Author: Komputer Korner
Date: 22:51:08 03/27/98
Go up one level in this thread
For a chess program there really isn't diminishing returns unless you
consider that at 14 to 16 ply there is somehow a tactical barrier at
which the program will not fall for any long range tactical tricks any
more. There isn't an absolute tactical barrier but the chances of being
caught tactically do approach 0 the deeper the ply. For the first 8
hours of thinking time the program will find better moves as the time
increases. After that it will be rare, as the program needs 2.5 -3 times
longer (than the total accumulated thinking time on all the previous
ply) on each succeeding ply. However the fly in the ointment is the size
of the hash tables. When they fill up the search will slow down by a
factor of 2 or more depending on the position (in the endgame it could
be as much as a factor of 4 to 5 or more). So for you the diminishing
returns are when your hash table fills up. The formula for Fritz 5
(Extreme hash table size should be nearly the same) is
HT in Kb = 2* Processor [Mhz] * t
where HT is hash table size in kilobytes and Processor [Mhz] is clock
speed of your processor in Mhz and t is the time available per move in
seconds. If your system doesn't have enough RAM then obviously your hash
size will have to be lower or else the thinking time will have to be
lower. In all cases, you do not want your Windows swap file to swap
virtual memory from your hard disk as this is too slow. If you find that
the swap file is going crazy on you when setting large hash tables then
in Windows you will have to set the
system.ini file.
Under the [Vcache] section put the
following line:
MaxFileCache=xxxx where the x's are the RAM in Kb that you want
to
limit the cache to.
It should be either 1024,2048, or 4096 or higher multiples of. If you do
this input/output processes that require lots of RAM will slow down but
if
all you do is run chess programs to play against, then this is
satisfactory. If you have a huge database or opening book operation to
do,
you can always change it back. But you shouldn't have to do this for
Extreme or Fritz because of their separate extensions created for the
engines.
On March 26, 1998 at 15:37:13, Timothy Vaughan wrote:
>I frequently do analysis with Extreme Chess. I am unsure what are
>reasonable values to use for the various parameters. For example,
>is there real benefit to letting it use 5 minutes per move instead
>of 4 minutes? When do diminishing returns set in?
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.