Author: Uri Blass
Date: 03:55:34 03/06/03
Go up one level in this thread
On March 06, 2003 at 06:38:15, Chessfun wrote: >On March 06, 2003 at 02:39:42, Uri Blass wrote: > >>On March 06, 2003 at 01:50:29, Mike Byrne wrote: >> >>>On March 06, 2003 at 01:01:32, Uri Blass wrote: >>> >>>>On March 05, 2003 at 22:37:56, Mike Byrne wrote: >>>> >>>>>On March 05, 2003 at 22:24:45, bobby palacios wrote: >>>>> >>>>>>On March 05, 2003 at 20:23:07, Mike Byrne wrote: >>>>>> >>>>>>>On March 05, 2003 at 19:56:30, Mike Byrne wrote: >>>>>>> >>>>>>>>Ruffian vs Crafty 19.03 "x", Blitz: Game /5 >>>>>>>> >>>>>>>>Dual 1.8 Ghz Xeon >>>>>>>>Ponder On >>>>>>>>Unified Hiarcs Tourn book >>>>>>>> >>>>>>>>Crafty 19.03 "x" +62 -53 =69 52.45% 96.5/184 >>>>>>>>Ruffian 1.0.1 +53 -62 =69 47.55% 87.5/184 >>>>>>>> >>>>>>>>This is clearly the best result a modified Crafty (for me) has ever obtained >>>>>>>>against Ruffian at Game/5. >>>>>>> >>>>>>>This is the last game played - Ruffian has the advantage and lost it. >>>>>>> >>>>>>> >>>>>>> >>>>>>>Ruffian 1.0.1 - Crafty 19.03 'x' [B33] >>>>>>> >>>>>>>W=11.2 ply; 512kN/s; 959 TBAs >>>>>>>B=10.9 ply; 748kN/s >>>>>> >>>>>>This is a bit strange. On my p4 2.47ghz, Ruffian generally outsearches Crafty >>>>>>19.03 by about 100-200kn/s, but certain engines run better on certain machines, >>>>>>so maybe this is expected. >>>>> >>>>>Since I compiled Crafty on my machine using profiled optimizations, it would be >>>>>highly optimized for my machine. >>>>> >>>>>:)))))) >>>> >>>>I am surprised by the number of nodes of Ruffian. >>>> >>>>Even on my slow P850 Ruffian can show more than 512 Kn/seconds >>>> >>>>Here is some analysis of one of the position of the game by Ruffian. >>>>depth=12 1/48(Rh5) +0.96 Rh5 Rc3 Qxc3 Qxh5 g4 Qg5 Qa3 h5 h3 Qh4 Qxd6 Qxh3 Qxa6 >>>>hxg4 Qxb5 >>>>Nodes: 34277608 NPS: 548266 >>>>Time: 00:01:02.52 >>>> >>>>Why did Ruffian search only 512 knodes per second on 1.8Ghz? >>>>Are you sure that Ruffian used 100% of the time of your second processor? >>>> >>>>Uri >>> >>>Yes - I do checked that frequently to make sure no are problems. My p4 behaves >>>more like a P3 1000(?guess) for many chess programs that are not optimized >>>specifically for the P4. >> >>How much speed can programs earn from optimizing them for p4? >> >>Do you get more than 50% speed improvement from your optimizations? >> >>748 nps is almost 50% more than 512 nps and I read in this thread that a person >>with pentium4 who did not do special optimizations said that Ruffian is faster >>than crafty on p4. >> >>Uri > >His average NPS was 512 k/Ns, this don't mean on a given position he would get >that. Plus you never posted the FEN for him to check. > >In this case Ruffian shows 959 TBAS, this itself slows down NPS in some cases by >a factor of 6 or 7. So to simply say Ruffian only did 512 k/Ns and I can get >more on my PC something is wrong, is itself wrong. > >Sarah. I did not say that I can get more average nps but the position that I analyzed was a middls game position from the game(I did not save it but I guess that there are not many positions from the game when the pv make sense so it is probably possible to find out). I thought that in endgames the search is faster thanks to smaller number of pieces(it is at least the case with movei). If tablebases make the program slower by a factor of 6 or 7 then something is probably wrong with the program. I also thought that his 1.8gh should be at least twice faster than my p850 (I was probably wrong in this thought). I thought that all this data probaly suggests that something may be wrong in the game. Uri
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.