Author: Robert Hyatt
Date: 11:16:41 05/19/04
Go up one level in this thread
By the way, I do have hard copy for most of my DTS data. The tests were all run using the pure FORTRAN version of Cray Blitz, which was highly vectorized for the Cray, but which ran very slowly on the Sequent (We also had a lot of bitboard stuff (64 bit) that was horribly slow on the sequent). Typical search speeds on one cpu for the 24 kopec positions look like this: 22 92 78 44 47 92 85 83 80 81 82 50 100 54 57 59 78 60 82 45 53 79 82 96 Yes, those are actual "nodes per second" using one processor. I also did verify that the searches were all done to a depth of 5 plies, period. For comparison, here are the 1-cpu times... 0:18 15:48 6:17 32:56 54:36 2:24 9:51 1:18 12:20 19:47 11:29 66:48 11:22 16:49 15:36 20:58 12:28 33:27 12:32 41:05 40:01 14:13 10:11 12:46 Care to speculate how long it would take to search them all to a depth of 10 plies today? We don't need to speculate. I just ran it here, using one xeon 2.8ghz processor, with null-move R=1 to come close to matching what I was doing in 1988 according to the output back then (R=1, but non-recursive compared to recursive null-move today). The above Sequent run took about 450 minutes total. Today's xeon takes 4:40, less than 5 minutes to search 2x deeper (10 plies vs 5 plies). To run the test to a depth of 5 plies takes under 1 second on the xeon. So is it any _real_ surprise that the C90 did better than the Sequent? The C90 was searching about 500K per second. My single xeon is about 2x that. That's what convinces me you have absolutely no business working in any sort of scientific endeavor. You have absolutely no ability to look at something, and by studying it, drawing meaningful conclusions based on those observations. You just wave your hands and spew garbage... Again, "lookin' good, Vincent... " For those that are getting tired of looking at real data all the time, I'd suggest you bypass my posts and read only Vincent's. That way you don't have to look at real data at all...
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.