Author: Robert Hyatt
Date: 09:41:50 05/02/01
Go up one level in this thread
On May 01, 2001 at 06:34:13, Tony Werten wrote: >On May 01, 2001 at 04:53:49, Uri Blass wrote: > >>On May 01, 2001 at 04:33:28, Tony Werten wrote: >> >>>It's already a long discussion but one thing is unclear to me. >>> >>>Why does GT have to be smp to participate in the challenge for Kramnik >>>tournament ? Last dutch championship Fritz participated on a dual machine and >>>finished third. Tiger won the tournament with 2 points more, the King (also >>>single processor) finished second. >> >>Yes but Fritz used a dual and not a 8 processor machine. > >There isn't a program that makes efficient use of 8 processors. If they're lucky >the get a speedup of 4, 2 processors gives 1.7, not a big difference. That is wrong. I have run on 4 and 8. I have given my speedup numbers. For 8, it was roughly 6x faster. For 16, it was nearly 12x faster. For my quad, it is roughly 3.1, although the most recent test I ran for a paper I am working on produced 3.3 on the kopec-bratko test suite. Even 3.3 would be significant... as it would represent 1 ply deeper than a single-cpu program. > >Even more, afaik they are not playing on a 8 processormachine right now. > Which means the qualifying match is simply very poorly thought out. You don't use a dual to predict performance on an 8-way box. >> >>> >>>So why exclude a single processor program that has proven to be capable of >>>beating multi processor programs ? >> >>Single processor program is capable pf beating multi processor in one game but >>the difference between programs is not so big that it can be better than 8 >>processor machine in a long match. >> >>Another point is that the sponsors want to support multi processor system so >>they cannot use single processor programs for that purpose. > >That's the point I don't get. Somebody should have told them about the silly >discussion that appear every year with the worldchampionships. Program A will >win because it uses 700 Mhz and program B only uses 550. > >Tiger ( and Shredder btw ) has proven to beat multiprocessor programs. It's >about the quality of the program, not the Mhz is uses. Else Zugzwang would have >won more tournaments. > >Tony > >> >> >>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.