Author: Robert Hyatt
Date: 07:59:28 07/17/04
Go up one level in this thread
On July 17, 2004 at 06:23:55, Peter Berger wrote: >On July 16, 2004 at 22:07:38, Robert Hyatt wrote: > >>Right. But don't forget that you are looking at NPS. > >No I didn't. But it really doesn't matter enough to continue discussion. If you didn't, I don't see how the quad could be 4x faster, unless you look at just one position and you happen to pick a "good one" for the parallel search. Certainly some positions will be 4x faster or even better, but an average set should settle in around 3x +/- .2 or so.. I was running the SMP test on the 4x2.4 machine last night, since it is still up probably until Monday. It was almost exactly the same speed as the 4 x 2.2 used in CCT6 because of the 10% loss in not having all the NUMA stuff working correctly for the WCCC. IE our box, 4 cpus, would be equivalent to 3.1 FX-51 processors, roughly. About 2.7 FX53 boxes... And if the single-cpu programs used a better compiler (ie windows and ms C) then there is at least another 10% advantage to the windows boxes based on compiler optimization advantages. > > That is not parallel >>search efficiency. NPS might be 4x. Actual search will be maybe 3.1X faster, >>or something pretty close to that...
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.