dave@dms3b1.UUCP (Dave Hanna) (03/18/89)
In article <13345@steinmetz.ge.com> davidsen@crdos1.UUCP (bill davidsen) writes: >In article <10866@pasteur.Berkeley.EDU> jsilva@cogsci.berkeley.edu.UUCP (John Silva) writes: > >| A 16Mhz 386 can be successfully used at 20Mhz, *as long as they are kept cool*. ... >| The reason for this is the chip will generate a certain amount of heat in >| normal operation (which the specs account for), and running them at a higher >| clock rate will generate excessive amounts of heat. This reduces the life >| of the processor, and the operating temperature range. > You may have a good conclusion, but I think you got there using >incorrect information... ... > Chips labeled as 20 MHz are tested at that speed and are guaranteed >for it. Chips labeled for lower speeds may be either (a) chips not >tested at the higher speed, or (b) chips which failed the higher speed >and passed the labeled speed. > > Intel says that they test under conditions which can not be duplicated >casually, using extremes of rated voltage and temperature. > The chip does not change by putting a label on top which says 16 MHz. >There is no reason to suspect that a chip rated at any speed will run >hotter than a chip rated a any other speed. I therefore disagree that >running a slow chip at a high speed will "reduce the life of the >processor" more than running a fast chip. They should run at the same >temperature and have the same mean time to failure. > I agree - running the chip at a higher speed will not "reduce the life of the processor" because of heat. But heat does enter in -- the physics are such that propagation delay increase with heat, and thus maximum clock speed slows down. That's why Intel (or any manufacturer) tests them at extremes of temperature. A chip that will run at 20 MHz at room temperature (about 20 degrees C) may very well fail at 70 degrees C. (the upper limit on commercial parts). And it's hotter inside the box, and hotter still at the semiconductor junction, which is what matters to the physics of things. Also, running them faster does dissapate more power, therefore raising the semiconductor junction temperature. But because they test them at such extremes, and usually place a guard band around their testing (e.g., if it's speced for 70 degrees, they probably test it at at least 75, and maybe 80 degrees), you can very often get it to work beyond what it's speced for in normal operation. It's just that you're taking the risk. If it fails, you can't take it back to Intel and say this doesn't work at 20 MHz. You might be able to take the CPU box back to the equipment vendor, but the failures you get are liable to be extremely erratic and hard to reproduce. Anyway, same conclusion, different reasoning. It will probably work, but don't ever bet your life on it. > I think we've gotten to the same point from different directions. >NOTE: > Keeping *any* CPU cool is a good idea, >regardless of it's rating for speed or temperature. Remember "a cool >electron is a happy electron." > bill davidsen (wedu@ge-crd.arpa) -- Dave Hanna, Infotouch Systems, Inc. | "Do or do not -- There is no try" P.O. Box 584, Bedford, TX 76095 | - Yoda (214) 358-4534 (817) 540-1524 | UUCP: ...!killer!gtmvax!dave |