Author: Robert Hyatt
Date: 17:50:53 08/11/05
Go up one level in this thread
On August 11, 2005 at 15:30:25, Peter Berger wrote: >On August 11, 2005 at 13:53:37, Robert Hyatt wrote: > >>That will be a miracle. A single opteron CPU (single core) is 2x as fast as a >>3.0ghz pentium. I, at least, will be running on 8 of 'em. That is a tough >>disadvantage to overcome... > >The question is how much this amounts to. Let's call the hardware 16 times >faster, how much faster will Crafty be when running on both of them? So far I >have only seen NPS comparisons reported but no time-to-ply comparisons. Do you >think you will get more than 10 times faster performance out of the faster >hardware? > >With 10 times faster you'd only have sth like maybe 200-250 Elo difference ( >simplified) by hardware. I ran some tests last night. The data is "rough" as I'd like to re-run the same 25 CB/mchesspro positions I used last year at least 16 times. The rough data was comparing the one cpu 875 (one physical process) to 8 cpus (8 physical processes). The average speedup was around 6.0, after yesterday's marathon tuning activities. So I suppose that factors into about 12X faster than a 3.0ghz xeon, if the math holds. Again, note that that 6.0 number is from one run of a batch of positions, just to get a feeel for what the tuning I had done was doing to the search. It is actually possible to do a bit better in middlegame positions, by different tuning, but that tends to impact endgame searches. The current tuning is set to achieve a reasonable balance on all positions... Sorry I don't have the kind of data I gave to Martin last year. But I will certainly rerun that same test again for 1, 2, 4 and 8 processors as soon as the WCCC is over, then I can supply some real data over multiple runs...
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.