Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: I do not understand why faster hardware is better

Author: Russell Reagan

Date: 23:45:02 11/15/03

Go up one level in this thread


On November 16, 2003 at 01:29:52, Timothy J. Frohlick wrote:

>To get to 14 ply with 11 to 12 good moves per ply would require a million
>billion choices to be searched.

I agree with Uri here. Taking into account the transposition table and iterative
deepening, it is not unreasonable to assume an effective branching factor of 6,
possibly 5 or even 4 (I'm not sure of the exact amount the transposition table
helps out). 6 is surely a safe conservative assumption.

6^14 = 78,364,164,096 nodes

At one billion nodes per second that's 78 seconds. Even at 7 million nps (which
Crafty gets on a quad Opteron 1.8 GHz), it only takes 3 hours for a full 14 ply
search. There are faster Opterons than what Eugene used to arrive at 7 Mnps, and
9Mnps could easily be reached with 2.2GHz quad Opteron. Then there are of course
8-way Opterons, and I've heard the Opteron will go as high as 2.6 GHz. A little
rough math and I estimate that Crafty could go at around 18 Mnps on an 8-way
Opteron at 2.6 GHz per cpu. That would mean a little over an hour for a full 14
ply search. If any non-pruning tricks could lower the branching factor to even
5, then we're talking 5 1/2 minutes to hit 14 ply. At a branching factor of 4,
15 seconds.

If we consider the best programs that are getting close to a branching factor of
2 sometimes, that amounts to a 32 ply non-exhaustive search in a middlegame
position in under 4 minutes. Wow! Maybe it won't be too long before chess is
just like checkers, hitting endgame tablebases shortly after leaving book.



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.