Author: Robert Hyatt
Date: 05:29:37 02/11/00
Go up one level in this thread
On February 11, 2000 at 01:02:36, Tom Kerrigan wrote: >On February 10, 2000 at 23:27:52, Robert Hyatt wrote: > >>The physics don't support you there. IE gates don't function faster at >>absolute zero, but the resistance goes to zero. I know of no physics law >>that says electrical signals propagate faster at lower temperatures... > >I'll try to avoid saying something rude here. > >I'm sure you know that gates are not wires that simply propogate electrical >signals. So while lowering the temperature will not cause the electricity itself >to flow faster, it WILL change the amount of time it takes to induce a channel >in the substrate between source/drain. Sure... but give me a citation for a faster switching time. I am looking at a data book right this second, that is describing the properties of a specific OR gate. It gives the settling time for the range of temperatures 0c - 60c, and the settling time is constant. I don't see any way to operate a gate below 0c in a real machine. IE Cray couldn't do it because any exposure to air causes a 'serious' problem. So please provide a citation as I am interested in the effect. I can't find a thing in the data books I have (covers all sorts of specific integrated circuits, gives their properties, pinouts, temperature range, settling times, etc...) that would suggest that if I drop the temperature significantly that the device will suddenly operate at a higher clock frequency. I did find several references that said "if the size of the device is reduced, and heat can be controlled, switching time can be driven down..." But I interpret that as a function of size, not temperature. Smaller is harder to cool. > >>Of course, you can ramp up the voltage to make them switch faster, and you >>can make them smaller, because cooler temperatures combat the heat rise for >>smaller junctions. But I sure don't see why they would switch faster. If >>they did, Cray would have run his stuff at really cold temperatures since he >>was speed-centric... > >I remember reading that several Crays required massive amounts of liquid >nitrogen to flow over the processors during operation. > >-Tom You remember wrong. They used an electrically inert fluorcarbon liquid, cooled to just above 0c. The circuit cards were immersed directly into this liquid during operation. The machine was the cray-2, and then the cray-3. But they did not use liquid nitrogen. It has a horrible effect on circuit longevity. A couple of cycles of hot and cold and you have silicon dust. You can find details about this at www.cray.com... Or you can go look at the prototype cray-3 at NCAR which I think is near where you live. There are also at least a couple of cray-2's there. Machine is impressive, but never was based on superconducting.
This page took 0.03 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.