Author: Peter Kappler
Date: 09:50:02 11/14/03
Go up one level in this thread
>> >>Grok averages around 250knps in the middlegame on an Athlon 1400. You >>didn't list your hardware, so I'm not sure how we compare. > >wow, that is much faster than mine. I have a 2500xp Barton, clocked to 3200, and >average probably 150Knps in the mid game. That's about 2.25 times faster than >chompster ... that would be nearly 600Knps on my hardware.... mmmm .... I am not >very impressed with myself. If I had 600Knps, I could take on any engine .... I >can't keep using the - "its java - its slower" excuse. Maybe its bitboards, >maybe its because I generate all moves first, maybe my eval is heavy, maybe I >get better move ordering ... or .... maybe I have sloppy code. > Whose VM are you using? Sun's is 30% slower tham IBM's. Microsoft's is somewhere in between. BEA's is almost as fast as IBM's but seems to be buggy. Another thing to avoid is creating new objects during your search. Pre-allocate whatever you need. Run a profiler against your code to see where you are slow. (Use the -Xprof switch with Sun's VM.) Finally, we might be counting nodes differently. In Grok, any time I update the board with a new move, that's a node. I think that's how most people do it. >I just worked out I'll be Overseas on holidays skiing at Whistler Canada when >cct6 is on. I might need to lean on a few people to make sure chompster enters >the comp. > >Is grok online much of the time? I'd love to play against it, not to compete, >but to see differences. > It plays on ICC occasionally. I'll probably be there this weekend. > >Are you getting a faster CPU for CCT6? > Setting up a new machine is enough of a hassle that I try not to do it unless I can get at least a 2x speedup. Unless I'm misinterpreting the benchmarks on the fx-51, I don't quite see a 2x speedup vs my old Athlon 1400. -Peter
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.