total chip power. And it will get worse.</b>

For what's it's worth as well, having worked on a 100W+ microprocessor (the Itanium 2), I can speak with experience in saying that <b>the problem with a 100W microprocessor isn't cooling it. It's getting power to it. </b>Changing the numbers slightly to keep the math simple, 100W at 1V is 100 amps. And that requires serious power delivery engineering in order to deliver 100A to a CPU that is constantly changing it's current loading while holding the voltage fixed at 1V. <b>No one is particularly worried about cooling at this stage... it's more how expensive the power regulation systems are going to get going foward. </b>

When I say that cooling is not a problem, I am specifically speaking to cooling a CPU. As far as cooling an entire system in which the parts inside are quite a bit hotter as well, I can agree that this will be a challenge, but it certainly is easier (and cheaper) than trying to design a 100A 1V power delivery system. If anyone thinks that cooling 100W is harder than powering 100W@~1V, I need to introduce them to my world.

I'd like to add one comment on this though; this guy has been working on Itanium, which even though it also dissapates >100W is a huge die. thermal density of Itanium 2 is 0.34W/mm². that is not all that much really. Prescott is going to have a thermal density 3x as high, around or over 1W/mm². To put that in perspective, the core of a nuclear powerplant reaches thermal densities of ~2W/mm². Cooling Itanium may not be much of an issue, especially considering its native loud airconditioned habitat, but I doubt cooling Prescott will be either easy or cheap. If power regulation is an even bigger problem even at 100W (~100A), that is going to get exponentially worse as voltages keep dropping and power consumption keeps increasing; things look not too bright for the medium term future..

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Parhelia]

problems will still come with the 0.09um, but for 1w/mm^2, [-peep-] this is a real chalenge for intel...

 

[Mephistopheles]

I think that maybe the problem is that costs will go high, because of the increased complications associated with voltage regulators and cooling... I mean, it's not impossible to implement watercooling on desktops, but hey, won't that be expensive and somewhat cumbersome?... So Intel has to think its way out of this carefully...

<font color=red><b>M</b></font color=red>ephistopheles

 

[sargeduck]

At 0.25um, leakage wasn't really an issue. At 0.18um, it was a minor annoyance. At 0.13um, it is a serious problem. Going forward at 0.09um, it becomes a dominant contributor to the total chip power. And it will get worse.

cooling isn't going to be the problem as he said. The reasons they are having problems with Prescott is becuase of the leaking gates. This is probably why Teja is pushed back, it was leaking too much. Amd is going to run into the same problem......

Is there anything really more to say?

 

[eden]

I'm really interested in seeing if the pattern of better process success continues up on the 0.09m with AMD having less trouble than Intel this time.

--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>This just in, over 56 no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A> :lol:

 

[Guest]

Captain Obvious says yes! That sounds logical.

Intel have shown their latest high end coolers based on a central copper core with spiral fins! Very effective!




<b><font color=red>Captain Obvious To The Rescue!!!</font color=red></b>

 

[guesswhosback]

maybe AMD's foward thinking with SOI will pay off with the .09 hammers

 

[Guest]

>I'm really interested in seeing if the pattern of better
>process success continues up on the 0.09m with AMD having
>less trouble than Intel this time.

No guarantees, but this is to be expected as AFAIK AMD is using the same lithography tools for .13 and .09. So .13 was the difficult shrink for them whereas intel is on a alternate path of using the same litography tools foor .18 and .13, making .13 relatively easy, and now they face the tougher shrink to .09. This being said though, I'm not sure they actually stumbeling over the .09 process, which seems to be relatively healthy, what we are seeing is probably more a prescott issue. I guess Dothan's numbers will prove me right or wrong.

Anyway I expect intel to have a relatively easy .065 conversion, while AMD will be facing a much tougher challenge going to this size. OTOH, AMD has already invested a huge ammount of effort in understanding and designing for SOI, which may pay off. Intel may have a lot of learning to do on that one.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Guest]

>cooling isn't going to be the problem as he said. The
>reasons they are having problems with Prescott is becuase
>of the leaking gates.

Hu ?? The only real problem with leakage is the resulting power consumption, which translates into current delivery issues and heat/cooling issues.

>This is probably why Teja is pushed back, it was leaking
>too much.

Where did you read that ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Guest]

>I think that maybe the problem is that costs will go high,
>because of the increased complications associated with
>voltage regulators and cooling..

Exactly. It is possible to cool >100W and it is possible to deliver the current, but at what price ? Price of these probably isnt much of an issue in the high end where prescott is going to compete initialy, but this will be a problem once prescott and prescott based celerons move down to mainstream and budget systems. Celerons have ASP's of maybe $50, if they require a ~$15 (intel price) copper cooler instead of a $5 one and a $15 more expensive PSU and motherboard, that is a huge problem.

BTW, AMD faces a somewhat similar challenge with their SOI process. SOI wafers are twice (? as expensive as bulk silicon, probably not an issue for >$400 cpu's but this may eat into their margins once SOI hammer based cores reach <$100 pricepoints.

Maybe we will be seeing the end of $50 budget cpu's ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Mephistopheles]

Oh man... I hope you're wrong!...

<font color=red><b>M</b></font color=red>ephistopheles

 

[eden]

Well, AMD already operated on profit loss with most of their sub 100$ CPUs anyways, so it's nothing new. :tongue:

--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>This just in, over 56 no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A> :lol:

 

[Mephistopheles]

Maybe you're right... I guess the price/performance curve is something to keep in mind... So if you take a bit from the lesser CPUs, the top of the line naturally has to get more and more expensive...

<font color=red><b>M</b></font color=red>ephistopheles

 

[slvr_phoenix]

If AMD is smart though they'll stop screwing themselves like that though. There's a reason that AMD has been losing money...

<pre><A HREF="http://ars.userfriendly.org/cartoons/?id=20030905" target="_new"><font color=black>People don't understand how hard being a dark god can be. - Hastur</font color=black></A></pre><p>

 

[HardWareBoss]

That's a big IF!!!!! they get smart.

 

[slvr_phoenix]

Yeah, well... It's not like AMD is a business that's trying to make money.

Oh wait...

<pre><A HREF="http://ars.userfriendly.org/cartoons/?id=20030905" target="_new"><font color=black>People don't understand how hard being a dark god can be. - Hastur</font color=black></A></pre><p>

 

[HardWareBoss]

Why make money? There stock is going up and their losing big doe.

 

[sonoran]

>>>If AMD is smart though they'll stop screwing themselves like that though. There's a reason that AMD has been losing money...

And Hector Ruiz clearly is aware that they cannot continue business like that. Their recent financial analyst calls have indicated they plan to get their ASP's (average selling prices) up to a profitable level, instead of giving away product to try and garner additional market share.

 

[sargeduck]

>cooling isn't going to be the problem as he said. The
>reasons they are having problems with Prescott is becuase
>of the leaking gates.

Hu ?? The only real problem with leakage is the resulting power consumption, which translates into current delivery issues and heat/cooling issues.

Ok, maby I'm wrong with this one, but when I read this article I thought of brake lines. If your brake lines were leaking, would the solution be to keep adding more brake fluid into your car? That's a temporary solution. The permanent solution is to stop the leaking. That's why in my first post I said cooling wasn't the problem, it was the leaking. If you don't fix your brake lines, it will only get worse, and you'll just have to start adding insane amounts of brake fluid. If they fix the leak, they fix all the power/cooling issues.

This is just the way I see it

>This is probably why Teja is pushed back, it was leaking
>too much.

Where did you read that ?

I didn't read that. It's pure speculation that I got from reading the article.



Is there anything really more to say?

 

[Guest]

> If your brake lines were leaking, would the solution be to
>keep adding more brake fluid into your car? That's a
>temporary solution. The permanent solution is to stop the
>leaking.

LOL.. funny analogy..what do you think is leaking out of those Prescott cpu's that needs to be added all the time ? if anything: current, therefore heat. Adding that all the time is one problem (power delivery) but getting rid of the excess "brake fluid" (ie heat) is also a problem.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[pied_piper2004]

Statements like "Amd is going to run into the same problem......" drives me me mmmmmmad...

How are you qualified to say such a statement? Prove it? Got a link?

People just can’t keep it to Intel, they have to drag AMD in there… why read the topic header… is Prescott an AMD processor?

Besides, how do you know really know? Do you fully understand SOI?

Its probable/more than likely they will have problems too... put don't post it as if it were true...

 

[pied_piper2004]

Yeah, because fighting Intel is an easy thing... hey you should be their new CEO?

Ever think no matter how smart or how perfect they are they just might not be able to fight Intel? AMD as to do what a lot of companies don’t have to face.

In the meantime I’ll enjoy their low pricing… (even if it doesn’t last)

Tell me, could they have not reduced their prices to extreme lows to combat Intel… what were their alternatives?

Sure AMD is not perfect, but all the business man wannabes here (inc. Omid) talking as if they could run a company better than AMD that just so happens to be fighting a top ten company of all time with a stock pile of cash to thow at anything(incredible uphill battle IMHO) is something I doubt very much.

Never take financial advice from someone who is poor

Never take business advice from someone who does not run a business of their own.

-pied

 

[sargeduck]

I was making that comment based on the following observations: AMD had trouble when switching to .13. Nvidia had problems when moving to .13. Intel is currently having problems with .09. You're right, I'm not qualified to say such a statement. But, looking at the past, and the present, I am allowed to make a prediction. And that's all that is, is a prediction. If you were to say, for example, "Nvidia's new fx 5700 will have extremely poor dx9 performance", you are just predicting. Is that a valid reason for me to get huffy and puffy, and then go on saying how mmmmmmmmmmmmaaaaad that makes me? You're just making a prediction based on past and present situations.

And you're right, I don't fully understand SOI. perhaps you could explain it to me?

Is there anything really more to say?

 

[sargeduck]

Interesting post from intel engineer on 100W cpu's

People just can’t keep it to Intel, they have to drag AMD in there… why read the topic header… is Prescott an AMD processor?

The Intel engineer was making a generalized statement. This statement would apply to Via processors, ATI, Nvidia, XGI, S3, Cyrix, 3dfx, Matrox, ect... Any company that does/did make cpu's/gpu's, is more than likely going to run into the same problem. (My prediction). If cyrix were to announce a new .09 processor, do you think they would be able to pull it off without any trouble? I merely said AMD becuase nobody cares about the other companies listed in this thread. If you want, I can edit AMD out and put the Via C3 in for you....

Is there anything really more to say?

 

[Guest]

Leakage would definately be a problem for AMD if they where using bulk silion (or strained silicon like intel). But AMD opted for SOI for this very reason, so its reasonable to expect AMD to have a bit less of an issue with this than intel; though time will tell...

= The views stated herein are my personal views, and not necessarily the views of my wife. =