heat generation increasing (in response to 100W prescott)

[Stunt]

hey.
with this talk of prescott being 100+ W and tejas being well into 100W, i am wondering if this is going to be a continuing trend.
a few years ago there was a real move towards less heat. (c3, transmetta chips)
now we see a leader in the industry going in the wrong direction.
i for one am into quieting my computer, and i need chips that run cool.
also servers which run in racks need to run cool too.
i am sad to see the new chips with massive cooling needs.
already intel has had to create a separate line of notebook cpus because of the inappropriateness of the p4.
does this mean in the future intel will need a low heat generating xeon/notebook line in addition to their itaniun and desktop line?
or are they going to offload all cooling to heatsink manufacturers?
i think that this is getting to be rediculous (p4 3gig, disipates around 100W)
now they have shrunk process and its over 100W!
when will this beging to hurt intel, and is this a trend that will continue?

 

[ST4RCUTTER]

Yeah, you can pretty much bank on this being a continuing trend. The physics that come into play will not change, only the materials and the methods used to counter them. Reducing the process to 90nm will help Intel and others to reduce the temperatures, as will the implementation of low-k dielectrics. It would be nice to see how SOI impacts the heat generation/dissipation of the Opteron. With those numbers one might be able to say it was worth the huge investment AMD appears to have made into that process technology. It will likely be that the Opteron/AMD64 is a cooler running chip than the Prescott.


:sun:

 

[pm]

It's a fact of CMOS manufacturing that is only going to continue to get worse.

In the past power has increased because the industry crammed more and more transistors into a smaller and smaller space. The smaller transistors burned less power as they were shrunk but designers kept adding more of them. So power gradually started to creep up. But more recently power has been increasing at an even faster rate because transistors have gotten so small that they are acting less like switches and more like leaky faucets; essentially they leak power. With each successive generation going forward (0.09um, 0.06um, etc.) it is going to get worse. A specific type of leakage called "gate leakage" increases by 10x with each successive generation. Another type called "subthreshold leakage" increases more slowly with each process shift, but it has a head start over gate leakage and so it's the bigger problem currently (pun unintended).

At 0.25um, leakage wasn't really an issue. At 0.18um, it was a minor annoyance. At 0.13um, it is a serious problem. Going forward at 0.09um, it becomes a dominant contributor to the total chip power. And it will get worse.

Design tricks can reduce it - usually with increased design complexity and usually with performance reductions. Manufacturing changes - such as using different materials like high-K gate dielectrics - can reduce it. But as transistors get smaller and smaller, they will leak more and more with each generation and there will be more effort required merely to hold power at a constant value (say, 80W as an example).

This is not a "the sky is falling" post. Moores Law is not over. But it is a fact that we are going to have to learn to live with higher and higher power processors. For what it's worth, in the global scheme of things, 100W isn't that bad. Most of my house has two-light sconces and they burn two 60W bulbs (120W total). Look at the power ratings on major applicances and you'll see that 100W isn't that bad.

Patrick Gelsinger, Intel's Chief Technology Officer, gave a great keynote speech at ISSCC (the International Solid State Circuits Conference - the largest annual chip design conference) two years ago. There's a summary here and here. It makes for interesting reading. And having attended it, I now start to feel a bit like people must have who heard Gordan Moore speak decades ago about Moores Law and then saw it happen around them.

For what's it's worth as well, having worked on a 100W+ microprocessor (the Itanium 2), I can speak with experience in saying that the problem with a 100W microprocessor isn't cooling it. It's getting power to it. Changing the numbers slightly to keep the math simple, 100W at 1V is 100 amps. And that requires serious power delivery engineering in order to deliver 100A to a CPU that is constantly changing it's current loading while holding the voltage fixed at 1V. No one is particularly worried about cooling at this stage... it's more how expensive the power regulation systems are going to get going foward.


On a slightly related note, I was looking at a 386 motherboard recently and I noticed that the 386 (or at least the version that I was looking at) didn't even require a heatsink. Looking at the bare ceramic CPU, it struck me how much things have changed.

 

[McCarthy]

Guess I'd disagree about 100w not being a big deal as it's 100w for one part of the overall system. Other devices like radios, tvs, stereos wouldn't be a good comparison as it's total vs component. I'm touchy about heat though as I sit here listening to my old AC straining to keep up and I think "I'm pumping out how many of watts to type text?" (entire system)

Which is somewhat solved at present. Though my AMD seems to like to pump out heat like a hand dryer even when cpu use is at 2%, the P4 2ghz I gave my mother never got the heatsink warm in general use. I probably turned off something in W2K power management I shouldn't have, point being that 100w peak doesn't mean 100w all the time. In the typical computer role, sure 100w bursts now and then are managable for the room environment.

In a laptop role, better be real short bursts. That's where I'm really interested in how things will turn out. Heard the other day that laptop sales exceeded desktop sales for the first time - as desktop replacements. (not sure if it was year or quarter) About the same time DVD rentals finally surpassed VHS.

Still got my 386sx-16 here with original components. Cool and quiet, can't say quick anymore though 🙁

 

[Barnaby W. Füi]

Originally posted by: pm
On a slightly related note, I was looking at a 386 motherboard recently and I noticed that the 386 (or at least the version that I was looking at) didn't even require a heatsink. Looking at the bare ceramic CPU, it struck me how much things have changed.

Yeah, the 386 I have is soldered to the mobo, and is just in black plastic packaging like you'd see on a chip that belongs on a sound card or something 😛

Originally posted by: McCarthy

I'm touchy about heat though as I sit here listening to my old AC straining to keep up and I think "I'm pumping out how many of watts to type text?" (entire system)

Well honestly, if ALL you're doing is typing text, you can very comfortably get away with one of those low power via's, or a tualatin, or an older apple, etc. The reason we put up with these power requirements is that we can do so much with computers.

the P4 2ghz I gave my mother never got the heatsink warm in general use.

But that doesn't mean a whole lot. You could have a cpu putting out 80w of heat with a heatsink that's cool to the touch, or you could have a 40w cpu with a heatsink that almost burns your hand. Heat and temperature are only somewhat linked.

 

[McCarthy]

Yes, only typing text at the moment. Not going to build a new computer to come write a post. Seems pretty obvious.

Later tonight I'll probably game for a bit at which point it will be worked. I think I'm fairly average for a home user who uses the full capacity of the system anywhere from 0~10% of a day. When the idling features work well that makes for a very nice combination. As it will be with a 100w peak output Tejas for most people.

The P4 computer, it was a Dell with the nifty green scoop-duct. Flipped it out of the way so the sink was getting no airflow, still sat there dead cold doing normal 'browse/email/quicken' stuff. Really the entire system is overkill for the need, but the price was right. CPU idling works nicely there, as did good old "Rain" with my P150 with it's worn out fan I unplugged except when I needed sustained CPU work. (One of those factory intel sinks, couldn't figure out how to replace the fan and wasn't going to warranty something that was nearing the end of it's life)

 

[draggoon01]

how much heat does opteron dissipate? athlon64 about same?

 

[zephyrprime]

... my AMD seems to like to pump out heat like a hand dryer even when cpu use is at 2%, the P4 2ghz I gave my mother never got the heatsink warm in general use.

It's because of the way AMD cpu's handle the halt instruction. It relies on the chipset or something to actually go into idle state rather than being able to do it of its own volition like Intel CPUs. I honestly don't know why AMD ever thought to make it weird like that.

 

[Budman]

Originally posted by: zephyrprime

It's because of the way AMD cpu's handle the halt instruction. It relies on the chipset or something to actually go into idle state rather than being able to do it of its own volition like Intel CPUs. I honestly don't know why AMD ever thought to make it weird like that.

That's exactly why I say Intel cpus run cooler,Not because the are cooler than AMD at full load but because when an Intel cpu idles it really idles & runs very cool.

 

[ViRGE]

Pm, while it's true that power delivery is an enormous challenge, I can't help but feel that your underestimating the cooling costs of a computer. CPUs of course are hitting the 100wt mark, and GPUs are starting to work their way up their too; this isn't to discard northbridges either, which are at the giant heatsink/fan stage. All of this is adding up to a lot of heat, with all of these devices being near-perfect electricity-heat converters, with totals reaching 200, 250, or even 300 watts of electricy being converted to heat. All of this can be drawn away from the device easily enough through ever-larger and complex heatsinks, heatpipes, and fans, but all this heat energy has to go somewhere, and unlike the power-hungy applicance, all of this heat is typically dumped in to a smaller, poorly ventelated room, resulting in an abnormally hot and uncomfortable room.

You may be worried about routing power to the Itanium 3, but I for one am starting to get worried about routing air conditioning to my den to keep it managable. At this rate, are home "server rooms" the wave of the future?

 

[pm]

When I say that cooling is not a problem, I am specifically speaking to cooling a CPU. As far as cooling an entire system in which the parts inside are quite a bit hotter as well, I can agree that this will be a challenge, but it certainly is easier (and cheaper) than trying to design a 100A 1V power delivery system. If anyone thinks that cooling 100W is harder than powering 100W@~1V, I need to introduce them to my world. 🙂

The room that I'm in right now as I type this is in the SE corner of the house. It's 100F outside, and I have my 3GHz Pentium 4 box and a 21" CRT monitor in here. It's about 88F in here, and I long ago decided that it's not particularly comfortable. If I could reduce the dissipated thermal power of my system without compromising screen performance (or size) or system performance and for very little money, I think that I would look into it. But it's not uncomfortable enough in here that I would, say, get rid of my Ti200. Or replace my Pentium 4 with something cooler (but slower) like a 0.13um Pentium 3.

I agree that I'm not looking forward to this future that I'm presenting, but I have had a lot more time to get used to the idea than you guys. I'm resigned to the fact that computers are going to get a lot hotter in the future. There's very little that can be done to stop it until the day when consumers are more interested in power than performance... something that happened a long time ago in the notebook segment, but something that isn't likely to happen in the near future in the desktop processor market.

 

[Lonyo]

Originally posted by: Stunt
hey.
with this talk of prescott being 100+ W and tejas being well into 100W, i am wondering if this is going to be a continuing trend.
a few years ago there was a real move towards less heat. (c3, transmetta chips)
now we see a leader in the industry going in the wrong direction.
i for one am into quieting my computer, and i need chips that run cool.
also servers which run in racks need to run cool too.
i am sad to see the new chips with massive cooling needs.
already intel has had to create a separate line of notebook cpus because of the inappropriateness of the p4.
does this mean in the future intel will need a low heat generating xeon/notebook line in addition to their itaniun and desktop line?
when will this beging to hurt intel, and is this a trend that will continue?

If I remember correctly, Intel are set to be introducing some new server chips using the Banias core (the one their mobile chips use) meaning they will run cooler and require less power.

I think the Banias chips may be constructed on 0.09 process, which could also reduce power consumption (but might not). They do seem to know there is a need for cooler products.

Also, with graphics cards, while the top end products may be producing more and more heat, ATi have their mid-ish range 9600 which runs fanless on their 0.13 micron process. It's not all bad.

 

[Dresdenboy]

There are some cooling tools (for example here and here) which make use of the HLT command for AMD CPUs.

What I remember is: The problem is and was the kind of bus which AMD uses (the Alpha EV6 bus AFAIR) which needs to be connected if the chipset (more exactly the northbridge) doesn't support a CPU HLT (which needs special support by the chipset). CPU Cool and other tools activate this feature if a chipset FSB disconnect and reconnect is OK.

With these tools one gets the same behaviour like with a P4. I don't need it since Prime95 is running all the time 😉

Regards