Author: Robert Hyatt
Date: 06:34:32 08/12/00
Go up one level in this thread
On August 12, 2000 at 00:50:18, Uri Blass wrote: >On August 12, 2000 at 00:03:55, Robert Hyatt wrote: > >>On August 11, 2000 at 18:04:16, Mark Young wrote: >> >>>On August 11, 2000 at 07:37:19, Vincent Diepeveen wrote: >>> >>>>On August 11, 2000 at 05:59:53, Uri Blass wrote: >>>> >>>>>On August 11, 2000 at 04:11:12, Jouni Uski wrote: >>>>> >>>>>>I wonder, if it's possible to arrange match between the two top rated (against >>>>>>humans) programs: Deep Junior 2702 (8*700Mhz) and PConners 2663 (160*300Mhz)? >>>>>>Mini match with 4-10 games will give interesting info about their strength! >>>>>>Just on idea... >>>>>> >>>>>>Jouni >>>>> >>>>>I guess that p.conners is going to lose. >>>> >>>>>P.conner is not commercial and it is a big advantage against humans and inspite >>>>>of this advantage Deep Junior did better results. >>>> >>>>>I think that a tournament with Deep Junior(8*700),Fritz6b(8*700) and Crafty on >>>>>the best alpha machine that it can use will be more interesting >>>> >>>>Diep 8*700 is more interesting than that. I don't see deep junior NOR >>>>fritz ever do better moves as they get like 15 ply anyway. DIEP hardly gets >>>>to tactical sufficiency. >>>> >>>>Your speculation is completely on the wrong ways. >>>> >>>>Speeding up positional weak programs like junior and fritz is not gonna >>>>help much. >>>> >>>>As for tactical power against p.conners, with 8 processors you don't outgun >>>>a 180 processor 3000$ an hour or so machine. >>>> >>>>I'm not sure that junior would in lippstadt have achieved the same >>>>result. Drawing junior is just TOO easy. >>>> >>>>I'm just 2255 fide rated , though very likely IM within a few years >>>>and 2400 rated, that's still not much nowadays to play computers. >>>> >>>>But especially drawing programs like junior, which because of their >>>>huge positional gaps is possible, is real easy. Winning is something >>>>else of course. >>>> >>>>What junior showed is that with even 2 pawns ahead against Khalifman >>>>it can't win. >>>> >>>>The only bad player of the tournament won from junior in 23 moves or >>>>so. >>>> >>>>Despite that all it's obvious that it's an advantage to know what >>>>program you play, (a huge advantage), nevertheless none of those >>>>players that junior played has ever used junior on a commputer, >>>>i'm quite sure of that. >>> >>>This is wrong, the players did play against Junior before the tournament. Many >>>of the players said so. >> >> >>That is only partially true. They _could_ play against Junior. But _not_ >>the Junior playing in the tournament. Nor could they play against an 8-way >>xeon 700 box either. >> >>This is similar to Kasparov using Fritz to prepare to play deep blue. > >This is not similiar because Fritz and Deeper blue is not similiar when Junior >and Deep Junior are similiar. > >I believe that Junior6a can produce 80-90% of the moves of Deep Junior if you >give it more time when kasparov had not something that can produce even 70% of >the moves of deeper blue. If you really believe that, then you should be able to pretty well prove it, by simply taking the last two annual releases of Junior and comparing them. This version of Junior is probably going to be released late this year, as that is when most new versions of commercial engines get released. I personally believe that there will be _significant_ differences between the released version of the program that played in Dortmund and the released version of junior from late last year... I _know_ there are serious differences between (say) crafty of december 1999 and crafty of december 2000. Ditto for Rebel, fritz, etc... > > It >>would be worse than useless, because what you think you see as weaknesses >>may or may not exist on a machine 8x faster than anything they might normally >>use > >The speed difference is smaller. >700*8 is not 8 times faster than 700 and they could also use Deep Junior on >machine that is 2 or 3 times slower. It ought to be 6 times faster. That is the number I get on an 8-way box. In fact, my number is above 6, although I have not run on a primergy and don't know how well their memory architecture scales. But the main point is _not_ speed, it is "different engine". > >Based on the number of nodes per second I think that the 8*700 that they used is >only 8 times faster than my 450 and if you consider only effective nodes then I >think it is even less than it. > >Uri remember you are comparing a released version vs an unknown experimental version. Who knows what has changed? My program is significantly slower than last year, due to evaluation additions. Has nothing to do with multiple processor scaling..
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.