TEAMANANDTECH Chess Brain 1-14-04 9:29 (GMT)

[sciencewhiz]

Yes, the high memory setting does cause problems with the stable client. It "hangs" causing Chessbrain to use 100% of the idle power, but not actually process any position. I did not hear of Carlos fixing it, but I also haven't heard it reported in a while, so its more then likely been fixed in the experimental clients.

Using more memory only gives small improvements in speed. The memory is used as a cache for positions. The problem is, there are many more positions then any computer could possibly cache, so there is only a small chance that your position will hit, in the cache. For this reason, there is little difference what memory setting you choose.

 

[BCinSC]

How about multiple instances per machine? Am I gaining or losing here? Would it cause higher abort rate?

 

[D347h]

Originally posted by: BCinSC
How about multiple instances per machine? Am I gaining or losing here? Would it cause higher abort rate?

I don't think so. The first one who complete move and send it back get points, the rest get info - Aborting this WU - too slow... So, how running 2 instances on one PC could help ??

BTW, I have 256 Mb RAM but still use LOW option.. I think it doesn't change anything...

 

[BCinSC]

The different instances get different jobs, or so it seemed. If there's only one move in a given moment and 1000 instances on tap, then ROI is very low.

 

[Coolkid]

Well im experimenting with running 2 instances on my main rig. With 2, my aborted rate had gone up slightly, its processing twice as much work, but SETI also takes more of a hit