Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: PDA hits 100,000+ NPS

Author: Mike Byrne

Date: 17:07:19 10/04/03

Go up one level in this thread


On October 04, 2003 at 12:33:27, Russell Reagan wrote:

>On October 03, 2003 at 20:13:05, Mike Byrne wrote:
>
>>A Dell Axim 400 Mhz (OS Mobile 2003) (overclocked to 472 mhz...
>                                       ^^^^^^^^^^^^^^^^^^^^^^
>
>Hmmm...
>
>I recall installing Pocket Fritz a while back on a 400 MHz XScale, and it only
>got around 4-5Knps, give or take. Jorge Pichard said that his comparable PDA
>(400 MHz XScale) was overclocked a bit, and that he got something like 75Knps.
>It turned out that when he overclocked, it messed up the internal clock and the
>nodes per second that the program reported were not accurate. I wonder if this
>is the case here, because going from Pocket Fritz 2's 5Knps to over 100Knps on
>hardware that is very similar is questionable IMO, even if it is a different
>program (over 20x faster on more or less equal hardware?).
>
>I wouldn't trust these numbers unless they came from a stock PDA with no
>overclocking or other "enhancements" that could fudge up the numbers.

The overclocking is roughly 19%.  472/400 == roughly 1.19.  Nps increase was
almost  1.18 from overclocking.  I can also tell you visually there was no
timing distortments.  The nodes are counted by the program.  This is legit, when
it is released (the new version)anyone with a Dell Axim 400 Mhz WITH Pocket PC
2003 (Mobile 2003) AND the AO5 ROM ( the critical words there are AND and WITH
-- you must have this configuration that is currently shipping from Dell).

I stake my 20 years reputation of working with chess computers that this is for
real.  This is also verified by the solution times and the node count maintained
by the programmer.  I wish could share more with you, but I am bound by my
agreement with author  not to reveal the program or the test problems at this
time.  He wants to have it available prior to any disclosure.  Note - the 100
Nps was just one problem - not the average.  I was asked to test several
problems that the program resolves in a reasonable amount of time , but not
instantaneously.  The Dell running straight @ 400 Mhz average 69,745 nps for the
problems tested.  Overclocked to 472 Mhz using Speedstepper, the nps was 82,141.
 The setting used on the program was the minimum hash setting.  The Nps are the
fastest when this is used.  Solution times are often faster when hash tables are
used, but the NPS is lower.  Using a 2MB hash table, the nps was 63651 (400 Mhz)
and 75910 (472 Mhz).  With 2MB hashtable, the nps increase as the result of the
overclocking was 19%.  The overclocking was very consistant in this regard.  But
even with the slower nps count, the bigger hash tables solved the problems 16%
faster.   This update will be significantly faster than the current version -
especially for those with Dell Axim 400 Mhz WITH Pocket PC 2003 (Mobile 2003)
AND the AO5 ROM.  The author is extremely pleased with these results as he
should be.

I am also pretty damn excited about it.

Best,

Michael



This page took 0.01 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.