Author: Robert Hyatt
Date: 13:11:40 01/08/98
Go up one level in this thread
On January 08, 1998 at 04:56:15, Kai Lübke wrote: >On January 08, 1998 at 01:42:58, Jouni Uski wrote: > >>Have anybody found good way to use TB dynamical? In my PC node speed >>goes to 1/10 of normal, even if all TBs are moved to hard disk! So >>it seems to be no sense to dynamical TB access. Is there special >>hardware requirements (50 MB cache) to succeed or what? > >I have noticed some heavy slowdowns as well, but only occasionally. >Typically, I get around 22000-28000 nps (P6-200) running with 60 MB hash >under DOS 6.22. With heavy tablebase probing, this goes down to 8000 nps >sometimes, on rare occasions to 3000 nps (but only for a few seconds). >However, this may be due to the fact that I run it under DOS 6.22 - >Crafty is also _very_ slow in tablebase access there (as low as 12000 >nps compared to 40000 nps under DOS 7.0 minimum, 55000 nps typical with >TB probing). >Reason is probably the 16 bit disk access? > >Under Win95, I haven't noticed nps counts less than 15000, which is more >than 50% of the usual speed. > >(All this is with the full 3/4 man set and a fast EIDE harddisk) > >--- >Shep with 4 piece files, I have *never* seen crafty's NPS drop below 50% of normal speeds, but I use a P6/200 with wide SCSI disks and am running Linux. with 5 piece files, on rare occasions I see speeds down to 1/4 or normal, but that is very rare. typical is still 50% of normal... I haven't run under dos or windows, and have no clue as to how bad or good their random file access is. but this is truly random I/O so it's got to be efficient or something will die... ]
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.