90mm Dothans cooler than Banias?

[Goldstandard]

I just finished looking at the latest Anandtech article about Dothan. Here is an interesting quote:

For those who look at Prescott as an example of how "bad" Intel's 90nm process is, take a look at Dothan as a more accurate measurement. Without any architectural changes, 90nm allows Dothan to run cooler and faster - the opposite of what we've seen on Prescott, leading us to believe that the reason for Prescott's heat issues isn't Intel's 90nm process, but rather the architectural changes to Prescott.

I wonder what we can expect in temps for the 90mm A64's?

[ClarkKent]

Well I guess it depends on architecture as far as heat is concerned. Or that article would have us believe that. Whatever happens I can't wait.

[Rio71]

it is true. :cool:
compare banias with 1.70 GHz (http://processorfinder.intel.com/scripts/details.asp?sSpec=SL6N9&ProcFam=942&PkgType=ALL&SysBusSpd=ALL&CorSpd=ALL) with new dothan 1.70 GHz (http://processorfinder.intel.com/scripts/details.asp?sSpec=SL7EP&ProcFam=942&PkgType=ALL&SysBusSpd=ALL&CorSpd=ALL).


very confusing for me is both cpu (pentium-m and p4) are only made of transistors, resistors and capacitors.
wy the big thermal difference :confused: :confused:
what's the secret ?
both cpu have near the same number of transistors. (norty~banias ; pressy~dothan)

[@md0Cer]

it is true. :cool:
compare banias with 1.70 GHz (http://processorfinder.intel.com/scripts/details.asp?sSpec=SL6N9&ProcFam=942&PkgType=ALL&SysBusSpd=ALL&CorSpd=ALL) with new dothan 1.70 GHz (http://processorfinder.intel.com/scripts/details.asp?sSpec=SL7EP&ProcFam=942&PkgType=ALL&SysBusSpd=ALL&CorSpd=ALL).


very confusing for me is both cpu (pentium-m and p4) are only made of transistors, resistors and capacitors.
wy the big thermal difference :confused: :confused:
what's the secret ?
both cpu have near the same number of transistors. (norty~banias ; pressy~dothan)

Well, to tell you the truth I really cannot say, because I do not know, but there are allot of architectural differences between CPU's. It could be what the transistors are made of, or the process they are made from. Just as an example, perhaps the strained silicon process will allow them to go further than 90nm more successfully, but 90nm could be too large for that particular process and it is letting the power escape through the transistor gates in the form of thermal energy.

That above was just an example that I thought up. But, as I was trying to say in that example, there can be many minor quirks like that, that could cause it to throw off more heat. Also to take note of, Prescott has many architectural differences, amount of L1 cache, more instructions, a modified branch predictor, a longer pipeline, I am not saying these could cause the CPU to lose efficiency as in losing power in the form of heat, but there are many reasons that could be the cause. I am sure the guys and girls at Intel are trying to figure it all out as we speak.

Again, I do not know much about Intel, or even the latest AMD's for that matter, but does Banias use Strained Silicon like Prescott? That could be your answer right there.

-0cer

[aNTiChRisT]

Manufacturing smaller transistors obviously causes problems, but i cant say exactly what the problem is.

~t0m

[NookieN]

There's basically two explanations for why Prescott is very high power and Dothan is relatively low power. The P-M chips (Banias, Dothan, Yonah...) are designed specifically with power optimization in mind. Even though the chips are fundamentally related to the PPro architecture, they use power even more efficiently than their predecessors. A PIII 1.4Ghz with 256k L2 uses about 23W, while a 2Ghz Dothan with 2M L2 uses only about 21W.

The Pentium4 chips, other the other hand, are not designed around reduced power. They were designed for performance (snide comments aside), most specifically high clock speeds. This is where the major problem with 90nm comes in. A 2Ghz Dothan shows it's clearly possible to make a good processor on 90nm. But in order for a chip like Prescott, even with its much longer pipeline, to hit higher clock speeds (e.g. 4Ghz) the transistors need to be made smaller.

Smaller transistors of course mean that the gate oxide needs to be thinner. The problem that Intel (and IBM, and probably AMD) is facing is that when they make transistors small enough to hit those speeds, the gates become very leaky. As a result the chips require tremendous amounts of current, which inevitably means more heat.

AMD will most likely encounter similar problems if they try to design their 90nm chips to run at speeds much greater than the 130nm chips. SOI may help, but strained silicon helps Intel's 90nm process and clearly its not enough to overcome leakage problems.