Why is Prescott so HOT?

xeenrecoil

Distinguished
Jan 7, 2003
842
0
18,990
heya hornsen:

The short answer is...
Intel decided to goto 0.09 micron, and the consequence of that decision is increased wattage demand which causes thermal bleed, they are trying to cram alot of transistors into a very small space, and it produces a hell of alot of heat, even though they reduced the voltage they still cant get the heat under controll, the only way around it is to further reduce the voltage, and well they cant do it because they couldnt run the CPU at 3Ghz and beyond if they did, so their decision was to step up production on their dual core CPU, which will debut about a full year ahead of schedule. They can increase processing power, and decrease voltage, which in turn decreases wattage demand and excess heat. They have finally hit the proverbial wall at around 4Ghz with a single core it just bleeds energy like a bastard, so the dual core solution is the obvious next step, slow down the Ghz and increase the number of physical processors on-die to achieve parallel processing directly on CPU, makes sense doesnt it, its a very ellegant solution. when they hit the wall with the dual core solution all they have to do is just add more on-die physical processors, 3, 4, 5, 6 inf, when you put more physical processing units on-die it provides a ton of power, in car terms you get raw horsepower and you dont have to crank it to the moon to get the performance out of it.

Imagine a Dual core CPU with HT enabled, thats wicked sick and its comming sooner then we could possibly imagine...
Oops that was the long answer, o well...


<font color=blue> <Archer> Cant this thing go any faster, I thought this was a Warp 5 engine?
<Trip> Yeah, on paper... </font color=blue>
 

ytoledano

Distinguished
Jan 16, 2003
974
0
18,980
But we'd have to have programs supporting multi-processing. HT will get you nowhere when working with a single program which isn't optimized for it.

<b>Behold, Mine anger and My fury shall be poured out upon this place upon man and upon beast and upon the trees of the field and upon the fruit of the land and it shall burn and shall not be quenched
 

endyen

Splendid
Two words - leakage current. When you have two conductors with different charges, the energy tries to balance out. The prescotts have more pipes so more conductors (interconnects) and because the die is smaller, the insulation between is thinner. You get little short circuits, which are pure heat loses.
 

Xeon

Distinguished
Feb 21, 2004
1,304
0
19,280
Design flaw it has nothing to do with their 0.09u process anything longer. The Pentium M's suffer none of the 0.09u pitfalls they claim the Prescott suffers from.

Longer pipelines need more juice to push that data down the pipeline.

Xeon

<font color=orange>Scratch Here To Reveal Prize</font color=orange>
<font color=white>In the case of over scratching contact THG for property damage claims.</font color=white>
 

P4Man

Distinguished
Feb 6, 2004
2,305
0
19,780
For once I somewhat agree with Xeon. There are a number of reasons that work together to contribute to what you are seeing now..

1) Increased transistor count (core). For some reason, Prescott contains a LOT more core transistors than NW. There have been tons of speculation on the why's and hows, but the truth is, no one really knows what they do, or even if they are all used. One thing is for sure though, more transistors (especially core logic as opposed to cache) means higher leakage, and therefore more heat.

2) 90nm.. Not that I am among those that believe something is terribly wrong with intels 90nm, but regardless, Prescott is tiny and packs an incredible ammount of transistors, which combined gives you a very high thermal density. 100+W is "easily" cooled on a big chip, but >100W/cm² gets HOT. Prescott has a >50% higher thermal density as NW and more than double that of Opteron.. Even if power consumption where comparable, the smaller chip will still be hotter.

3) higher than expected Vcore. Vcore has been dropped compared to 130nm, but not nearly as much as typical process shrink allows (at least so far). Expect improvements here.

4) leakage.. as process nodes get smaller, leakage gets exponentially higher. It was neglectable at 180nm and previous nodes, noticeable at 130nm, and getting a real problem at 90nm. SOI or similar technologies can counter this though, which is probably why you will not see the same problem with AMD (at least not same magnitude), and I expect intel to cure it when they move to 65nm with their (FD?) SOI implementation.

5) last but not least: hyperpipelining. Intel dramatically increased the length of prescotts pipeline (+50%), which inevitably leads to higher power consumption. The longer pipeline should allowed prescott to scale to much higher frequencies (and overclocked with adequate cooling, it seems to live up to this), but if intel can not get power under control, we may never see the full potential of prescott.

Put these 5 factors together, and you end up with the hottest chip in history. Hopefully, the hottest we will ever see..

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

Xeon

Distinguished
Feb 21, 2004
1,304
0
19,280
Put these 5 factors together, and you end up with the hottest chip in history. Hopefully, the hottest we will ever see.
NSA uses much hotter ones check it out.

Xeon

<font color=red>Post created with being a dickhead in mind.</font color=red>
<font color=white>For all emotional and slanderous statements contact THG for all law suits.</font color=white>