Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: HW Challenge - Revisited & Revised

Author: Robert Hyatt

Date: 07:55:08 12/31/03

Go up one level in this thread


On December 31, 2003 at 03:52:03, Sandro Necchi wrote:

>On December 30, 2003 at 01:33:52, Slater Wold wrote:
>
>>On December 29, 2003 at 23:03:50, Robert Hyatt wrote:
>>
>>>On December 29, 2003 at 14:14:03, Slater Wold wrote:
>>>
>>>>There is a lot of talk that the HW experiment(s) Peter & I have run are no good,
>>>>because the commercial is getting 'crappy' hardware.  That a Celeron vs a P4 is
>>>>not the same as a 2x2.8Ghz vs a 4x2.0Ghz Opteron.
>>>>
>>>>I don't want to get into this discussion.  I just want to see who is right.  ;)
>>>>
>>>>Here is what I propose:
>>>>
>>>>Crafty   = Crafty 19.8 on a P4 3.0Ghz with 8h/G.
>>>>Shredder = Shredder 7.04 on a P4 2.0Ghz with 2h/G.
>
>The best thing you can do to get close to the real tournament is to run Shredder
>7.04 under UCI and with tournament book option on.
>This will not be the same as in tournaments the book selection has not so many
>alternatives.
>
>The commercial version needs to have many alternatives not to be boring as it is
>made for the users to get fun with it.
>This means that you cannot create exactly the same, but it could be similar.
>
>Also here you are running a version which is about one year old, since the new
>Shredder 8 will be available soon, while Crafty is with the latest version and
>may include also some lines specifically prepared against Shredder 7.04.
>This needs to be taken into consideration as well.
>
>About Crafty ask Robert.

My book (as distributed) has _no_ prepared lines against anybody.  It is made
from a PGN game collection with about 2 million games, condensed into something
small enough to be reasonably trustworthy.  In fact, most build their own book
for Crafty since it is very fast to do this.

>
>Sandro
>>>
>>>My advice is to run crafty on both machines, and compare the NPS.  Then play
>>>with shredder at 2h/G and crafty at the appropriate multiplier to achieve the
>>>time advantage you are trying to test.  IE if you want to do an opteron
>>>comparison, I could give you a bench run to give an NPS value for the quad
>>>2.0ghz opteron I was running on...  You could use that to scale the Piv/3 to
>>>that speed if that is the point here...  However, I am not quite sure what you
>>>are trying to show?  IE I have run on an 8-cpu opteron and know what it looks
>>>like speed-wise as well.  I've not run on any opteron beyond 8-way yet...
>>
>>The P4 3.0Ghz is 66% faster than the 2.0Ghz.  It benches 1.23M nps.  Crafty had
>>a 4x time odds (8h vs 2h).
>>
>>The 4-way did more NPS than that.  Way more.  (1.23*4=4.92M).  Even at 5x time
>>odds it'd be tough.  That's only 6.15M nps.
>>
>>>>The games will be played on ICC.
>>>>
>>>>Bob, feel free to send me learning files, books, special instructions, whatever.
>>>> This will be just like a championship.
>>>
>>>I would use the same books on the ftp.  Re-grab the to clear any learning
>>>if you have run any with learning enabled.  Then run both engines with
>>>learning enabled...  let 'er rip...
>>>
>>>
>>>
>>>>
>>>>Same with you Sandro.



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.