This is a big difference vs a stand-alone engine. So GM Benjamin was able to use his human intuit as to what move was best, or when to let Deep Blue keep on searching for a better move. They were able to pre-program certain opening lines in anticipation of Kasparov’s repertoire, and they, during the game, they could “force” Deep Blue to make the move that, at any time, was being evaluated as the best. There was one critical difference in Deep Blue vs Kasparov – Deep Blue had at least one GM operator (Joel Benjamin, I think).
Well the old rule was 90% of the program is in 10% of the code, so you could get gains by rewriting 10% of the code, but that may well be significant… and can’t in anyway be “automatic”… you can’t 64bitify an old executable or even the source code just by clicking “fix” !) (software has to be re-written for 64 bit. Double that for 8 cores, and you can get servers that run 32, 64 cores… and you can inflate the MIPS a little because they are 64 bit, if you are comparing to 32 bit data or 16 bit data CPU, but that for the same program, it won’t help.
Well the PC software looks more moves a head but does a less thorough search by missing (assumed) useless territory.ĭeep blue did chess steps at the rate that a 3 million MIPS regular computer would, but it was custom, somewhat chess specific, hardware so it wasn’t actually a 3 GIPS CPU system.Ī modern intel cpu with 4 cores is only 100,000 MIPS. Pretty much they are saying they don’t bother making Deep Blue software since its pointless having a software that is impossible to beat. They are saying that the PC software searches less positions per second. I’d hazard a guess that a powerful desktop computer would outclass 1997’s Deep Blue. The Computer Chess Rating Lists website shows the top ranked program Houdini with an Elo ranking of 3256 using a 4 core 64bit processor.Īccording to Wikipedia, Magnus Carlsen has an Elo rating of 2872. I’m no chess expert, but the question brings up two informative websites.
0 kommentar(er)
