Author: Dann Corbit
Date: 09:23:26 04/18/01
Go up one level in this thread
On April 18, 2001 at 10:40:50, James T. Walker wrote: >Seems to me that the "Qualifying" tournament for the right to play Kramnik >should be open to anyone stupid enough to pay $5000 for the right to have his >program crushed by a very well prepared Champion. Where is the >advertising/publicity gain from this? 1. This has similar potential to Deep Blue verses Kasparov and whoever is running the show is a genius. The computer won't have to win the match at all. Remember IBM's first try? They *lost* but still got enormous positive publicity. Now, suppose that an 8 CPU machine plays Kramnik ten games and Kramnik gets 5 wins, three draws and two losses (very realistic possibility). The advertizing headlines will blare: "This is the computer system that beat the best player in the world! TWICE!" So no matter what happens, the equipment manufacturer can turn it into a giant victory. And if Kramnik should get psyched out and freak and somehow lose the whole match, then they will really have something to tout. 2. Now that I have heard all the facts, I understand where the sponsorship is coming from and think it is completely reasonable to demand that multi-processor versions of chess programs must be used. That is only reasonable, because of who is footing the bill. However, I think that other multiple CPU systems should have been allowed to try. Actually, it is a mistake not to. I suspect that crafty or ferret (perhaps Amy as well) would be formidable opposition. Now, a 32 CPU Alpha machine would probably not be allowed in the contest because: A. Not sponsored by Compaq B. It would beat the pants off of the Intel based machines and really make them look bad. Therefore, crafty would have to compete on the same 8 CPU intel system. We must consider economic realities. Nobody is going to pony up 7 figures of cash and then let someone waltz in who is going to clobber them. I do suggest that it would be a good publicity stunt for Compaq to challenge the victor afterwards. A fully loaded Alpha machine with a full complement of memory and latest generation CPU's would be in nearly the same order of magnitude as Deep Blue. (Perhaps 2 million NPS per chip * 32 SMP chips = 64 million NPS, with some loss so perhaps 50 million NPS.) I don't know if Ferret can be easily ported to Alpha, but I understand that Ferret is SMP. I suspect that the commercial programs are optimized with Intel assembly at some points, but perhaps not. In any case, if they did not perform the same port they would probably take it on the chin. Amy also might be a big surprise on that platform. At higher speeds and slower time controls Amy does better and better. On a platform like that, I would be very curious to see how Amy performs.
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.