Author: Dann Corbit
Date: 08:52:52 11/02/98
Go up one level in this thread
On November 02, 1998 at 05:58:17, Jari Huikari wrote: >Well... I don't know if the swedes already does it. >(And anyone save me knowing it already as usual...) > >But it game into my mind: > >How about fixing some current hardware to be used in testing chess program >strenghts (playing games program versus program)? Same hardware to be used >by all the programs. Using the same standard hardware year after year to >get certainly comparable results and to see how much progress really >happeness in software. > >What do you think? I think this will be very hard to achieve. Besides which, hardware keeps moving forward exponentially. How much will you care how well a program does on a 486? I think a better idea might be a standard calibration suite. We run the same program on many versions of hardware both solving test positions and in actual combat against computers and people. Short, medium and long time controls. It would be a lot of effort, but when complete, we would have a pretty good idea of the value added by a new architecture. Hence, we would have projected ELO for any calibrated machine for which we have run the suite. A non-linear curve fitting program might be useful. I think that there is a danger people would simply try to extrapolate without actually testing though. I have noticed that a 450Mhz PII does better than 450/300 times as well as a 300Mhz PII. So to really know, you will have to test it.
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.