Author: Kirill Kryukov
Date: 17:29:30 11/07/05
Go up one level in this thread
Hi Ray, >I've run thosands of ponder-on games with tablebases in a single location, and >haven't noticed anything unsusual. But I don't know exactly what to look for >that would constitute a problem. Sure there will be some contention, but with >fast hard drives today I just don't think it is a significant issue. >One suggestion here was to use two hard disks in a raid configuration. That >appears as a single disk but physicaly two, hopefully the tablebases would >physically be spread across the 2 disks when you copied them on, thereby >avoiding some of the access issues I support the RAID idea. RAID-0 configuration is the best for chess - everything is spread across two disks and all accesses are much faster. In this particular case (ponder on match) I think RAID-0 will work even better than two independent disks, since when two engines try to access the tablebase it will likely be the same tablebase anyway (Or the same set of tablebases). I personally don't like "ponder on" testing because it introduces a random factor, thus reducing reliability of the results. The random factor is whether or not the expected move was played, this will affect effective time spent on thinking about the move. In ponder off mode, on the other hand, all time spent on thinking is controlled by the engine, which should give more consistent play. Best, Kirill
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.