Author: Uri Blass
Date: 12:10:33 08/30/04
Go up one level in this thread
On August 30, 2004 at 11:55:52, Peter Skinner wrote: >On August 29, 2004 at 21:27:55, Robert Hyatt wrote: > >>Here is a quote from ICC: >> >>quote on=================================================== >>DIEP(C DM)(64): because they claimed having 'solved' chess and people like hyatt >>supporting that indirectly (by saying that nothing ever can get better than that >>old program) >>quote off================================================== >>In fact, I have been quoted more than once where I predicted that hardware >would eventually take us well beyond DB's speed/performance. > >Well DB was incredibly fast, and to my current understanding it also had >knowledge that no _current_ program has. I could be wrong as I am not 100% sure >what most commercials are doing. > >Based purely on results vs GMs, DB was and still is the best. Of course that is >relative as the strength of GM play and the play of computers has increased as >well. > >DB completely dominated the ACM events almost as Shredder has in recent years. >Shredder although being #1 in the SSDF list, has considerable problems facing >human opponents. There have been many times that Shredder has played a GM on the >ICC server and was made to look like a complete patzer. And these GM's were >hardly the strength of Kasparov. > >Belle was another hardware based program that did _extremely_ well, but like all >hardware programs there is a wall they hit and there is little they can do to go >past it, thus falling into the catagory of "what once was". > >>My quote was that in 1997, it would take 10+ years for a micro-computer based >>program to approach DB's speed. Today I can hit 10M nodes per second on a quad >>opteron, 20M on an 8-way. Probably approaching 40M on a 16-way box. That isn't >>as fast as DB, but it is in the ballpark. And I still have 3 more years on my >>"prediction". Next year AMD has promised a dual-core opteron, so that 16-way >>box will instantly become a 32-way box. 80M if there is no clock speed >>improvement, yet they say it will be faster via clock as well. So 2007 may be >>enough time to hit 200M roughly, if not more. > >As I stated above, speed _and_ knowledge were part of DB. Even if today's >programs were to get to DB's speed, I have serious concerns that the knowledge >factor would still play an important role in it maintaining it's legendary >status. It seems that you forget that knowledge is not only function of hardware and today people who develop programs have more knowledge then the knowledge that they had in 1997. Programs of today can have in many ways better search and better evaluation than deeper blue simply because people today know better and tested for more time. deeper blue may have the possibilities to add things to the evaluation for free but more knowledge is not always better knowledge. The top commercial programs of today can beat the top commercial programs of 1997 even if the programs of 1997 get 3:1 hardware advantage and the reason is simply more knowledge of the programmers. I do not know if the commercial programs of today can beat the commercial programs of 1997 with 10:1 hardware advantage(they cannot do it in blitz) but I suspect that slower time control can help. 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.