Quote:
|
Originally Posted by jaydee116
One thing your over looking is a peice of junk OEM heat sink will handle 80watts with no problem and keep the computer operation without fault. Remember the majority of users don't care about over clocking and what not. They just want something that works. If a standard OEM heatsink can handle todays computers then our current after market heat sinks can easily handle tomorrow's. Like I said 200watts should be easily cooled by air. It will be sometime, if ever, CPU's will get that warm. IMO anyway.
|
Piece of junk OEM heatsinks do dissipate that much heat. I personally don't think watercooling will become a OEM type mainstream solution necessarily because of thermal stress, at least at first. Average user doesn't care about heat or overclocking, but ten fans in the 'puter is going to irritate the housewife running Create A Card! I think the idea of "quiet" is going to become an issue before "heat" does to the soccer-mom and her Dell XPS junk.
But, you are especially right about one thing, namely HSF tech is still pretty primitive. I have always been impressed by the Shuttle Micros and their heatpipe trick. The new BTX form-factor is going to facillitate that kind of heatsink, namely taking the heat of the sink and moving it to a big radiator with a slow moving 120mm fan.
But the other issue is power output
density per unit area. Think about it, a P-IV Presoctt, according to Intel, is designed for ~100w output typical heat. The heatspreader on the chip is approx. 900mm squared. A square meter is 1000,000 mm square. So the equivilant heat output from a square meter at the same power denisty is like a hundred kilowatts! That's almost a hundred and thirty horsepower in equivilant work. We're starting to approach the same thermal "density" as heat engine with that number. A P-IV Prescott is on .09nm process. The next gen will probably be on .065 process, which reduces the die area on the chip to about 30 percent the size of the .09nm chip, assuming identical transistor count. So even if power output is halved on that hypothetical chip, the thermal density of the die is going to increase somewhat, ~40%.
This is what's been happening all through the last decade. Eventually the thermal conductivity of even silver is going become a barrier, it just can't suck enough heat off such a small area. That is why I think boron doped diamond is going to be the next big step if air-cooled chips are to continue. The thermal conductivity of the stuff is around 8 times that of silver (it is the best heat conductor known) and can take temps of more than 2000C. Or of course, a different kind of fluid instead of air cooling the silicon, like water .Just my opinion I guess. Great discussion, though!