Author: Uri Blass
Date: 02:31:02 06/14/01
Go up one level in this thread
On June 14, 2001 at 03:50:31, Tony Werten wrote: >On June 14, 2001 at 03:14:04, Uri Blass wrote: > >>On June 14, 2001 at 02:55:02, Tony Werten wrote: >> >>>On June 13, 2001 at 11:49:35, Ed Panek wrote: >>> >>>>On June 13, 2001 at 11:00:11, James T. Walker wrote: >>>> >>>>>On June 12, 2001 at 20:44:58, Ed Panek wrote: >>>>> >>>>>>Athlon 1.3Ghz 192 MBhash tables.. >>>>>> >>>>>>14:16:34.9 -0.50 25 -1715898192 >>>>>>07:27:40.6 -0.44 24 -344954985 >>>>>>00:00:31.9 -0.42 14 13011609 >>>>>>00:12:59.1 -0.36 20 339922649 >>>>>> >>>>>>Same position PIII 800 48 MB HASH tables >>>>>> >>>>>>17:57:26.9 -0.44 24 -1100754853 >>>>>>00:00:45.8 -0.42 14 13392336 >>>>>>00:23:58.1 -0.36 20 436641487 >>>>>>Bigger difference than I thought >>>>>> >>>>>> >>>>>>This is from the current Gt2 vs the web position >>>>>> >>>>>> >>>>>>Ed >>>>> >>>>>Hello Ed, >>>>>I have no idea what your numbers above mean but it seems to me that if you want >>>>>to compare processor speed you need to use the same amount of hash memory. I >>>>>can't get to the game to see what the position is like but if it's close to an >>>>>endgame or deep into an endgame the hash tables can make a very big difference. >>>>>Jim >>>> >>>> >>>>Hi Jim thanks for the reply..typically in computer chess the above means >>>> >>>>time-score-depth-nodes >>> >>>Then what does the - mean ? >> >>The last column does not mean number of nodes. >> >>It mean that value of an integer variable that is called nodes. >> >>This variable can get values between -2,147,483,648 and 2,147,483,647 and is >>updated by the C commands nodes++; >> >>nodes++ does not mean always increasing the value of nodes by 1 and when >>nodes=2,147,483,647 and you do nodes++ you get -2,147,483,648 > >That makes the comparison quite useless. > >ie pos 2 > >If the counter went around once then 2^32-344954985 positions are searched in >26860 seconds wich means 147 Kn/s wich is too low in the endgame on 800Mhz. So >it went round more than once. How often ? If normal speed is 800Kn/s it went >around 6 times. > >On a 1.3 Ghz speed is different and so is the number of times the counter went >around. I'd know how often, if I would now the speed, but that would make the >test unnescessairy. > >Tony The test was not only to compare nodes but to compare times because there is a difference in the hash tables and not only in the speed. Comparing nodes per second make sense only at depth 14 and depth 20 but it is not all the difference because more hash tables helped tiger to get the same depth with less nodes. You can see that the Athlon1.3 Gh 192 Mbytes hash is more than twice faster than pIII800 48 Mbytes hash at depth 24(the score is the same score when the time is 7:27:40.6 instead of 17:57:26.9) Uri
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.