Author: tony watson
Date: 17:28:26 02/11/00
Go up one level in this thread
On February 11, 2000 at 19:48:09, Dann Corbit wrote: >On February 11, 2000 at 19:43:40, tony watson wrote: > >>Having just overclocked my P166MMX to 200mhz, :>) >>I ran a couple of benchmarks using Crafty's built-in bench command, >>to confirm an increase in speed of around 18%. :>)) >>what I didnt expect was the large speed difference between >>the 2 versions (17.5 and 17.8), so I checked another 3 versions >>and got the following results:- >> >> Version Total Nodes Nodes per Second Time (seconds) >> >> 16.6 74,921,651 63,118 1187 >> 17.5 87,931,266 79,720 1103 >> 17.6 70,362,201 60,138 1170 >> 17.7 72,591,308 61,413 1182 >> 17.8 72,591,308 61,206 1186 >> >> interesting, it appears that the last 3 versions share the same results, >>but 17.5 is 30% quicker, cancelling out the larger number of total >>nodes searched and returning with the best time. >> >>Im sure this will have been discussed before on the forum, but i'm kinda >>new here and I am just interested if the latest version is the finest, or are >>there some real gems hiding in proff. Hyatt's ftp. >> >>All benches were run using identical settings and 16Mb 0f HASH :>) > >You can't tell how strong a program is by counting how many nodes it examines. >Turn off NULL move and it will search more nodes with less effect, for instance. > Remove all positional checking and it will search even more nodes. By a >judicious removal of tests, you could probably double the node counts and halve >the playing strength. > >The proof of the pudding is in the eating. Play a 500 game match between two >versions and tell us how it turns out. Use separate machines, and turn >pondering on. Dont hold your breath then Dann,:>) The time taken to complete the EPD test must surely be of some significance here - or so i thought, but im no expert tony.....
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.