Author: Tom Likens
Date: 10:42:36 04/26/02
Go up one level in this thread
The physics is correct, but the understanding of how a microprocessor works is incorrect. A microprocessor runs off a fixed rate crystal and PLL, which do *not* vary with temperature even if they have been running for hours, days, weeks or months. Of course, there is cycle-to-cycle jitter and long-term drift but good phase-lock loops (PLL) compensate for these effects. When the microprocessor is designed it is designed based on a fixed cycle time for the system. In other words, a 1-GHz CPU has a cycle time of 1ns between rising edges of the system clock and everything must happen in that 1ns cycle time (exactly what must happen depends on the architecture- is it pipelined, superscalar etc). Before taping out the chip it is tested against all the process corners to verify that it works at low voltage, high temperature, slow silicon or high temperature, low temperature and fast silicon, etc. etc. This is where most hardware designers spend their time. But (and this is the important part) unless it is an asynchronous design, it will *NOT* run any program faster if you cool off the system or raise the voltage. Yes, the actual silicon will be faster, but that matters not one wit. If it works it works, regardless of these other conditions. That's the beauty of synchronous design and it's what makes the chip industry possible. The processing speed, which is the part chess software cares about, is fixed by the crystal and the system's operating frequency. Binning parts is a whole different matter. When you bin parts you test them at manufacturing test to find out how fast the will reliably run and then you select a crystal of the appropriate speed. Also overclocking a chip refers to running the basic clock frequency faster (for example our 1ns cycle time now becomes 900ps and thus we do the same amount of work in less time). regards, --tom On April 25, 2002 at 15:19:58, David Dory wrote: >On April 24, 2002 at 23:37:50, Ratko V Tomic wrote: > >>>"It's just physics -- the longer a processor runs, the more it heats up and the >>>less efficient it becomes. Your program will play better chess right after you >>>turn your computer on than it will after the machine's been running for hours or >>>even days" >>> >>>Say what?? >> > >What the original quote referred to is that as a conductor heats up, the >resistance increases. Basic physics/electricity. > >If the CPU continually increased in temperature, of course, it would be true, >but then your CPU would be a puddle of silicon, etc. Of course, that's not the >case at all. Never was that way that I'm aware of. > >Since modern CPU's heat up so high, so quick, this isn't a factor. I can recall >testing many a 286/386 where _only_ a test done in the first 3-5 seconds would >give you the highest CPU performance index. After that it leveled out, and never >would give you the very best index again until you shut it off for quite a while >and let it cool all the way down. > >darrz
This page took 0 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.