Author: Robert Hyatt
Date: 12:26:55 11/20/01
Go up one level in this thread
On November 20, 2001 at 10:26:40, Gordon Rattray wrote: > >If a dual machine is used to run two different chess programs (non-SMP), each >using a single processor to itself, will this be equivalent to having two >separate single processor PCs (assuming the processors were the same)? Or does >the use of a shared memory affect the performance? single bus?! There is a cost. But since it is equally spread over both processors, you can assume that the match is equal in that regard. > I'm also >assuming that the dual machine would have the combined memory capacity of the >two single PCs, so hash table sizes would be the same. > >For the dual, what if both programs were accessing endgame table bases from the >same disc? Both programs analyising the same position, so maybe cache hits?! >Different positions would cause more disk access?! Yes. But again it should be pretty equal, unless one program intentionally probes less than the other (or more). > >Can anyone give an opinion on the efficiency of crafy/fritz/junior/shredder (SMP >versions) when running on a dual? Is 80% a fair estimate? Can it vary >significantly depending on position? Crafty will generally _average_ about 1.7X faster on the dual than on a single cpu at the same clock frequency. It will be faster on many positions (closer to 2.0 at times) but the average over a wide range should be 1.7. I have no idea how efficient Fritz and the others are. I once heard Frans answer that question with an answer something like "mine does pretty well, not as efficient as crafty but it does well." Best way to answer that is to test them yourself... test on a set of positions using one cpu and then two, to a fixed depth, so see how much faster the dual performs... > >thanks in advance for any answers > >Gordon
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.