Author: Robert Hyatt
Date: 10:26:34 07/27/00
Go up one level in this thread
On July 27, 2000 at 12:19:58, blass uri wrote: >On July 27, 2000 at 11:22:53, Robert Hyatt wrote: > >>On July 27, 2000 at 09:55:14, blass uri wrote: >> >>>On July 27, 2000 at 09:00:30, Robert Hyatt wrote: >>><snipped> >>>>>Chess Tiger was computing in average 375,000 nodes each time it had to play a >>>>>move. >>>>> >>>> >>>>DB _only_ looked at 1.5M moves _total_ for each move it played. I thought >>>>you were searching much longer. >>> >>>Tiger searched for more time but did not search more moves because it used slow >>>pentium and no permanent brain. >>> >>><snipped> >>>>Yes.. but didn't you use more than 1 second? It only used 3/4 second of >>>>computation for each move it played. I thought you were using 30 seconds >>>>or some such? >>> >>>Tiger on slow pentium cannot see 375000 nodes in a second. >>> >>>The 1.5M vs 375,000 advantage is after considering the fact that deep blue >>>Junior used 3/4 second and tiger used more time. >>> >>>I also read that deep blue Junior used more time when it failed low so the 3/4 >>>second for each move may be wrong. >>> >>>Uri >> >> >>I believe the machine ws a 150mhz pentium? Crafty used to do 35K+ on a P5/133, >>and tiger is faster, although I don't know by how much. > >This is copied from previous post about it. >"I set up Chess Tiger 11.9 (Paderborn version) to play the game in 15 minutes. >Permanent brain was turned OFF." > >Game in 15 minutes with no permanent brain does not give you 30 seconds for >every move. > You are right. I had thought I remembered 30 seconds per move. I don't know where I read that, but perhaps Christophe was not the one that wrote it. But it did stick in my mind from somewhere... But it was obviously wrong here as your quote shows... >I also do not know if tiger is faster than crafty in nodes per second. Several on ICC have run both... it was significantly faster running both on the same machine. Which isn't surprising as I have not spent a lot of time worrying about raw speed (yet) other than in the parallel search. Speed in the eval makes it harder to read and modify. I am still modifying too much to stop and make it more complicated to do this... > > > I think he said he >>used 30 seconds per move? that would be in the same ballpark as the number of >>nodes the WebDB machine could search in 3/4 second, roughly... >> >>That I don't know about (the more time on fail low) particularly when it is very >>hard to determine that the program did fail low. IE I doubt it showed _any_ >>analysis locally... Lag probably caused more delays than fail lows did... > >I understood from a previous post of Ed that Hsu admitted that Deeper blue used >more time when it failed low. > >Uri That is possible, although an observer would not be able to say "that search took 15 seconds due to network delays" or "that search took 15 seconds because it used 1 extra second for a fail low and the extra 14 seconds were network delay." All you would know is it took 15 seconds of elapsed time to move. It would be very hard to explain _where_ the 15 seconds went... search or network lag...
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.