Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: DB vs Kasparov Game 2 35. axb5

Author: Ernst A. Heinz

Date: 10:19:44 11/20/98

Go up one level in this thread


On November 20, 1998 at 12:49:47, Dave Gomboc wrote:
>
>Are you going to be able to keep this running?  You've got faster hardware than
>most of us, [...]

SPECint95-wise the 500MHz Alpha-21164a is *not* faster than a 400MHz PII.

"Fritz 5", "Junior 5", and "Nimzo 98/99" all normally process more nodes per
second on a 400MHz PII than "DarkThought" on a 500MHz Alpha-21164a.

>I wonder if you've just "nice'd" everything to
>the maximum, so that people can still use the machines.

No, "nice'd" runs do not make much sense if you want to make progress at these
extreme search depths (especially because "DarkThought" uses neraly all the
available RAM of the machines --> "nice'ing" leads to immediate swapping).

>Can you estimate:
> a) when 21.01 (Qb6) will be searched?
> b) when 21.all will be searched?

It looks like the effective branching factor increases to roughly 4 at these
extreme depths with the completely overloaded hash tables.

>I guess I am not even sure from the above that 20.all has been searched yet.

You should re-read my post then because it states quite clearly that iteration
#20 got completed (all moves searched) after roughly 70 hours. The computation
of the PV (first move) finished after about 52 hours.

Assuming an effective branching factor of 4, iteration #21 will then take an
estimated 280 hours to complete with the computation of the first move
finishing after an estimated 208 hours.

=Ernst=



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.