Author: Robert Hyatt
Date: 12:50:16 02/06/06
Go up one level in this thread
On February 06, 2006 at 15:38:56, chandler yergin wrote: >On February 06, 2006 at 15:03:08, Robert Hyatt wrote: > >>You keep providing stuff that has _nothing_ to do with the topic being >>discussed. >> >>Here is a test I ran just now, to show you how silly your statements look: >> >>I varied the hash size with Crafty as follows: 3M, 6M, 12M, 24M, 48M, 96M, 192M, >>384M >> >>I ran a normal middlegame position to depth=13, using just one cpu to avoid any >>SMP variance. Here's the times/nodes: >> >>log.002: time=4:45 mat=0 n=247340776 fh=89% nps=865K >>log.003: time=3:52 mat=0 n=201952429 fh=89% nps=867K >>log.004: time=2:52 mat=0 n=150627132 fh=89% nps=870K >>log.005: time=3:00 mat=0 n=157180245 fh=90% nps=870K >>log.006: time=2:53 mat=0 n=152600033 fh=90% nps=877K >>log.007: time=2:27 mat=0 n=128306131 fh=90% nps=870K >>log.008: time=2:27 mat=0 n=127485719 fh=90% nps=863K >>log.009: time=2:22 mat=0 n=124561460 fh=90% nps=872K >> >>3M is the default. I'd personally want to use something on the upper end of the >>hash sizes here. Or would you _really_ want to give up a factor of 2x in speed >>and use the default? (4:45 / 2:22 == 2.0) >> >>So as you can see, your facts are wrong, your argument is invalid, and your >>conclusions are based on who-knows-what... >> >>Now, for the last time, stop this before you get stopped. >Thank you, now a logical proposition for you. >OK? >Post the position and let us see the Depth of search, total nodes searches >and the Quality of the analysis from the Position. >k/N/s is the key to speed in Depth of search. >nps at various time controls mean nothing, it's the "Depth" of search >You know that! >cy Did you read what I wrote. _every_ search above was to a fixed depth of 13 plies. Every last one. here is the position: r3kb1r/pppq2pp/1n2b3/8/1P1Pp3/2N1B2N/1P3PPP/R2QK2R w KQkq - The nodes and time are given above. time=xxx and n=xxx are the relevant numbers. The search chose the _same_ move each time. So I don't see what your point could possibly be, other than to show the world exactly how _little_ you understand about this subject. You are doing an excellent job of that, btw. To explain this where even you can understand: 1. set position to the above FEN string 2. set hash size to the value you want to test, I used 3M, 6M, ..., 384M for my tests. 3. set search depth to 13 (you might change this to something that makes your engine of choice search for about 3-4 minutes). 4. type "go" or whatever needed to make the engine do the above search. 5. when it finishes depth 13, write down the time (and nodes if you are interested, but time is the important number here). 6. Repeat for each different hash size. 7. Decide which hash size produced the shortest total time used. Now if you can't accomplish that simple task, go away permanently. These discussions are basically pointless.
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.