CPU Heat Limiations?

[GreatestOne]

I was watching a special on nanotechnology, and made me curious in that are these silicone based chips not able to go higher than ~4 Ghz on modern cooling techniques? I.e., unless a drastic change is made, like nanotech or biotech, is it impossible to get to 6 or 7 Ghz mostly bec of the heat output?

We do see an increase in Cores etc etc... but I wonder when we will see the 6,7, 10 Ghz chips... I am not up on future tech like this so any insight or current research is appreciated. Thanks.

 

[westrock2000]

Leakage is probably more at play in regards to clock speeds.

Heres an excerpt from an (old) Arstechnica article

Transistor leakage is a complex phenomenon, and decent explanation of it would make for a large article in itself. With that in mind, the following summary treatment represents the best stab at it that I plan to make for the foreseeable future.

As most Arstechnica readers already know, transistors are just tiny electrical switches which are always in one of two states: on and off. At least, that's how it's supposed to work. The reality is that even when a transistor is off, it's never really off. Rather, a small bit of current is still flowing through the switched off transistor, kind of like water trickling through a leaky faucet.

Now, one or two leaky faucets in a house is annoying, and if the leaks are relatively small then the impact on the monthly water bill is likely to be negligible. But imagine a whole apartment building full of leaky faucets; if every faucet in every apartment leaks the same amount, then the more apartments in the building the more total water ? and money ? will go down the drain each month, without doing any useful work.

As transistors get smaller, CPU architects can fit more of them onto a chip, and since each transistor leaks a certain amount then the overall leakage current flowing through the chip increases with the total number of transistors.

If leakage current only increased linearly with the number of transistors on a chip, then it wouldn't be as much of a problem. But there are other conditions that can cause the leakage current of each individual transistor in a chip to increase. First, as individual transistors get smaller they also tend to leak more, though semiconductor makers have a variety of tricks for counteracting this tendency. Much worse from Prescott's point of view, though, is that as the CPU die gets hotter the leakage current increases, and we all know what causes a CPU die to heat up: higher clock speeds, which are accompanied by higher power consumption and higher voltages. All of this adds up to big trouble for a processor architecture whose primary approach to performance lies in simultaneously increasing clock speeds and shrinking transistor sizes.

So heat just by itself can be handled. But changing the way the processor fundamentaly operates is proving to be harder for engineers.

 

[myounage]

There's some promise of making dies out of artificial diamond, which is currently on par with silicon for price ... and much more conductive to heat than copper, aluminum, silver, etc. The barrier before was that no one had made it into a semiconductor successfully. Scientists recently figured out how to convert it via boron doping. Anyway, there was a rumor that intel got a prototype all-diamond(Apollo made) CPU to over 80GHZ.

 

[GreatestOne]

Sweet guys... 80ghz, nice... that should be good for some real VR and interactive holograms right...

So the whole nano/bio tech chips are out, a farce?

 

[myounage]

For now I think. They may resurface .. but diamond seems like a more realistic successor to silicon at this point due to cost effectiveness and the fact that its a less revolutionary step. The heat transfer difference is the big'un though. Your die would be more conductive than your heat sink!