Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: WMCCC Hardware

Author: Graham Laight

Date: 02:19:55 10/06/97

Go up one level in this thread



On October 03, 1997 at 13:37:18, Chris Whittington wrote:

>
>I'be been following the  rgcc thread over WMCCC and hardware. These are
>my thoughts:
>
>I don't think this WMCCC is about who wins it, but about which of the
>'groups' of programs do best.
>
>With the danger of annoying loads of people, it seems to me that group
>(1) consists of what I shall loosely call the Hyatt-paradigm programs.
>These include IMO Crafty, Ferret, Shredder, Nimzo-Hydra, Fritz et al.
>
>Any one of these programs could win, my guess is that it will be a
>mostly a matter of luck which one. To paraphrase Napoleon, Ferret is
>making his own luck with a few gigaHertz's of alpha and stands a good
>chance.

As I've said before, in motor racing, you may use different
manufacturers to supply the engine. However, size of engine, use of
turbochargers, and stuff like this is strictly controlled. The
equivalent in computers would be to control the clock speed and RAM
IMHO. 200 Mghz and 32 Mb RAM, say I.

>Group (2) consists of the older knowledge and tuned programs, many of
>these commercial; and which some people think may have been superceded
>by Group (1).
>
>Group (3) may consist of CSTal all on its own.

>As Bob has said many times, the knowledge-speed thing  is a trade-off.
>My own guess is that depending on the time of year, and the year, and
>the hardware, and engineering competence; one side gets an advantage
>over the other for a time. Maybe the time of the Hyatt-paradigm is now.
>
>So the relative performance of these groups interests me.

On RGCC, Mclane seems to have resigned on your behalf, stating that the
faster  the processor, the more likely it is that Crafty will beat
CSTal.

This is exactly the opposite of what one would expect. Normally, the
less knowledge there is in a program, the more it suffers at higher
speed because, at greater search depths, the expectation is that tactics
count for less and positional knowledge counts for more.

The results that Mclane reports are difficult to reconcile with the
prevailing doctrine.

>On the issue of the fast alphas: hmmmmm. Its going to confuse the issue.
>For sure a win by a program on faster then PC hardware is going to open
>up questions. Personally, I'ld prefer to see it restricted to top
>Pentium Pro speeds.
>
>
>CHris



This page took 0.12 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.