Future Intel Power Numbers

[TraumaRN]

Well looks like HP accidently released some future power consumption numbers on Intel processors...I dont know if this was posted or known already but...

The quite interesting bits begin on P19 though, where you find Potomac, the 8MB cache Prescott with a 115W TDP. Not bad at all considering where Prescott is with 25% of the cache. Tulsa, the dual core variant comes in at 175W, and you can almost hear AMD exec giggle when you mention it in print. Dempsey, the DP version with less cache comes in at at a (gack) much more reasonable 150W.

But wait, it gets better. Woodcrest, the 4M Merom based Xeon has a 70W TDP, which is a little higher that the 60 or 65W I have been hearing for Conroe, but this could simply be the conservative calculation that Mr Blade was talking about earlier. The best part it that Whitefield, the four core, 16MB cache CSI bearing Xeon is listed as 100-130W TDP. It is also listed in '06, which falls into the 'not a chance in hell' category. It may be talking about the unnamed four core Woodcrest variant, but until Intel decides if is going to do it with MCM or JAB, there is little use in talking about features.

They dont sound like Intel lovers but 175 watts..holy smokers....

Linkage

Linkage to HP pdf

 

[ryanv12]

Is that right? At this rate if I ever get one of these buggers, I'll just apply some AS5 to my frying pan and make grilled cheese sandwhiches.

 

[TraumaRN]

I'm assuming it's legit....I mean goto the PDF, it's directly from HP and dated June 7 2005...*shrugs* It is kinda crazy I couldnt imagine a power bill on one of those babies

 

[imported_michaelpatrick33]

They have a 70W mobile Intel part listed in their PDF. I saw on the Intel roadmap 15W dualcore mobile chips so something is definitely off. I also don't see have moving down to the 65nm process is raising their power up so much unless they are messing with the pipeline and trying to increase frequencies again. Of course the dualcore Prescott's now reach 140W under load at the 90nm level so trying to get to 3800 even at 65nm on the Prescott may cause those wattages.

 

[Sentential]

I agree, its really bad, but these are Xeon MPs. For a server it isnt AS big of a deal as it would be for people like us.

 

[TraumaRN]

Now I'm not server admin but when you pack those babies in a bunch of blades wouldnt you rather have them not cranking heat like that?

 

[Continuity27]

Originally posted by: DeathBUA
Now I'm not server admin but when you pack those babies in a bunch of blades wouldnt you rather have them not cranking heat like that?

Oh of course you'd rather them not, but have you seen the cooling systems employed on blade servers? :Q

 

[imported_michaelpatrick33]

When you compare the existing 90nm dualcore Opteron numbers to these (if accurate) it is downright frightening. AMD is completely dominant in power and heat. If accurate. Intel, I thought was coming out with a more Pentium M like architecture to replace the PIV. Let's hope it is sooner rather than later with these power numbers.

 

[Continuity27]

Originally posted by: michaelpatrick33
They have a 70W mobile Intel part listed in their PDF. I saw on the Intel roadmap 15W dualcore mobile chips so something is definitely off. I also don't see have moving down to the 65nm process is raising their power up so much unless they are messing with the pipeline and trying to increase frequencies again. Of course the dualcore Prescott's now reach 140W under load at the 90nm level so trying to get to 3800 even at 65nm on the Prescott may cause those wattages.

It's simply not the best architecture. We run into very harsh barriers at extremely high frequencies, Intel lowered their process to 65nm from 90nm and the problems didn't get much better because they stick with the same architecture that needs high frequencies. It will continually get worse and worse for them unless they switch to another architecture (kill Netburst), or keep perfecting their process. If they do the latter, the competition will pull even further away from them very quickly. They should redistribute their manpower to work on a desktop Pentum M variant instead of trying and failing at keeping Netburst alive and viable.

 

[TraumaRN]

I mean not to smack on Intel but why would they even continue with Netburst, would AMD really get that far ahead of Intel if Intel went ahead and tried to create something like Pentium-M like arch for the desktop???

And as another thought so right now the dualcore Opteron spank the Xeon's in power and heat, but is AMD really cranking out *that* many DC's Opteron's as to steal more of Intel's marketshare in the server side of things??

 

[Duvie]

Originally posted by: michaelpatrick33
They have a 70W mobile Intel part listed in their PDF. I saw on the Intel roadmap 15W dualcore mobile chips so something is definitely off. I also don't see have moving down to the 65nm process is raising their power up so much unless they are messing with the pipeline and trying to increase frequencies again. Of course the dualcore Prescott's now reach 140W under load at the 90nm level so trying to get to 3800 even at 65nm on the Prescott may cause those wattages.

How do you figure that??? look at that link...looks like that is an idle load and full load is muchmore...That is the type of spin Intel likes to place on thier numbers...usually figure there numbers as low end, like their current dual core chips....I have no respect for a company who spins (deceit) data like they do...

http://www.techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=15

 

[clarkey01]

Originally posted by: Sentential
I agree, its really bad, but these are Xeon MPs. For a server it isnt AS big of a deal as it would be for people like us.

For higher temps in a server I think it is important.

 

[imported_michaelpatrick33]

Originally posted by: Duvie

Originally posted by: michaelpatrick33
They have a 70W mobile Intel part listed in their PDF. I saw on the Intel roadmap 15W dualcore mobile chips so something is definitely off. I also don't see have moving down to the 65nm process is raising their power up so much unless they are messing with the pipeline and trying to increase frequencies again. Of course the dualcore Prescott's now reach 140W under load at the 90nm level so trying to get to 3800 even at 65nm on the Prescott may cause those wattages.

How do you figure that??? look at that link...looks like that is an idle load and full load is muchmore...That is the type of spin Intel likes to place on thier numbers...usually figure there numbers as low end, like their current dual core chips....I have no respect for a company who spins (deceit) data like they do...

http://www.techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=15

Lost Circuits shows Prescott power load for the cpu alone using Prime95

These numbers are measured cpu power numbers, not the system load.

 

[Duvie]

Originally posted by: michaelpatrick33

Originally posted by: Duvie

Originally posted by: michaelpatrick33
They have a 70W mobile Intel part listed in their PDF. I saw on the Intel roadmap 15W dualcore mobile chips so something is definitely off. I also don't see have moving down to the 65nm process is raising their power up so much unless they are messing with the pipeline and trying to increase frequencies again. Of course the dualcore Prescott's now reach 140W under load at the 90nm level so trying to get to 3800 even at 65nm on the Prescott may cause those wattages.

How do you figure that??? look at that link...looks like that is an idle load and full load is muchmore...That is the type of spin Intel likes to place on thier numbers...usually figure there numbers as low end, like their current dual core chips....I have no respect for a company who spins (deceit) data like they do...

http://www.techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=15

Lost Circuits shows Prescott power load for the cpu alone using Prime95

These numbers are measured cpu power numbers, not the system load.

gotcha!!!

<The systems(Intel dual cores) based on these chips suck up over 100W more than the systems based on the competing X2 processors.>

Interesting how Lost Circuits doesn't have the same 100 watt differnec unless the intel systems also draw more...I would expect the difference to be similar with similar system setups configured in the test....

 

[Viditor]

I posted that link yesterday. The pdf is flawed, so you need to take it with a bunch of salt.

1. It lists the mobile part as Woodcrest, but that is a Xeon part.
2. It lists Whitefield as a 32bit part, when that is the Xeon/Itanium part coming out in 2007

 

[Concillian]

Originally posted by: Continuity27

Oh of course you'd rather them not, but have you seen the cooling systems employed on blade servers? :Q

No, I can't get close enough to see them without fear of permanent hearing loss.

 

[composer777]

Yes, and those cooling systems are not free or cheap. For every watt you spend on processing power you have spend at least another watt on cooling. We've reached a point at my workplace where they literally cannot expand due to power requirements, at least not without building another server room and running another line from the power company. We've maxed out the power and cooling, and we have a raised floor in a very large building for our computing center. I work in bioinformatics, we have over 1,000 processors working at once (I think close to 2,000, but I'd have to check), and meeting the power requirements for all the computing we do is not trivial. Also, if you are colocating your server in someone else's rack, there's a high chance that you will have to purchase more space in order to cool the chip effectively. More space and more power consumption translates into higher costs.

Originally posted by: Continuity27

Originally posted by: DeathBUA
Now I'm not server admin but when you pack those babies in a bunch of blades wouldnt you rather have them not cranking heat like that?

Oh of course you'd rather them not, but have you seen the cooling systems employed on blade servers? :Q