Author: Mark Young
Date: 11:23:38 05/29/98
Go up one level in this thread
On May 29, 1998 at 09:10:26, Enrique Irazoqui wrote: >On May 29, 1998 at 07:34:34, Mark Young wrote: > >>Here is some more data for people to chew on. Tests ran on a P II 300 64 >>meg sdram. You will notice a very uneven upward trend, as hash is added. >>This happend no matter how many time I ran this profile. This may be >>caused by Windows 95, and not fritz 5. When I ran this profile on win 98 >>beta with the same computer. I got more of a steady upward trend. Maybe >>a chess programmer could explain this. >> >>Hash size Fritzmark Time >>64kb 50 445.8s >>1meg 99 226.3s >>2meg 112 201.1s >>3meg 119 190.5s >>4meg 133 169.4s >>5meg 139 162.1s >>6meg 145 156.6s >>7meg 152 148.8s >>8meg 151 149.2s >>9meg 160 141.2s >>10meg 161 140.6s >>11meg 163 138.0s >>12meg 167 135.7s >>13meg 172 131.8s >>14meg 179 126.6s >>15meg 165 137.5s >>16meg 189 119.1s >>17meg 179 126.7s >>18meg 170 133.6s >>19meg 193 117.7s >>20meg 185 122.7s >>21meg 191 118.3s >>22meg 205 110.1s >>23meg 202 111.4s >>24meg 176 128.0s >>25meg 206 110.5s >>26meg 202 112.7s >>27meg 206 110.5s >>28meg 207 109.1s >>29meg 196 115.2s >>30meg 184 122.2s >>31meg 210 107.2s >>32meg 187 120.4s >>33meg 198 114.9s >>34meg 201 112.9s >>35meg 199 113.4s >>36meg 200 113.0s >>37meg 207 109.8s >>38meg 198 114.1s >>39meg 223 101.9s >>40meg 209 108.7s >>41meg 226 100.0s >>42meg 205 110.0s >>43meg 197 115.7s >>44meg 202 111.4s >>45meg 206 109.4s >>46meg 223 101.9s >>47meg 206 110.6s >>48meg 232 97.7s >>49meg 199 113.3s >>50meg 207 109.9s >>51meg 205 110.8s >>52meg 210 108.6s >>53meg 228 99.9s >>54meg 204 110.4s > >I don't think the long Fritzmark gives an accurate idea on faster >machines if what you want to measure is performance at 40 in 2. That's >why on your table there is no difference between 22MB and 54MB. You need >a position that takes about 2 to 3 times as long to solve. > >By the way, how did you manage to get 54 MB on a 64MB machine? My >maximum is 50, and with lots of swapping. :( > >Enrique ----------------------------------------- Thats not what I was trying to do. Just show the increase in speed as you add hash. People can make what they want out of the data. There is a increses in speed at certain hash settings. I found that on the fritzmark it liked certain hash settings better. Likes 53 meg, 48 meg, 41 meg, 39 meg ect. This is not random. It will repeat no matter how many time I run the test. Now in windows 98 this stopped. and say after 39 meg. each values as you added hash will gave you a better time. And all times were in the 220 to 230+ fritzmark range with a smooth upward tread as you add more hash. So this is not true. I don't know why I could go higher on my hash settings. I do know with windows 98 beta I could go even higher. Up to 56 meg. I wanted to show that fritz could still think fast with a small amount of hash. It did not stop or do little as the manule says. Look at what 2megs gives you. And that amount of hash is filled in secs. So I do think it is worth doing over night analysis with fritz 5. Even though It will fill up its hash tables of 50 to 100 meg in 2 to 4 min.
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.