Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: So why *does* Fritz beat Crafty?

Author: Eugene Nalimov

Date: 18:13:26 03/28/99

Go up one level in this thread


On March 28, 1999 at 19:17:15, James T. Walker wrote:

>On March 28, 1999 at 15:46:53, Eugene Nalimov wrote:
>
>>On March 28, 1999 at 09:15:11, James T. Walker wrote:
>>
>>>On March 27, 1999 at 15:43:42, Eugene Nalimov wrote:
>>>
>>>>On March 27, 1999 at 12:45:05, Bruce Moreland wrote:
>>>>
>>>>>
>>>>>On March 27, 1999 at 03:55:25, blass uri wrote:
>>>>>
>>>>>>It is not fair because part of the effort in doing the program crafty is by
>>>>>>doing it a parallel machine.
>>>>>>I believe that Bob could do in the same time a better program if he did not
>>>>>>waste time for doing a SMP program.
>>>>>
>>>>>If someone wants to compare Crafty and Fritz I think it would be fair to compare
>>>>>them on uniform high-end (single-processor) hardware, since they are both
>>>>>designed to work on that hardware.
>>>>>
>>>>>If one of them would only run on a 286, I don't think it would be fair to make
>>>>>them both run on a 286.
>>>>>
>>>>>But multiprocessor machines are still a super- high-end thing so it's probably
>>>>>not fair to say:  Here is the machine, it has 4 processors, feel free to use
>>>>>them in this match.  Oh, what did you say Fritz, you can't use 4 processors, you
>>>>>can only use one?  Well, that's too bad for you.  You might as well put them
>>>>>both on an Alpha and expect Fritz to use an emulator.
>>>>>
>>>>>In a few years, maybe, because everyone will have a multiprocessor machine, but
>>>>>of course everyone will be multiprocessor then.
>>>>>
>>>>>Bob's put time in being SMP, sure, but I think he supports single-processor
>>>>>machines and runs well on them.
>>>>>
>>>>>bruce
>>>>
>>>>Bob spent his time working on SMP. Also, he deliberatly lost some
>>>>performance by using C instead of assembly.
>>>>
>>>>Author of Fritz decided not to include SMP code, as well as write
>>>>his program on assembly to squize last pieces of performance.
>>>>
>>>>By using single-CPU x86-compatible machine you favor Fritz - he will
>>>>be running on the best possible platform.
>>>>
>>>>Maybe it's better to put some dollar limit - e.g. "on a machines
>>>>that cost not more than $7,500".
>>>>
>>>>Eugene
>>>
>>>Maybe it's better to say "For the average user".  The average person does not
>>>have a $7500 machine.  For the average user ($2000 PC or less) Crafty comes up
>>>way short of Fritz and Junior and the other top programs made for PC's.  So for
>>>comparable speed machines, Crafty gets beat more often than not.  If you want to
>>>put Crafty on a 4 processor machine  which increases it's speed by a factor of
>>>say 3.5 then give Fritz a comparable speed increase and it will still come out
>>>on top.  This is not a put down of Crafty.  I have often wondered the same thing
>>>about why Crafty gets beat by the top programs when Crafty seems to have all the
>>>modern techniques of chess programming.  The question begs for an answer not to
>>>put down Crafty but to search for weakness which can be overcome.  I believe
>>>this will take some analysis by master chess players which I am not.  I believe
>>>this question was given in the sense of trying to find an answer which will
>>>eventually make Crafty a better program.  Everyone appreciates the fact that
>>>Crafty is portable to different platforms because of the C language.  This has
>>>to cost some rating points but I don't believe it accounts for the majority of
>>>the rating difference between Crafty and the top programs.
>>>Jim Walker
>>
>>I was not the first who started the talk about "top of the line
>>computer". I just pointed that for the same money you can buy
>>computer that will be much more 'Crafty-friendly'.
>>
>>If you want to go to more reasonable price range - Ok.
>>Here is a surprise, too - computer that will give Crafty
>>advantage in computing power over Fritz can be *cheaper* than
>>top of the line 'Fritz-friendly' computer in the same price
>>range. For example, dual PII/350 cost less than PII/450;
>>dual PII/400 cost less than PIII/500. Also, Crafty needs
>>less memory than Fritz for its' hashes.
>>
>>Eugene
>
>Hello Eugene,
>I can appreciate your line of reasoning.  Crafty has an advantage with multiple
>processors.  I believe Crafty is showing other programmers the way of the
>future.  I don't think processors will get much faster than 1-2 giga-hertz if
>that fast.  Speed increases beyond that will naturally turn to multiple
>processors and everyone will have to learn to use them (Of course only my
>opinion).  None of this explains Crafty's inability to beat or even play equal
>to the top commercial programs on equal hardware.  To me this suggest there is
>something missing in Crafty which limits it's strength.  I am only interested in
>what it is.  What would make Crafty on par with the top commercial programs?
>Jim Walker

Another one-two years of development, of course.

Eugene



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.