Author: Uri Blass
Date: 01:05:49 05/11/04
Go up one level in this thread
On May 11, 2004 at 03:32:25, Slater Wold wrote: >On May 11, 2004 at 02:06:46, Uri Blass wrote: > >>I plan to buy new hardware for WCCC and my question is which hardware is best >>for movei. >> >>If people can run movei in the opening position and tell me nodes per second >>then it may be productive for me in deciding about buying new hardware. >> >>I want to know nodes per second that I can get in the opening position >> >>I am interested in the result that I can get in different hardwares. >> >>I do not plan to implement parallel search but having 2 processors have the >>advantage that I can test with pondering and get results that I can trust so I >>did not decide if to buy a machine with one processor or 2 processors and if the >>difference in speed between one processor and 2 processor is small enough then I >>may decide to go for 2 processors. >> >>Uri > >AMD FX53 (2.4Ghz): > >Movei00_8_178 WB2 17 MB: > 1 00:00 0.35 b1c3 > 2 00:00 0.09 b1c3 b8c6 > 2 00:00 0.10 d2d4 > 2 00:00 0.16 d2d4 b8c6 > 2 00:00 0.17 e2e4 > 2 00:00 0.21 e2e4 b8c6 > 3 00:00 0.40 e2e4 e7e5 f1c4 > 4 00:00 0.06 e2e4 e7e5 d1h5 b8c6 > 5 00:00 0.34 e2e4 d7d5 b1c3 g8f6 e4d5 f6d5 f1b5 c7c6 c3d5 d8d5 > 6 00:00 0.10 e2e4 e7e5 b1c3 f8c5 d1g4 g7g5 > 7 00:00 0.12 e2e4 d7d5 e4e5 b8c6 d2d4 c8f5 b1c3 > 7 00:00 0.13 d2d4 > 7 00:00 0.30 d2d4 e7e6 c1f4 b8c6 g1f3 f8b4 c2c3 b4e7 > 8 00:01 0.09 d2d4 g8f6 b1c3 d7d5 e2e3 c8g4 f1e2 g4f5 > 9 00:01 0.11 d2d4 g8f6 e2e3 e7e6 g1f3 b8c6 f1b5 f8b4 c1d2 b4d6 > 9 00:01 0.12 g1f3 > 9 00:01 0.11 d2d4 g8f6 e2e3 e7e6 g1f3 b8c6 f1b5 f8b4 c1d2 b4d6 >10 00:02 0.10 d2d4 g8f6 b1c3 e7e6 g1f3 f8b4 c1g5 h7h6 g5f4 b8c6 >10 00:03 0.11 b1c3 >10 00:04 0.18 b1c3 b8c6 d2d4 e7e6 g1f3 f8b4 e2e4 d7d5 c1g5 g8e7 >11 00:06 0.15 b1c3 d7d5 d2d4 g8f6 g1f3 e7e6 c1f4 f8b4 e2e3 b4c3 b2c3 b8c6 >11 00:07 0.16 d2d4 >11 00:12 0.19 d2d4 g8f6 g1f3 e7e6 e2e3 b8c6 f1b5 a7a6 b5c6 d7c6 b1c3 >12 00:24 0.11 d2d4 g8f6 g1f3 e7e6 e2e3 d7d5 b1c3 f8b4 f3e5 e8g8 a2a3 b4a5 > >676,440 N/sec > >I would venture to say optimized and profiled with Whidbey, it'd be over 700k. > >Send me the source, and I'll tell you what it runs with Whidbey 32-bit & 64-bit. > ;) Thanks for the information. It is more than 3 times faster than the best hardware that I have today and it is also more than 50% faster than the hardware that was used in the last CCT tournament. I am sure that it is possible to get speed improvement from better compiler. I know what I did not do and I did not do profile based optimization(I have only microsoft visual C++6 and I do not have newer compiler and it is impossible to do profile based optimization or I do not know how to do it with the compiler that I have). I will consider if to send the source. I will not make it public in the near future in any case but I may decide to send it to somebody to get some speed improvement before the tournament. I plan to release a new version in the next week. no promise for big improvement but I plan to give the user options to change personalities by changing weights in the program and I also hope to implement a book that knows about tranposition but no promise about it. Another reason to release a new version is that I corrected a bug that can cause movei to lose games on time in rare cases. 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.