Author: Chris Carson
Date: 06:50:12 04/15/99
Go up one level in this thread
On April 15, 1999 at 09:31:32, Christophe Theron wrote: >On April 15, 1999 at 09:11:16, Chris Carson wrote: > >>On April 15, 1999 at 08:10:14, Tord Romstad wrote: >> >>>On April 15, 1999 at 06:36:04, Chris Carson wrote: >>> >>>>Best program (impressed me the most): >>>>Strength (Single processor): Deep Blue Jr, great record againt a >>>> wide range of opponents, not enough >>>> games, I wish they would have played >>>> more. >>> >>>Are you sure this is a single processor machine? I always thought >>>Deep Blue Jr had 12 processors... >>> >>>Tord >> >>See: http://www.chess.ibm.com/press/html/g.6.4.html >> >>"DBjr. is based on a single-node RS/6000 workstation, and uses a total of 16 >>chess accelerator chips. DB jr. also runs in a non-parallelized (serial) mode, >>meaning it does not divide the processing work up between processors; all of the >>calculations are done on a single processor." > >This sentence contradicts itself! > >There is one master processor and 16 chess accelerators. AFAIK, the "chess >accelerators" conduct a 4 plies search at high speed. So this is a parallel >chess computer... There are 16 processors conducting 16 searches at the same >time, and the RS6000 processor collects the results. > > > Christophe > > > >> >>Best Regards, >>Chris Carson You could be right. It could be that they work in parallel or it could be that one processor sequentially calls the appropriate accelerator. I am impressed with DBjr either way. I do not know enough about the design to make a stand one way or the other. I just base my opinion on the printed material. Best Regards, Chris Carson
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.