Author: Bo Persson
Date: 15:53:14 06/11/03
Go up one level in this thread
On June 10, 2003 at 14:48:23, Ricardo R Santana wrote: >Hello All >so, with the following information (thanks Bo) > >>Double the speed, maybe +50 ELO points? > >Deep Blue >>200M NPS >> >>>compared >>>with today speed (nodes/second ??) >> >Today computers: > >>1-2M for normal PCs, 3M for 8 processor servers. >> > >Lets say we have normal PC with 2.2M (nodes/second). So, we expect >2.1NPS*2^6 = 140 NPS and 2.1NPS*2^7=280 NPS. So, today PCS should increase about > or 7 times to get 200M NPS !!. This is the first conclusion. > >So, since 200M NPS are about 6 or 7 "doubles", we can expect deepblue to have >6*50 to 7*50 rating points ahead todays PC. So, this is about 300-350 points. >SSF presents a lot of programs with 2600. So, DeepBlue is about 2950 !!! This is >too crazy for me...please, help ME !!! You must remember that SSDF rating and USCF rating cannot be comapread, as they are not correlated. Also, Deep Blue may have used more nodes for a particular search than the micros, because it was relatively cheaper for them. We know, for example, that Hsu didn't like null moves and thought he could do without them because of the speed advantage. On the other hand, the 200M NPS was an average figure over games where the processors were idling some of the time. With a faster host, maybe we would get a higher number for a hypothetical reassembly of the system? :-) Bo Persson bop2@telia.com
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.