Author: Dann Corbit
Date: 17:55:40 11/25/03
Go up one level in this thread
On November 25, 2003 at 17:47:33, Uri Blass wrote: >On November 25, 2003 at 17:27:27, Drexel,Michael wrote: > >>On November 25, 2003 at 16:45:29, Uri Blass wrote: >> >>>On November 25, 2003 at 15:18:37, Drexel,Michael wrote: >>> >>>>On November 25, 2003 at 15:00:35, Joachim Rang wrote: >>>> >>>>>On November 25, 2003 at 14:28:55, Drexel,Michael wrote: >>>>> >>>>>>On November 25, 2003 at 14:03:45, Joachim Rang wrote: >>>>>> >>>>>>>On November 25, 2003 at 13:58:51, Jorge Pichard wrote: >>>>>>> >>>>>>>>Standing after round 5: >>>>>>>> >>>>>>>>WCCC 2003 >>>>>>>> >>>>>>>>1 Brutus 4.0/5 >>>>>>>>2 Shredder 4.0/5 >>>>>>>>3 Quest 3.5/5 >>>>>>>>4 Deep Junior 3.5/5 >>>>>>>>5 Diep 3.0/5 >>>>>>>>6 Green Light 3.0/5 >>>>>>>>7 Falcon 2.5/5 >>>>>>>>8 List 3.0/5 >>>>>>>>9 Deep Sjeng 2.5/5 >>>>>>>>10 Jonny 2.5/5 >>>>>>>>11 Chinito 2.0/5 >>>>>>>>12 Quark 2.0/5 >>>>>>>>13 ParSOS 1.5/5 >>>>>>>>14 Ruy Lopez 1.0/5 >>>>>>>>15 Nexus 1.0/5 >>>>>>>>16 Hossa 1.0/5 >>>>>>> >>>>>>> >>>>>>>Junior has to play against Shredder AND Brutus. >>>>>> >>>>>>>Quest already played all >>>>>>favorites. >>>>>> >>>>>>Are Deep Junior and Diep not favorites? >>>>>> >>>>>>Michael >>>>>> >>>>>>>It is _very_ interesting I do not dare to predict something. >>>>>>> >>>>>>>regards Joachim >>>>> >>>>> >>>>>Deep Junior is of course a favorite, therefor I mentioned, that it has yet to >>>>>play against Shredder and Brutus. >>>> >>>>...and Quest :) >>>> >>>>> >>>>>Despite the 500 proc diep is for me not a favorite and therefore I didn't >>>>>include it in the fantastic 4. >>>> >>>>A thing on 500 processors can beat any program. (, with better opening lines ;) >>>>) >>>> >>>>Port Crafty to that supercomputer and it might beat the hell out of Quest >>>>on 4*2.8 Ghz in a long match. >>> >>>I am surprised that you expect Crafty on a super computer to beat Quest on >>>4*2.8Ghz. >>> >>>You predicted that Crafty with 10 times faster hardware is going to lose against >>>Junior when the programs use their books and the hardware of Junior is clearly >>>slower than 4x2.8 GHz. >> >>Crafty on 512 500MHz processors is much more than 10 times faster than Crafty on >>4x2.8 GHz. >> >>Michael > >I am not sure about it. > >I do not know the speed improvement from 512 processors and it may be only 100 >times faster than one 500 Mhz processor when 4*2.8 Ghz is more than 10 times >faster than one processor of 500 Mhz. There may be some other effects that come into play at that scale. For instance, sometimes SMP searches will "get lucky". Now, we know that 'luck' is not really what is involved. Rather, a bit of randomness sometimes makes it find a good solution faster than it ought to. Perhaps with 512 CPUs, that effect will magnify. So you might get only 100x speedup (in NPS) but you might get effectively 400x solution improvement. (Or the effect might even be reversed so that it actually slows down as you add processors -- no way to know without testing). I certainly think it would be an interesting experiment. Having a NUMA architecture sure lifts the lid off of the ceiling of what you can try, though.
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.