Author: Shep
Date: 05:00:09 08/13/99
Go up one level in this thread
On August 13, 1999 at 05:39:46, Terry Ripple wrote: >I'am using a AMD K6-2/266Mhz with 64 Ram & Hash Table size = 44Mb. > >I find that the Hi-Mark does not work properly in the Fritz interface because it >shows my best Hi-Mark for the long test at 195 Mark with only 18Mb Hash and as i >increase the Hash size the Hi-Mark decreases. I get only 185 Hi-Mark when i >increase Hash to 44Mb but the KN/S remains the same at 23KN/S for both. So, i >find the Hi-Mark not accurate for Hiarcs7.32 in the Fritz interface, because the >less Mb of Hash size the better my Hi-Mark gets in both the Short & Long Tests. I suppose the problem is that a) Hiarcs takes a little time to clear the hash tables before a Hi-Mark run; the larger the tables, the more time it takes, and that time may be counted as "used calculation time" by the Hi-Mark b) slow engines tend to take more "hash maintenance" (or "hash access" or whatever (?)) time with large hash tables; this is no problem at longer time controls, but at the "Blitz" time of the Hi-Mark, it is definitely an issue if you spend an extra N/10 seconds per ply. Remember the advice to keep hash tables small for Blitz games frequently given here! >I will only show the KN/S for the Short and Long Tests with Time control >set for 40/2hrs. The time settings and Hash size both affect the test! Time settings as well?? I would find that hard to believe since the *-Marks are all run at fixed ply depth. Are you sure this is more than the "natural" fluctuation between single runs (which is, among others, caused by Windows itself). --- Shep
This page took 0.01 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.