Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How does Maastricht Diep compare to Deep Blue?

Author: Uri Blass

Date: 11:02:12 06/29/02

Go up one level in this thread


On June 29, 2002 at 13:25:01, Vincent Diepeveen wrote:

>On June 29, 2002 at 12:30:05, Roger D Davis wrote:
>
>>I see Diep is running on some incredible hardware for Maastricht. How does
>>today's Diep compare against Deep Blue?
>
>it will be answerred. Note that the last 13 years not a single supercomputer
>won a world title, including deep blue (which never won under
>the name deep blue a world title. Instead it lost in hong kong and drew
>a few horrible games too).

It drew only one game against wchess in hong kong.
>
>Comparing 2002 software with Deep Blue is of course insane in itself,
>because the progress, especially algorithmically and weakest chain
>from a program has progressed a lot. For example Deep Blue had 4000
>hand given in bookmoves by a grandmaster, the rest was a random book.
>
>Any serious program of today has more like 100000+ hand tuned bookmoves,
>majority of them top grandmaster moves which have proven themselves, as
>well as refutations of pretty recent lines.
>
>I would be pretty amazed if in the top10 of the world champs a program
>would end which has a hand tuned openingsbook of 4000 moves or less
>(of course using after that a database is something else).
>
>Remember a very old pc, P90 could beat deep blue in 1995. Imagine what
>happens nowadays. The 1997 version was not so much faster in nodes a
>second than the 1995 version, especially when taking their branching
>factor into account.

The 1997 is supposed to be much faster than the 1995 version.

I remember numbers of 200M nodes per second for the 1997 version when I am not
sure about the exact number for the 1995 version but I remember something like
2M nodes per second.

I believe that the top programs of today are better than deep blue of 1997 but I
also believe that deep blue of 1997 is at least 100-200 elo better than the deep
blue prototype of 1995.

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.