Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: More TSCP nps results

Author: James Swafford

Date: 20:23:16 07/24/99

Go up one level in this thread


On July 24, 1999 at 23:03:26, Will Singleton wrote:

>On July 24, 1999 at 22:45:54, Peter Kappler wrote:
>
>>On July 24, 1999 at 21:42:56, Will Singleton wrote:
>>
>>>Latest results for the TSCP nps test:
>>>
>>>Platform      Compiler       nps     Dann's build
>>>
>>>P3/500        msvc opt       63,246      x
>>>K6-2/400      msvc opt       57,163      x
>>>Cel/450       msvc opt       56,075
>>>K6-3/450      msvc opt       53,634      x
>>>P2/450        msvc opt       49,338      x
>>>K6-2/350      msvc opt       45,679      x
>>>Mac G3/300    CW 3.1 opt     40,519
>>>P2/300        msvc opt       38,223      x
>>>K6-2/266      msvc opt       38,159
>>>K6-2/400      cygwin opt     37,516
>>>K6-2/350      borland opt    32,638
>>>K6-2/350      symantec opt   32,168
>>>P2/300        gcc opt        29,205
>>>K6-2/400      msvc no opt    28,745
>>>Cyrix 300     msvc opt       25,345      x
>>>K6-2/400      cygwin no opt  23,885
>>>P6/200        msvc opt       23,581      x
>>>Cyrix 200     msvc opt       20,398      x
>>>K5/116        gcc -0         17,200
>>>P6/200        watcom opt     16,018
>>>
>>>
>>>Note that the fast K6-2/400 result looks a bit out of place.  Perhaps a mistake?
>>>
>>>The other result that jumps out is that the Mac G3/300 is NOT faster than
>>>comparable or higher mhz machines when those machines run the MSVC compiled
>>>version.  Is the MSVC compiler really that much better than any other?
>>>
>>>Will
>>
>>
>>I was just discussing the K6-2/400 result with James Swafford.   Neither of us
>>can make any sense of it.  We both ran the Crafty benchmark, and my K6-3/450 was
>>34% faster than his K6-2/400.
>>
>>Perhaps everybody could re-run the TSCP test, and search a bit deeper, say 7 or
>>8 plies.  The problem with 6 is that the test is over in 3-4 seconds - I think
>>we will get more reliable results with a longer test...
>>
>>--Peter
>
>I think ply 6 is deep enough; the node count is high enough (203k) to give
>accurate timing results.
>
>Does NT report clock time the same as Win 98?
>
>Will

It is not a mistake.  I ran it twice.  The second run was slightly
better than the first.  I'm not sure about the timing differences
between NT/98, but I agree with Pete - I'd like to rerun the test
with a 7 ply search.  Dann?  :-)

--
James




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.