Author: Robert Hyatt
Date: 11:14:59 01/16/02
Go up one level in this thread
On January 16, 2002 at 09:53:17, Uri Blass wrote: >On January 16, 2002 at 09:27:25, Robert Hyatt wrote: > >>On January 16, 2002 at 06:38:05, Uri Blass wrote: >> >>>On January 15, 2002 at 23:40:46, Robert Hyatt wrote: >>> >>>>On January 15, 2002 at 21:46:45, Jorge Pichard wrote: >>>> >>>>>It feels good to read the history of computer chess. >>>>> >>>>>http://www.geocities.com/SiliconValley/Lab/7378/comphis.htm >>>> >>>> >>>>This isn't nearly as amazing as the fact that they could do 13 years ago >>>>what the very fastest machines of today can do _now_. >>>> >>>>:) >>> >>>I do not find it amazing but the fastest machines of today can play better >>>thanks to better software and better hardware. >> >>Nobody has said "better software". _purely_ better hardware was the >>topic. Raw speed. Today's PCs are roughly as fast as deep thought was >>in 1989. That was all that was said... >> >>> >>>Nodes per second are not everything. >>>The hardware of 1989 could not detect repetition in the last plies when the >>>hardware of today has no problem. >> >>So? DB1997 could also do this... > >The comparison is between the hardware of DT1989 and the hardware of today and >DB1997 is not relevant. OK... let's get on the _same_ page then. I was talking about hardware, period. DT 1989 was about as fast as today's programs on today's fastest hardware. _you_ added the "software" part which was irrelevant... Lots of software programs in 1989 didn't even handle 50 move draws correctly. So what? If you compare simple hardware speed, then comparing today's NPS with 1989 NPS is a good point for comparing. But if you take some particular feature of today's programs and try to say "that is better than what DT did in 1989" the response _must_ be "so what?" because they fixed the repetition issue in DB. Of course, DB was over 100X faster than DT, at a minimum, so taking 1997 DB2 speed and comparing that to today's PCs leaves a +lot+ to be desired and a long way to "catch up". And 1997 DB didn't have the repetition problem at all... I prefer to simply discuss hardware in this context, since the repetition issue was not a +/-25 Elo problem... It was smaller... > >If the hardware of today can detect repetitions in the last plies when Deep >thought could not do it then it is relevant for the comparison. Why, when it wasn't an issue in 1989? Some things have improved since 1989, to be sure. Hardware speed is one of them. But computers have _just_ now caught up to DT/1989 in terms of speed. The repetition problem did not have great effect on DT's skill, as evidenced by its ability to beat GMs at 40/2hr games at the time. >> >> >>> >>>It is also clear that there was a big progress in the software from 1989. >>> >>>Uri >> >>"some progress"... > >I guess that there was more than 100 elo improvement in comp-comp games >because of better software and I consider more than 100 elo as big improvement. > >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.