Author: Robert Hyatt
Date: 07:57:19 11/11/03
Go up one level in this thread
On November 11, 2003 at 10:23:26, martin fierz wrote: >On November 11, 2003 at 08:50:53, Aaron Gordon wrote: > >>It would be better had they used a quad or 8-way Opteron running 2GHz or more. >>From some testing I've done in the past you can figure a single Opteron 2GHz == >>a P4-3.6GHz in Fritz 8 (32bit mode). So, a Quad Opteron 2.0 == Quad P4-3.6. >>Almost 30% faster, plus the memory bandwidth available would probably push it a >>bit over that with large hash table sizes. 8-way Opteron 2.0 would of course be >>like 8 p4-3.6's (however with some 40gb/s+ memory bandwidth available depending >>on bus speed). >> >>Why not use the best hardware? Seems like if you'd want to promote your new >>'awesome' chess program you'd want to give it the best chance of winning. > >i believe hardware is vastly overrated in human-comp matches. this whole theory >of improving with increasing speed is based (AFAIK) on comp-comp games, starting >out with thompson's experiments with belle, letting the *same* program play >against itself with increasing search depth. *of course* it shows that searching >deeper helps. however, when you consider the search depth that these monster >machines reach nowadys (is it 16 or 19 ply?), you will see that they easily >outcalculate the human all the time *if there is something to calculate*. > >e.g. when hiarcs (or was it another program) allowed smirin to set up a fortress >in an endgame that was lost for smirin, it could have searched 10 ply deeper and >would not have seen that it's a draw. similarly, had kramnik played on in that >famous game he lost to deep fritz in bahrain he would have ended up in a >fortress position with R+P vs Q+P (IIRC), where fritz could have searched for >years and still believed to be winning. > >IMO the big thing in human-comp matches for the computer side is making sure >that you get a position where there is something to calculate. if you get that, >you can be running on a 1GHz machine or on a supercomputer, it doesn't matter - >you'll always outgun the human. the other option is to improve the program so >far that these holes disappear, which will happen sooner or later too... > >cheers > martin This can be debated from both sides with lots of facts and examples. IE in some positions, going 2x deeper would not help one bit without some understanding. Fortresses are one example. And there are lots of others. But, in other positions, having even another 30 seconds would let a program find a better move. Whether that turns a draw into a win, or a loss into a draw is another issue. So faster hardware won't help in _all_ cases. But it will certainly help in a fair number of cases. And any help at all is worthwhile...
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.