My Circuit Theory I professor told us that as heat increases, conductors raise in resistance. Which of course is true. He also said however, that semiconductors raise in resistance as they get colder. Now at first I thought maybe he slipped, because he does that sometimes. But he used that example in other lectures as well...If this is true, how is it that supercooled microprocessors (phase change for example) run so much faster than with air cooling? Is this guy an idiot or is there something I am missing here?
He's right for bipolar transistors, that's why thermal runaway is a danger in some power amps. They get hot, resistance drops, more current flows so they get even hotter... With cpus, the problem is getting rid of the heat, so supercooling them lets you run them faster (making more heat). The middle of the chip is still pretty damn warm even if the outside is in liquid nitrogen.
Because the thermal diode is often located on outlying regions of the chip. The Pentium 4's thermal diode, for instance, is situated near a corner of the die.
Perhaps, but the temperature difference can't be that great I would imagine. Now with the excellent thermal properties of silicon...Besides, there are still plenty of transistors on the outer edge. Surely -50 C on the edges and warmth in the middle would bork up resistance?
The effect of increasing resistance in semiconductors is called "carrier freeze-out". The name itself explains quite well what actually happens. Resistance of a semiconductor is greatly dependant on the number of electric carriers (electorns in n-type or holes in p-type semiconductor). The nuber of carriers is dependant on doping (how much "additives" were added to the semiconductor in the manufacturing process), type of semiconductor and temperature. Oh and forgot - the temperarue at which this effect starts to manifest itself is about 100K or -170 degrees celsius. If you're interested - here's the Britney's Guide to Semiconductor Physics
Are you sure they do, with the processor running? Supercooled systems normally delay cpu startup until the cooling system has had time to get going and the chip sensor is below freezing. I thought it was the temperature difference that counted. Say you've got a heat-exchanger rated at 0.5C/W on the core. To keep the core below 70C with coolant at 20C (50C head), processor can generate up to 100W. Lower the coolant to -30C (100C head) and the processor can generate 200W without going over 70C.
