Author: Robert Hyatt
Date: 09:36:58 12/04/03
Go up one level in this thread
On December 04, 2003 at 12:33:35, Slater Wold wrote: >On December 04, 2003 at 12:20:08, Robert Hyatt wrote: > >>On December 04, 2003 at 12:09:02, Slater Wold wrote: >> >>>Please take into consideration that Bob is *NOT* done with the 64-bit Crafty, >>>and these numbers aren't 'solid'. >> >>Let me add that these are positions that misbehave for various reasons, >>which is what I think originally caught your eye. I would not particularly >>use WAC for a speedup test myself, I have a >>set of positions I normally run that includes a few tactical positions a >>few quiet positions, a few opening positions and a few endgame positions... >> >>Not that that is a better test, but it does have a bit more diversity. > > >Well, the Opteron + 19.7 does a LOT better now than before...at least on the 2x >level. What I call "bad" positions are not just those that produce oddly low speedups, but those that produce speedups with significant variance. Good positions are relatively stable, although even they have some variance. But not that 16x stuff. :) All this probably means is that there are now other positions that are worse than these. :) > > >>>I ran a test a while back, on a dual AMD 1.73Ghz and two specific positions >>>where speedup is pretty ugly (or at least, odd). You can read the original >>>thread here: http://www.chess-archive.com/ccc.php?art_id=250157. >>> >>>My original results were: >>> >>>WAC.41 >>>1k6/5RP1/1P6/1K6/6r1/8/8/8 w - - >>> >>>NPS using 1 CPU: 1503k >>>NPS using 2 CPUs: 1786k (1.19x) >>> >>>11th ply 1 CPU: 48.36s >>>11th ply 2 CPUs: 2.99s (16.17x - LOL) >>> >>>WAC.76 >>>r1b1qrk1/2p2ppp/pb1pnn2/1p2pNB1/3PP3/1BP5/PP2QPPP/RN1R2K1 w - - >>> >>>NPS using 1 CPU: 804k >>>NPS using 2 CPUs: 1306k (1.62x) >>> >>>11th ply 1 CPU: 19.39s >>>11th ply 2 CPUs: 13.42s (1.44x) >>> >>> >>>And now the Opteron results: >>> >>>WAC.41 >>>1k6/5RP1/1P6/1K6/6r1/8/8/8 w - - >>> >>>NPS using 1 CPU: 2480k >>>NPS using 2 CPUs: 4687k (1.89x) >>>NPS using 4 CPUs: 8660k (3.49x) >>> >>>11th ply 1 CPU: 1.72s >>>11th ply 2 CPUs: 0.93s (1.85x) >>>11th ply 4 CPUs: 0.69s (2.49x) >>> >>>12th ply 1 CPU: 4.74s >>>12th ply 2 CPUs: 2.56s (1.85x) >>>12th ply 4 CPUs: 1.71s (2.77x) >>> >>> >>>WAC.76 >>>r1b1qrk1/2p2ppp/pb1pnn2/1p2pNB1/3PP3/1BP5/PP2QPPP/RN1R2K1 w - - >>> >>>NPS using 1 CPU: 1573k >>>NPS using 2 CPUs: 3046k (1.94x) >>>NPS using 4 CPUs: 5902k (3.75x) >>> >>>11th ply 1 CPU: 10.61s >>>11th ply 2 CPUs: 5.76s (1.84x) >>>11th ply 4 CPUs: 2.83s (3.75x) >>> >>>12th ply 1 CPU: 57.85s >>>12th ply 2 CPUs: 35.24s (1.64x) >>>12th ply 4 CPUs: 25.83s (2.23x)
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.