Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Coparing two Identical Programs using Different Processors Speed !

Author: Ed Schröder

Date: 13:56:06 01/29/01

Go up one level in this thread


On January 29, 2001 at 16:22:51, Bertil Eklund wrote:

>On January 29, 2001 at 13:09:10, Christophe Theron wrote:
>
>>On January 28, 2001 at 19:19:55, Hristo wrote:
>>
>>>Christophe,
>>>I do beleive you are "wrong" (! ;-) ) and
>>>Jorge is correct. However Jorges test doesn't undoubtedly prove
>>>his conclusion. In some cases it is not a prove at all. ;-)
>>>
>>>It is much more likely that some programs benefit more from
>>>increased CPU (memory, ...) performance than others.
>>>This is the case with many computer aided algoritms in general!
>>>Take for example linear search versus binary search. Then use those
>>>algorithms on a slow computer than can only generate 10 items to be searched
>>>and another faster computer that can generate 1000 items. This is self evident,
>>>no?! Computer chess programs present us with a significantly more
>>>complicated algoritm which in its own right is not a perfect solution
>>>to the problem at hand (chess). Firstly the benefit from improved performance
>>>might not be large enough to measure. Secondly the "benefit" (extra more ply
>>>than the opponent) might cause worst game results. (!!!)
>>>
>>>Perhaps someone has done this before.
>>>Take two computers C1 and C2. Where C1 is half the speed of C2.
>>>Take two programs A and B.
>>>Play a match of 100 games using the same program on both computers:
>>>dA = A-on-C1 vs A-on-C2
>>>dB = B-on-C1 vs B-on-C2
>>>
>>>? dA > dB then A benefits more from higher speed.
>>>
>>>This is not perfect test. However I'm sure you are going to get consistently
>>>different (dA != dB) results.
>>>It would be interesting to know what a test like that yelds ... ;-)
>>>
>>>hristo
>>
>>
>>
>>Of course it would be interesting and I'm ready to change my mind if a relevant
>>experiment shows I'm wrong.
>>
>>But nobody cares about doing it.
>>
>>On the other hand, there is data proving (or at least suggesting) that faster
>>hardware does not impact on relative playing strength: have you noticed that
>>blitz tournaments results almost always look like the SSDF list?
>>
>>A huge blitz tournament has been played recently (a lot of games where played,
>>which makes the final result interesting), and a member of the SSDF has pointed
>>out that the result looked exactly like the top of the SSDF list. You can still
>>find the messages on this forum.
>>
>>
>>
>>    Christophe
>
>Hi!
>
>You are right except for two exceptions (if you take the top 10 program in SSDF)
>I have played several houndred ( in some cases thousands) blitz-games with
>almost everyone of them. There is no big difference except for Nimzo7 that is
>clearly weaker in blitz vs tournament-time control and Hiarcs that is better in
>blitz (Uri says it is because the "hash-bug") As you know Genius can still
>compare in blitz with all programs on a "slow" computer but is almost without
>chance on 2h/40. I believe Marcus Kästner has the same impression of the above
>programs as he is aware of a lot of blitz-games.
>
>So in this case Jorge are right about Nimzo but I can't understand that he is
>sure after 9 games!
>
>Bertil

You forget Mchess and Cstal, both are much better at longer time controls.
The same applies for Reb6-10 and Century 1.0, better at longer time controls
than at blitz. I believe Bob also claims this to be true for his Crafty.

Ed




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.