Author: Mike Byrne
Date: 07:40:20 12/20/02
Go up one level in this thread
On December 20, 2002 at 08:38:32, Bob Durrett wrote: >On December 20, 2002 at 03:47:40, Aaron Gordon wrote: > >>On December 19, 2002 at 21:30:00, Bob Durrett wrote: >> >>>On December 19, 2002 at 19:28:24, Bob Durrett wrote: >>> >>>> >>>>Recently, there was a chart titled "Crafty v18.11 Benchmark" presented here. It >>>>showed Kn/s [what Hyatt calls "raw Kn/s"?] for various computers. The absolute >>>>worst computer, a P4-1.5GHz, had more than 500 Kn/s. >>> >>>><snip> >>> >>>>Bob D. >>> >>>It appears that the size of the hash table has a big impact on Kn/s in the case >>>of Fritz 8. What about Crafty v18.11? What hash table size is used to obtain >>>the data for the Crafty v18.11 Benchmark testing? >>> >>>We already talked about the impact of the choice of position. The way that >>>problem is addressed in the test was explained by Bob Hyatt. But no information >>>yet on the hash table size used in that test. Are several different hash table >>>sizes used and then an average Kn/s calculated? >>> >>>Bob D. >> >>When I ran the Crafty v18.11 tests I used no RC file, default hash size for >>Crafty. Hash = 3mb & pawnhash = 640k. > >Having separate Hash and pawnhash makes life more difficult since Fritz seems to >have only one Hash size that the user can adjust. How does one compare the two >engines insofar as hash is concerned? It's almost like trying to compare apples >and oranges. > >Confused, > >Bob D. Benchmarks are not for comparing different programs against each other on the same machine, it is for comparing the exact same program with exact same settings on different machines. The purpose of the benchmark is to assist you is setting your machine up to properly maximize your chess program full potential or to assist you in deciding what machine you may want to acquire for your favorite program.
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.