Computer Chess Club Archives


Search

Terms

Messages

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

Author: Peter Berger

Date: 14:32:05 01/29/01

Go up one level in this thread


On January 29, 2001 at 13:01:58, Christophe Theron wrote:

<snip lot of stuff >
>
>
>No, it's easy to do. Nothing tricky here. All you need is the hardware (many
>people have it) and a little time (maybe one week of computer time).
>
>But nobody does it, because it's more easy and fun to just guess and discuss
>endlessly about it.
>
>But by doing so people miss the real beauty of computer chess. Too bad.
>
>
>
>    Christophe


Thanks :-) ! In fact I couldn't agree more ! I also noticed Severi tends to
agree ..

I really think it is pretty easy to set up and to verify . Someone already tried
it before but used one of the worst examples with Fritz and Crafty in my opinion
( which whatever its author says is neither influenced by pondering or hardware
much ) ; in fact my private results even suggest a negative correlation for this
engine .

I definitely _will_ do this test although it might take a little time as the
amount of my spare time has currently reached an all time low ( same as the
amount of "lazy" CPUs ) .

My best canditates for the potential fast-hardware-gainers : Gandalf and Nimzo .

My best canditates for engines which won't gain _as_ much with faster hardware :
the Tigers or Crafty .

My candidates for engines which maybe should be avoided as they probably will/
might  prove something else : Hiarcs and yace .

Time control should be blitzy ; I will go for 5 minutes blitz , 3 secs increment
but anything similar should do .

With ERT it's a piece of cake ; 500 games on "fast" hardware , 500 games on
"slow" hardware ( only thing needed is a quick look at the final positions to
cross-check everything went OK ) . Maybe its worth to think a minute about
eliminating learning effects but shouldn't be a _too_ big issue anyway IMHO , so
I vote for ignoring it and simply play with default learning settings.

Also easy to set up for every Winboard fan ( /mg 500 , hint ) .

The _good_ thing : you just need _one_ free comp , set this up ( _every_ Rebel
11 owner or Winboard user has all that is needed !! ) and post the result here .
No need to do it on x compis ; the result of others should do the rest of the
job . If you read this : why not waste a little CPU time for the sake of science
??

I volunteer to give help to anyone who chooses a Winboard engine to do this test
with ERT under Rebel Tiger II and has problems setting it up ( simply email me )
.

Cheers .

pete



>
>
>
>
>
>> questioning a result created this way is pretty easy , creating
>>it pretty time-consuming IMHO .
>>
>>pete



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.