Author: Vincent Diepeveen
Date: 06:42:32 09/12/01
Go up one level in this thread
On September 12, 2001 at 04:03:00, Bernhard Bauer wrote: >On September 11, 2001 at 10:32:54, Robert Hyatt wrote: > >>On September 11, 2001 at 03:39:52, Bernhard Bauer wrote: >> >>>On September 10, 2001 at 15:53:36, Robert Hyatt wrote: >>> >>>>On September 10, 2001 at 15:46:38, Robert Hyatt wrote: >>>> >>>>>On September 10, 2001 at 13:46:10, Dann Corbit wrote: >>>>> >>>>>>On September 10, 2001 at 13:35:19, Robert Hyatt wrote: >>>>>>[snip] >>>>>>>BTW crafty generally gets 299/300 on a 750mhz machine. If you are using an >>>>>>>800 and getting 295/300 something seems amiss... >>>>>> >>>>>>I only allow 5 seconds per position. If your version can do that on 750 MHz, it >>>>>>will rule the world. >>>>>>;-) >>>>> >>>>> >>>>>OK... didn't see the 5 seconds. I do know of at least one program that >>>>>used to get 'em all at 5 seconds. :) >>>> >>>> >>>>I just checked my log files. A 600mhz 21264 gets 298/300 in 5 seconds, and >>>>gets 299 in 10 seconds, and gets them all in 60 seconds. Single-cpu. Tim Mann >>>>ran the test last year some time. >>>> >>>>nice box. :) >>>> >>>>Actually that 21264 will get 298 in 3 seconds/move >>> >>>Here are some benchmarks from SPEC including the time for the crafty part. >>>Sortet by Crafty-time, Int-peak and Ftp-peak. >>>The Alpha 21264C has a small advantage. >>> SORT >>>SPEC benchmark suite MHz Fp base Fp peak Int base Int peak Crafty time Time / >>>Ghz >>>Compaq Alpha 21264C 1001 585 756 561 621 124 124,12 >>>AMD Athlon 1400 426 458 495 554 126 176,4 >>>AMD Athlon MP 1200 433 481 495 522 144 172,8 >>>Compaq Alpha 21264A 833 571 644 511 533 156 129,95 >>>Dell Pentium 4 1800 615 631 599 619 170 306 >>>Dell Pentium 3 1000 329 340 454 462 191 191 >>> >>> SORT >>>SPEC benchmark suite MHz Fp base Fp peak Int base Int peak Crafty time Time / >>>Ghz >>>Compaq Alpha 21264C 1001 585 756 561 621 124 124,12 >>>Dell Pentium 4 1800 615 631 599 619 170 306 >>>AMD Athlon 1400 426 458 495 554 126 176,4 >>>Compaq Alpha 21264A 833 571 644 511 533 156 129,95 >>>AMD Athlon MP 1200 433 481 495 522 144 172,8 >>>Dell Pentium 3 1000 329 340 454 462 191 191 >>> >>> >>> >>> SORT >>>SPEC benchmark suite MHz Fp base Fp peak Int base Int peak Crafty time Time / >>>Ghz >>>Compaq Alpha 21264C 1001 585 756 561 621 124 124,12 >>>Compaq Alpha 21264A 833 571 644 511 533 156 129,95 >>>Dell Pentium 4 1800 615 631 599 619 170 306 >>>AMD Athlon MP 1200 433 481 495 522 144 172,8 >>>AMD Athlon 1400 426 458 495 554 126 176,4 >>>Dell Pentium 3 1000 329 340 454 462 191 191 >>> >>>BTW testing with WAC will not lead to any progress! >>> >>>Kind regards >>>Bernhard >> >> >>I disagree to your last comment on two grounds: >> >>1. new programs will _definitely_ improve as their WAC scores improve. Getting >>the right extensions to solve these quickly is the fastest way to eliminating >>the ugly tactical blunders likely in newer programs. >> > >I wasn't refering to new programs. For new programs anything may be good. >I was thinking about crafty and how to make it even better. >Crafty has reached a high standard and solves 290+ positions. Improvements for >playing better may not be visible by running test with very short time. > >>2. running WAC after significant changes will _also_ help a lot. If you >>start missing some that you were getting instantly, you have broken something >>important. > >I guess you know wac very well from many years and if you mean that it helps a >lot to see that nothing serios is broken, I agree. > >BTW crafty was always very good in solving the famous Ortueta-Sanz position. >The latest versions do not so well. This shows that you have changed something, >but is anythin´g broken? >Here is the position. >[D]8/pR4pk/1b6/2p5/N1p5/8/PP1r2PP/6K1/ b >Kind regards >Bernhard May i remind that diep solves all tactical wacs, except 1 or 2 endgames which are not important, it doesn't solve. If you can't solve all tactical WAC positions then a program has a major problem as it's a testset about basic tactics.
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.