Author: Eugene Nalimov
Date: 12:11:40 04/20/01
Go up one level in this thread
On April 20, 2001 at 11:06:34, Vincent Diepeveen wrote: >On April 20, 2001 at 08:48:39, Robert Hyatt wrote: > >>On April 20, 2001 at 08:41:41, Chris Carson wrote: >> >>>Uri, >>> >>>ref: http://www.specbench.org/osg/cpu2000/results/cint2000.html >>>Specint2000: >>> Crafty >>> Peak Peak Ratio >>>P3 1.13GHZ 464 589 >>>P4 1.5GHZ 536 497 >>>AMD 1.33GHZ 539 759 >>> >>>For Crafty the AMD was best then the P3 followed by the P4. However >>>for other programs the P4 was superior (thus the Peak overall improvement). >>> >>>However, the P4 was better for the TSCP chess programs: >>> >>>Ref: http://ucsu.colorado.edu/~kerrigat/bench/ >>> TSCP >>> MIPS Benchmark >>>Pentium 4/1700 1891 >>>Pentium 4/1500 1749 >>>Athlon/1200 1338 >>>Pentium III/1000 1168 >>> >>>As I read this data for the two programs, it depends on the chess >>>program and other factors (compiler options, design, ...). >>> >>>How do you read this? >>> >>>Any other thoughts/reasons for the descrepancy? >>> >>>Best Regards, >>>Chris Carson >> >>The P4 has a long pipe. And then there is the RDRAM problem. Both hurt >>Crafty because the eval has so many if then else type comparisons. I've >>seen the same sort of performance differences in every case it has been >>tried. > >There is a way simpler explanation besides all those problems. >Biggest difference between TSCP and crafty is that TSCP fits within the >L1 trace cache of ONLY 8192 instructions > >If a program falls outside then it gets dicked, to use an understatement. > >Of course the other factors are also very relevant, but that's all peanut >details compared to this. GCC also does not fit into L1 I-cache of P4. But it's almost 2 times faster on 1.5GHz P4 than on 1.33GHz Athlon. Eugene
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.