Author: Robert Hyatt
Date: 11:48:11 05/06/04
Go up one level in this thread
On May 05, 2004 at 18:49:39, Eugene Nalimov wrote: >On May 05, 2004 at 18:43:47, Robert Hyatt wrote: > >>On May 05, 2004 at 18:21:28, Vincent Diepeveen wrote: >> >>>On May 05, 2004 at 18:10:57, Robert Hyatt wrote: >>> >>>[snipping nonsense] >>> >>>Do you take the match for $1000 or do you NOT take the match for $1000. >> >>I believe I did. I don't play for money. >> >>> >>>So if i win, you pay $1000. If you win i pay you $1000. >>> >>>A public offer, just because you cry victory without showing up at a world >>>championship. >> >> >>Just pop over to ICC anytime you want and show your stuff... >> >> >> >>> >>>Most likely your hardware won't be worse than mine also, your cct machine for >>>example was very convincing quad opteron 2.2Ghz, holy smokes. I would *love* to >>>have such a baby work for me... ...you'll get 7.75MLN nps or so at it. >> >>Or 10 as I did in many positions... >> >> >>> >>>Most likely you are roughly 10 times faster in nps than i am. >>> >>>If you show up with a 32 node dual opteron machine (=64 processors), NO PROBLEM. >>>I would be HAPPY in fact if you do so, though i'm guessing that a quad is faster >>>for you currently. >> >>I haven't run on a 32-way opteron. An 8-way is significantly faster than a >>4-way. With Itanium 16 way is significantly faster than 8-way. I have no >>32-way data... >> >> >>> >>>So no cheap words please. Just say whether you take it or do not take it. >> >>I'll play any time you want. I just don't do childish bets,. >> >>Whenever you want... I'm on ICC. I'll even set up my good book to be used if >>you use the "diep" handle so I don't even have to be present.. Just give me an >>idea of when. >> >>BTW still waiting on that journal article citation. I notice you are trying >>your normal "run - hide - switch-the-topic" approach when caught red-handed in a >>lie... > >Just agree to play with him after he either show the article citation, or >announce he made a mistake :-) > >Thanks, >Eugene I'm now just waiting for what he says after looking at all the log files for the CB test positions run on the opteron AMD loaned me (thanks for the contact, by the way). The speedup = 1 + (NCPUS - 1) * .7 looks pretty reasonable based on the first two four-cpu logs I looked at. I'll leave it to Martin to publish the final number he gets to see how close it is to 1.7 and 3.1. Looks like 1.7 is a bit low, 3.1 is pretty close to 2.97 and 3.1 produced in the first two mt=4 logs...
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.