Ouch 103 Watt heat from the Prescott 3.6GHz CPU

[Wingznut]

Originally posted by: THUGSROOK

Originally posted by: Wingznut
First off... Let me go on record saying that I'm not going to confirm if this rumor is true or not, and that I'm not going to speak specifically about future products. But for the sake of this post, let's presume that it is true, and that the TDP is ~100w....

So what? Do you think that if PSC is ~13w more than the current Northwood, that is going to change ANYTHING? We're talking 13w here. How many watts do you suppose your overclocked/overvolted 3.5ghz P4's are dissapating?

And when people talk about overclocking, do you realize that heat dissapation is not the roadblock? It's the gate leakage and elecromigration... And that's only going to become more of an issue (exponentially) with every process shrink.

Are we done overreacting to a rumor now?

thx for the great posts in this thread

the reason we are concerned?
cause we see whats happening with northwoods @ 3400+mhz ~ heat and power consumption galore.
if its gonna get worse with prescott, we wont be able to OC them (much).

:disgust:

But that's my point, Thugs.... Whatever the limit of your overclock might be (good or bad), it won't be because of the extra 13w of TDP.

(Again, I'm just presuming that the rumor is true, for the sake of this thread.)

 

[THUGSROOK]

Originally posted by: Wingznut
But that's my point, Thugs.... Whatever the limit of your overclock might be (good or bad), it won't be because of the extra 13w of TDP.
[/i]

understood ~ we already have this problem
hehe~

 

[DAPUNISHER]

As Intel will probably wait till the last minute to point out though, that MHz is not everything. AMD already points this out, albeit sometimes less warranted in their "PR" schemes. As well as Apple showing that clock rate does not equal performance.

Nonsense. Intel will not make any such comment concerning MHz, and while you are pointing out how AMD and Apple have strong IPC proving clockspeed doesn't equal performance you fail to mention the Pentium-M which has very strong IPC, why?

 

[batmang]

so now the question is, which one will be cheaper / overclockable....
hell, if the athlon 64's run the same price, but can oc a bit....who knows. i still think prescott is gonna own athlon 64.

 

[Wingznut]

Originally posted by: DAPUNISHER

As Intel will probably wait till the last minute to point out though, that MHz is not everything. AMD already points this out, albeit sometimes less warranted in their "PR" schemes. As well as Apple showing that clock rate does not equal performance.

Nonsense. Intel will not make any such comment concerning MHz, and while you are pointing out how AMD and Apple have strong IPC proving clockspeed doesn't equal performance you fail to mention the Pentium-M which has very strong IPC, why?

Stronger than AMD's and Apple's, actually.

Btw, whatever happened to all the people who were saying that the Pentium-M would have to use some sort of PR rating or that it would fail.

 

[Jeff7181]

People seem to be stuck on heat being an issue. I just saw a thread where someone suggested that overclock an XP1700 to 2.7-2.8 Ghz would be possible with water cooling. I suspect they deduced this from the fact that AthlonXP's are getting close to 3.0 Ghz on prometia setups and liquid nitrogen. Then don't seem to understand it's not really an issue with the heat being generated, it's more of an issue with the physical abilities of the CPU. Transistors can only turn on and off so fast. And with the AthlonXP's relatively short pipeline, that limits the maximum clock speed as well. Wingznut can probably explain why better than I can. But the fact is, it's not the heat that's the problem... both Intel and AMD are at the limit of what their processors are physically capable. That's why AMD is switching to SOI, and Intel is switching to Strained Silicon, and will be switching to SOI in the future also.

Within 5 years Apollo Diamond expects to be able to manufacture 4 inch diamond wafers that are able to be used instead of silicon wafers. I see the switch being made, maybe around a .04 micron or smaller process. Heat would not be an issue because diamond CPU's could withstand 10-20 times as much heat as silicon. And now that Apollo Diamond can grow synthetic diamonds more perfect than natural diamons for $5 per carat, cost shouldn't be an issue. So I'd say sometime 5-10 years from now, the switch will be made from silicon to synthetic diamond.

 

[DAPUNISHER]

Btw, whatever happened to all the people who were saying that the Pentium-M would have to use some sort of PR rating or that it would fail.

Probably the same goofballs that were screaming how NV30 would crush the 9700pro months before it appeared Hard to talk with both feet stuck in your mouth.

 

[Wingznut]

Jeff7181... I wouldn't go investing in the diamond wafer market any time soon. Right now, since cpu's are being manufactured on 12" wafers, you can get hundreds of die from one wafer. With 4" wafers, you'd only get a few from each. The manufacturing costs would be phenomenal.

 

[Jeff7181]

Originally posted by: Wingznut
Jeff7181... I wouldn't go investing in the diamond wafer market any time soon. Right now, since cpu's are being manufactured on 12" wafers, you can get hundreds of die from one wafer. With 4" wafers, you'd only get a few from each. The manufacturing costs would be phenomenal.

Consider it though if they were using a .03 or .04 micron process... they'd be getting 3-4 times as many cores per wafer.

*EDIT* Assuming transistor count didn't dramatically increase with each step down in transistor size.

 

[Wingznut]

Originally posted by: Jeff7181
*EDIT* Assuming transistor count didn't dramatically increase with each step down in transistor size.

Ah... But it will.

 

[Def]

It will take a VERY long time for any switch to be made from silicon to any other materials(who knows what it will actually be?). I'd be surprised if it happened sooner than 15-20 yrs from now, very surprised.

 

[OverVolt]

I wanna see Dell Gateway HP ect ect try and make the prescott run silent !

 

[zShowtimez]

I wanna see Dell Gateway HP ect ect try and make the prescott run silent !

Maybe they should talk to Zalman.

 

[Wingznut]

Originally posted by: OverVolt
I wanna see Dell Gateway HP ect ect try and make the prescott run silent !

I bet people have been saying the same thing, year after year.

I don't know... I just don't see a rumored 13w being that big of a deal. *shrugs*

 

[Jeff7181]

Originally posted by: Wingznut

Originally posted by: OverVolt
I wanna see Dell Gateway HP ect ect try and make the prescott run silent !

I bet people have been saying the same thing, year after year.

I don't know... I just don't see a rumored 13w being that big of a deal. *shrugs*

Especially since the 3.2 Ghz P4 runs perfectly within spec with stock cooling...

One thing no one has mentioned... will the Prescott's core size increase? If so, that will help with heat dissipation.

Another thing I'm curious about is whether increased heat might spark some companies to redesign cases so that air flows up and out, instead of back and out. But I'd have to wonder if that's a good idea since Apple is doing the climate zones instead of up and out... and normally they stick with their own design even if it sucks... like the new cheese grater G5.

Who the hell designed that?!?

 

[RaynorWolfcastle]

Originally posted by: Wingznut

Originally posted by: OverVolt
I wanna see Dell Gateway HP ect ect try and make the prescott run silent !

I bet people have been saying the same thing, year after year.

I don't know... I just don't see a rumored 13w being that big of a deal. *shrugs*

well if the rumour is true (and we'll assume it is for speculation's sake), then the 13W isn't the big deal. The big deal is about what will happen next; that is if the 3.6 GHz has a TDP of 103W, then the 3.8 GHz will be higher (near 110W assuming a linear progression), the 4.0 GHz will be higher, etc.

Traditionally, a die shrink has meant less power required at the originally released frequency which, at least in part, allows for scaling the CPU speed on the shrunk process. For example, a Willamette P4 2 GHz had a TDP of 75W while a Northwood 2.2 GHz 57W TDP. In other words, a 10% clock increase (and a 50% boost in L2 cache size) resulted in a 31% decrease in power dissipated. Only with the 2.8C does the TDP of the (HT-enabled) Northwood core approach a 2 GHz Willamette's TDP.

By comparison a 500 MHz Katmai-based (.25u) P3 had a TDP of 28W and it took until the Coppermine (.18u) core hit 1 GHz for the TDP to get back to that level.

The worrysome part is that the power density in the Prescott core will be very high and will only get worse as clock speeds scale. As long as Intel provides adequate cooling guidelines this is not a problem so much as it is a concern.

I really hope that either this rumour turns out to be false or that Intel is able to lower the TDP for the final product; keeping my room nice and cool is getting more and more difficult

 

[SinfulWeeper]

Originally posted by: THUGSROOK
looks like ill be skipping another Intel platform.

WTF!!!:Q:Q:Q

Thugs, when it comes out, I highly recommend you get one. Then download some Jenna Haze and see what kind of encoding times you can get with her in some DivX compression.

I think you'll be much happier