I started thinking about Heat, apply a resistive load and you'll get heat somewhere right no matter how small, take a CPU, get's pretty hot, is it possible to ever get to the stage where a CPU could generate no heat?

 

Probably not. The reason CPUs as a class get so hot is they have millions, possibly billions, of transistors turning off/on/off/on.

The most efficient states for a solid state switch (a transistor) is on or off. In either of those states no heat is created. But when the switch is in between, neither on or off, there you have heat. Since CPUs also switch large numbers of transistors at billions of times per second it adds up, fast. The average heat per gate is less than a picowatt, but it is the numbers that tell the tale.

 

I think maybe in future if CPU's are ever able to use light as a switching device as opposed to the transistor circuits currently in use, then perhaps heat would be less of an issue due to the lower power consumptions, however I'm sure there will still be heat produced in the other required processes.

 

I suspect using photons for computation is a long ways away, quantum is probably closer, and both will be cooler.

The early 8 bit CPUs didn't heat up a bit, but that was because they had a few thousand transistors. The transistors keep getting smaller and more efficient, but the number also keep going up.

 

Agreed Bill. I think we're still far from the day heatsinks will be made redundant

 

The most efficient states for a solid state switch (a transistor) is on or off. In either of those states no heat is created. But when the switch is in between, neither on or off, there you have heat.

Is that so? Don't you still have some resistance (especially Rds) that causes heat? Or do you simply mean it is so small it is neglible?

 

Is that so? Don't you still have some resistance (especially Rds) that causes heat? Or do you simply mean it is so small it is neglible?

It is so small it is negligible. Basically the only current going thru Rds is the leakage of the next stage gates, which is... negligible.

There will always be a power loss, that's just the first law of thermodynamics. Technology may make is smaller but it cannot become zero.

 

A similar question came up on a forum dealing with superconductors. There the on resistance is truly zero ohms, and the off resistance is so high as to be a true open. But our old friend switching time is still to create some heat, because while it can be fast, the transition between off/on or on/off is not going to be zero.

 

In a modern processor about 50% of the waste heat is a product of leakage current; the current that flows through transistors when in the OFF state. The other 50% is switching losses. Leakage losses can be reduced by turning off sections of the CPU when not in use. For example the FPU and MMX/SSE units don't need to be used most of the time so they are in a low power mode.