Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Kaspy vs. DJ (any idea what kind of hardware????)

Author: Uri Blass

Date: 22:15:02 01/23/03

Go up one level in this thread


On January 24, 2003 at 00:52:18, Robert Hyatt wrote:

>On January 23, 2003 at 21:05:55, Vincent Lejeune wrote:
>
>>On January 23, 2003 at 14:21:17, Robert Hyatt wrote:
>>
>>>On January 23, 2003 at 10:55:10, Uri Blass wrote:
>>>
>>>>On January 23, 2003 at 10:32:10, Robert Hyatt wrote:
>>>>
>>>>>On January 23, 2003 at 04:30:50, Daniel Clausen wrote:
>>>>>
>>>>>>On January 23, 2003 at 04:19:47, Jorge Pichard wrote:
>>>>>>
>>>>>>>On January 23, 2003 at 00:18:54, Robert Hyatt wrote:
>>>>>>>
>>>>>>>>On January 23, 2003 at 00:04:18, Chris Kantack wrote:
>>>>>>>>
>>>>>>>>>I have seen the rules and other "tidbits" of the upcoming Kasparov vs. Deep
>>>>>>>>>Junior match.  There's even a site where you can bet on the outcome.   But what
>>>>>>>>>of the hardware?   2, 4, 6, 8 processors or more????   Processor speed??
>>>>>>>>>
>>>>>>>>>Any official info yet?
>>>>>>>>>
>>>>>>>>>Thanks,
>>>>>>>>>Chris Kantack
>>>>>>>>
>>>>>>>>
>>>>>>>>The only viable choices from the Intel world are a dual xeon 2.8, or a quad
>>>>>>>>xeon 2.0...  I don't think anything else comes close.  There might be some
>>>>>>>>quad 2.8's in the hands of a select few, but they will probably be hard to
>>>>>>>>get access to.
>>>>>>>
>>>>>>>
>>>>>>>It would be great if Amir gets his hands on a Quad 2.8, but a Quad 2.0 would do
>>>>>>>just fine. Why they have not posted any information on the hardware to be used
>>>>>>>will be simply a surprise :)
>>>>>>>
>>>>>>>Pichard
>>>>>>
>>>>>>
>>>>>>It's interesting to see that people are so focused on the hardware used in these
>>>>>>matches. Of course hardware is important, but I guess that most people (after
>>>>>>some thought ;) that the actual hardware used in a comp-human match is a bit
>>>>>>less important as opposed to a comp-comp match.
>>>>>
>>>>>absolutely correct.  The influence of going 2x faster against a human is
>>>>>minimal.  If
>>>>>it is all about search, the human will lose no matter what.  But it isn't, yet,
>>>>>all about
>>>>>search, if all you do is search deeply to find lost positions. :)
>>>>
>>>>If you find lost positions by search then you can avoid the losing blunder.
>>>>
>>>>I do not think that we have data to say that the influence of being twice faster
>>>>is minimal.
>>>
>>>I have _lots_ of data.  I have run on ICC for years now, at in blitz games,
>>>switching to
>>>2x faster hardware doesn't have an appreciable effect on playing humans...
>>>
>>>
>>>>
>>>>games on ICC when Crafty is twice slower do not prove much because humans lose
>>>>most of the games in any case.
>>>
>>>Not really.  There are a few gms that get a significant number of draws to go
>>>along with
>>>their losses, and they even win a game here and there.  Going from 4x400 to
>>>4x550 to 4x700
>>>and even 2x2800 does not make that much of a difference.  I have even changed
>>>from 4x550 to
>>>2x2800 and then later asked them if they could tell any difference.  The answer
>>>was "no".
>>
>>
>>If it's real that the change 4x550 to 2x2800 don't change the quality of moves,
>>i'd suggest you to play with the 2x2800 and divide the thinking time of crafty
>>by 2 or 3, this would give you some more wins ! ;)
>>
>>
>
>I'm not saying more time doesn't mean better "quality".  I am saying that
>against humans, the "quality" is not the issue.  If you don't understand
>something that they do, then all the depth in the world is not going to help
>you make the right move.  It will play tactically better of course, but not
>necessarily positionally.  I've only noticed this against GM players.
>
>And I _have_ done as you have suggested many times...  :)

No this is different.

Dividing the time of the computer by 2 is not the same as running on hardware
that is twice slower.

If you divide the time of the computer by 2 the opponent has less time to ponder
on crafty's time and this is a real problem for humans.

Uri



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.