Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The Limits of Positional Knowledge

Author: Robert Hyatt

Date: 11:04:25 11/12/99

Go up one level in this thread


On November 12, 1999 at 12:28:26, blass uri wrote:

>On November 12, 1999 at 11:49:12, Robert Hyatt wrote:
><snipped>
>>I believe that 'diminishing returns' does happen, _if_ you take a static engine
>>and run it on faster and faster hardware.
>
>It is possible that the new programs earn more from fast hardware relative to
>old programs and that this is the reason that I do not see the diminishing
>return.
>
>I compared the difference of the same programs of the ssdf list
>between p200 and p90 and between K6-450 and p200 and I was surprised to find
>that the improvement from K6-450 relative to p200 is bigger than the improvement
>between p200 amd p90
>
>I admit that it is possible that the programs I compared between K6-450 and p200
>earn more from time and that there is a diminishing return from being twice
>faster.
>
>The only way to check it is to use the same programs on 3 different hardware or
>with 3 different speeds.
>
>Uri


The main thing I have noticed (often) is that deeper searches require re-tuning
of the eval to play optimally.  I got burned in 1985 for tuning on a vax and
playing on a Cray, and barely found this problem after round two of the 1986
WCCC tournament.

I try to 'play' on what I 'test' on when possible now.  IE I would definitely
feel more at ease playing anyone using my quad xeon, as opposed to a 16-way
alpha machine.  The alpha would be nice, but no telling what kind of eval
glitches might show up with the 2-3 additional plies it would get...



This page took 0 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.