Intel's Smithfield to consume 130W?

[DrMrLordX]

http://www.tomshardware.com/ha...s/20050117_225242.html

Judge for yourself. Doesn't this sort of defeat the point of dual-core CPUs? They seemed to overlook the fact that two cores should amount to more die space for thermal dissipation. However, with a die shrink, that might not be the case . . .

 

[StrangerGuy]

Intel: "Here we go again..."

 

[n7]

Hmm...i have to admit to being curious to see what AMD's dual core babies will suck in & put out comparably.

They might have similar issues for all we know.

 

[zakee00]

i seriously doubt it. amd will probably manage to do dualcore great, low power consumption and heat to boot.
intel is seriously loosin' it. im not a fanboi either, i just buy whatever is better...and amd is better right now. if intel cut their prices, increased performace in games, and decreased heat, and made their memory controler on die...then i would look at them.

 

[AnandThenMan]

If the thermal numbers turn out to be true, then Intel has indeed lost it. 130+ watts is totally unreasonable. Totally. There is no way to use a chip like that in a normal case, with a normal cooler, or a normal motherboard. This pretty much proves that Intel is essentially putting forth a dual core'ed Prescott. They need to go back to the drawing board.

 

[Zebo]

I got news for you.. presscot already consumes 150W dispite Intels tdp which is an average. AMD OTOH quotes max for the highest rated chip in iteration.

 

[clarkey01]

why they scarp Tejas again ? Power consumption and heat ?

Oh, nevermind.

 

[clarkey01]

Theres a dual core prescott ( no hyperthreading) which is called smithfield and theres a dual pentium M in the works too. But, I think this heat is the result of Intel rushing and making on to anyone dumb enough to listen that they saw dual core coming "for years" and are very ready.

Whats really happened.. They'v hit a speed wall and power wall, droped the "Ghz means everything", and then said " Yeah, we'll do 32-64 too, and NX disable, but we'r call them something else, and er (looks to see whats on AMD's stall) were doing Dual cores too.

Nehalem, Tejas were suppose to be out this year if you look back on the 2003 roadmaps . For a company that insits it was/is "Very Ready" for dual cores, hmm why would they do that ? (put tejas and nehalem up them scarp them near them tapping out) Must of cost a few million atleast.

What did pat gelsinger say " Its not a race, we'r perfecting out product much more".


Ah ok then.....

 

[stevty2889]

I don't think Intel will use an on-die memory controler, because they like to leave options open when it comes to choosing memory. Just think of how much less they would be able to sell if the LGA775 chips had on die memory controlers that forced you to use DDR2..after the whole Rambus thing, I doubt they would try that again. Anyway, having a dual core only 30w higher than a single core really isn't as bad as it could be. In any case, I think AMD will come out on top again. Dual core A64 will still probably run cooler and use less power than a single core prescott.

 

[DrMrLordX]

Originally posted by: Zebo
I got news for you.. presscot already consumes 150W dispite Intels tdp which is an average. AMD OTOH quotes max for the highest rated chip in iteration.

If that is the case, then Smithfield will probably wind up consuming more than 130W at full load.

 

[LTC8K6]

We need to know the size of the core. If the core is larger, then it could be easier to cool a 130W Smithfield than a 115W Prescott.

 

[LTC8K6]

presscot already consumes 150W dispite Intels tdp

So what? TDP is how much power the cooler needs to dissipate. I would expect TDP to be well below total power consumption.

The implication that Intel is lying is getting a bit silly.

 

[clarkey01]

"The implication that Intel is lying is getting a bit silly. "


Zebo was not trying to give the impression there lying, more of a fuzzy picture with these "averages" and AMD's gives the full whack. Intel would never give you the highest tdp, they'd laugh at it themselves.


As for die size its two prescott's put side by side, so double the current dize and a little extra ( lol, Im sure someone on here has got to know the die size of presshot). I just cant be arsed searching.

 

[clarkey01]

"Smithfield is little more than two 90nm Prescott cores built on the same die. There is a requirement for a very small amount of arbitration logic that will balance bus transactions between the two CPUs, but for the most part Smithfield is basically two Prescotts. "

http://www.anandtech.com/cpuch...ts/showdoc.aspx?i=2252

 

[LTC8K6]

Why would they give you a lower TDP number which would cause too small a cooler to be fitted?

 

[DrMrLordX]

Originally posted by: clarkey01
"Smithfield is little more than two 90nm Prescott cores built on the same die. There is a requirement for a very small amount of arbitration logic that will balance bus transactions between the two CPUs, but for the most part Smithfield is basically two Prescotts. "

http://www.anandtech.com/cpuch...ts/showdoc.aspx?i=2252

Interesting. The Tom's Hardware article I linked above seems to imply that Smithfield may be manufactured on the 65 nm process rather than 90nm.

 

[clarkey01]

Didnt Intel recently say they were delyaing 65 nm production ?

http://www.overclockers.com/tips00689/

" Sometime in 2006"


I also spoke to a guy who works in the D12 (Ireland fab) or fab(s) which are getting 65 nm test runs first, he named the fabs to me, another which I think is oreigon ( Cant spell american states)

"neil says:
fab 12c/fab24-2/D1D are all the 65nm fabs
neil says:
only D1D is up and running
neil says:
thats where im going on assignment
"

 

[Duvie]

D1D is part of the Ronler Acres complex in Hillsboro, Oregon....Basicall my backyard... I can almost see the exhaust from the air handlers from here.....

D1D was the latest fab to come on line and is part of the series of the "Developmental " fabs. D1c got the 90nm ball rolling with the Prescotts....Older and first D1B fab (completed and tooled in 1996-1997) was converted to a full fledge fab back a few years ago....D1a is at the Aloha, Oregon Plant ad was built in the early 1990's.

D1D is new but I beleive back in the mid of the year it hit a 20-25% tooled milestone..not 100% sure on that as it was a goal I heard one of the contractors mention....


lets hope they figured somehting out cause th eleakage will only be gerater with the smaller process if they have ot figure out SS or implemented SOI....

 

[clarkey01]

I wonder how well Intel would benifit from SOI and that new breakthrough IBM & AMD's recently mentioned


I wonder how cool a 3.8 Ghz P4 would run with SOI...


http://www.vnunet.com/news/1160036

"BM and AMD today claimed a "semiconductor manufacturing breakthrough" based on strained silicon transistor technology that improves processor performance and power efficiency.

According to the companies, the process results in a potential 24 per cent transistor speed increase at the same power levels, compared to similar transistors produced without the technology.

The new strained silicon process, called 'dual stress liner', enhances the performance of both n-channel and p-channel transistors by stretching silicon atoms in one transistor and compressing them in the other.

The technique works without the introduction of costly new production techniques, allowing for rapid integration into volume manufacturing using standard tools and materials.

IBM and AMD said that the jointly developed technology marks the first time that strained silicon has been made to work with silicon-on-insulator technology, resulting in performance and power saving benefits.

The production technique aims to solve the problem of transistors operating faster as they get smaller, but doing so at higher power and heat levels due to electrical leakage or inefficient switching.

Dirk Meyer, executive vice president at AMD's Computation Products Group, said: "Our shared progress in developing advanced silicon technologies allows AMD to deliver today's best performance per watt."

The chip maker intends gradually to integrate the strained silicon technology into all its 90nm processor platforms, including its future multi-core AMD 64 processors. The first 90nm AMD 64s using the technology are scheduled for the first half of 2005.

IBM plans to introduce the technology on multiple 90nm processor platforms, including its Power Architecture-based chips, with the first products slated to begin shipping in the first half of 2005.

"Innovation has surpassed scaling as the primary driver of semiconductor technology performance improvements," said Lisa Su, vice president of technology development and alliances at IBM Systems and Technology Group.

"This achievement with AMD demonstrates that companies willing to share their expertise and skills can find new ways to overcome roadblocks, and help lead the industry to the next generation of technology advancements."
"

 

[stevty2889]

Yeah, Smithfield is definatly 90nm. From what I can tell D1D should be mostly filled up now, and then 65nm will be going to Ireland and Arizona. D1D is practicly in my backyard too.

 

[DrMrLordX]

http://www.tomshardware.com/ha...s/20050120_182657.html

I guess we all saw this one coming, didn't we?

I didn't expect this kind of news out of Tom's, though . . .

 

[AnnoyedGrunt]

Originally posted by: LTC8K6
Why would they give you a lower TDP number which would cause too small a cooler to be fitted?

Well, Intel gives a TDP number that they believe will represent a loaded system under "normal" use. This means that their TDP numbers are not actually the max possible power the chip can pull (if you were to multiply the chip voltage by the maximum current it can take).

AMD on the other hand rates their TDP as the maximum power the chip can take (they basically multiply the voltage and current).

Intel probably uses their method because they don't want system integrators to put is over designed heatsinks when almost no one is going to require them. However, the fact that the latest Intel chips end up throttling performance in some cases definitely indicates that they have underspec'd their heatsinks a bit (IMO at least).

AMD once had a reputation for excessively hot chips, and it was shown (I think @ THG) that some of the AMD chips could actually catch on fire because the thermal protection did not act quickly enough. Perhaps AMD has changed their TDP rating methods in order to be certain their HSF's are large enough to easily handle the power output of their chips.

In the end, I'm not sure why Intel and AMD have chosen the different ways of specifying TDP, but I do know that the two methods are extensively documented and that AMD's method gives more conservative (but not necessarily more accurate) ratings than Intel's method. No one is lying, since they both state how they find their numbers, but at the same time the numbers are not readily comparible between the two companies, and they actually make AMD's power consumption look worse than it would be if the two companies used the same method (or Intel's numbers look better, depending on your point of view).

Anyhow, I'm thinking the power issue will only concern a small portion of the market, so it probably won't affect Intel too much. With that said, I guess it is somewhat ironic that one of the big reasons I am looking towards an A64 is because of the lower power consumption compared to a P4.
Later,
D'oh!