Author: Timothy J. Frohlick
Date: 23:19:30 06/07/01
Go up one level in this thread
Mr. Burcham, Good idea but Ed Schroder did this with his Rebel programs about one year ago. It would certainly give a hint at the playing strength on faster machines. This sort of testing would be tedious but informative. Don't get too scientific on us K. I think that with increasing depth the return on investment diminishes. Just consider how Kasparov punished Deep Blue. DB was doing a whale of a lot more computations than a puny 4000 Mhz AMD processor. Knowledge is the key. We need faster machines in order to access the enormous databases of rules and exceptions to the rule. Tim Frohlick On June 07, 2001 at 21:11:15, K. Burcham wrote: > > >maybe the 4 gig amd processors will be out in two years, maybe even next year. >so i was wondering how accurate it would be to test this now. > >lets say we know that tiger can average 400 kns with a 1500 mhz amd. >so lets set the clock to four minutes per move. >this will convert to 96,000,000 nodes in four minutes. >this is with permanent brain off. > >lets say we estimate the 4000 mhz processor to get about 1100 kns with >tiger. this will convert to 264,000,000 nodes in four minutes. > >so the idea is to see what the 4 gig will do against the 1.5 gig, if we had the >timer set to four minutes per move, permanent brain off. > >so based on the above figures we could take two seperate 1.5 gig pcs, >play tiger against tiger with all settings equal, except for time control. > >on 1500 mhz computer #1 we will set the time per move to 4 minutes. > >on 1500 mhz computer #2 we will set the time per move to 11 minutes. > >now we have 1500 mhz computer #1 getting 96,000,000 nodes per move. > >now we have 1500 mhz computer #2 getting 264,000,000 nodes per move. > > >this type of test opens up several questions for me. > > >please dont tell me how the next generation of processors will be 500 megagig >current pulses traveling in the water of a desktop fishbowl. >and how our ram will hang from the ceiling in huge grids. > >please dont tell me how the software changes every six monthes. >please dont tell me how the hardware changes every three monthes. > > >the above discribed test is based on the hardware as we view it by todays >standards. > > > >please comment. > >thanks
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.