Author: Robert Hyatt
Date: 17:00:50 10/24/05
Go up one level in this thread
On October 24, 2005 at 10:59:01, Uri Blass wrote: >On October 24, 2005 at 10:29:29, Robert Hyatt wrote: > >>On October 23, 2005 at 21:07:43, Uri Blass wrote: >> >>>On October 23, 2005 at 15:03:00, James Swafford wrote: >>> >>>>On October 22, 2005 at 23:47:45, Uri Blass wrote: >>>> >>>>>On October 22, 2005 at 22:16:17, James Swafford wrote: >>>>> >>>>>>On October 22, 2005 at 14:45:16, Drexel,Michael wrote: >>>>>> >>>>>>>On October 21, 2005 at 14:48:36, James Swafford wrote: >>>>>>> >>>>>>>>On October 21, 2005 at 14:39:07, Ryan B. wrote: >>>>>>>> >>>>>>>>>On October 21, 2005 at 13:43:40, Charles Roberson wrote: >>>>>>>>> >>>>>>>>>> >>>>>>>>>> Make your predictions on placement for all programs. >>>>>>>>>> >>>>>>>>>> Also, make your prediction for biggest upset. >>>>>>>>>> There's always an upset; which will it be this time? >>>>>>>>> >>>>>>>>>Sure I'll play. I think HiarcsX will shock us all and win and Fruit is come in >>>>>>>>>second. As for biggest upset I will go with Crafty beating Fruit. >>>>>>>> >>>>>>>>Crafty beating Fruit would not be that big of an upset, IMHO. >>>>>>>>Calling that a huge upset implies Crafty is *significantly* weaker, >>>>>>>>and I don't believe that. >>>>>>>> >>>>>>>>-- >>>>>>>>James >>>>>>> >>>>>>>Well, if about 200 rating points is not significant, what is significant for >>>>>>>you?? >>>>>>>Topalov even thinks 60 points is a "different class altogether". >>>>>>>Not that I agree with him. >>>>>> >>>>>>200 points is significant, if that is true. >>>>>>Where do you get 200 points difference? >>>>>> >>>>>>-- >>>>>>James >>>>> >>>>> >>>>>4 minutes /40 moves >>>>>http://www.husvankempen.de/nunn/eloblitz.html >>>>> >>>>>Fruit2.2 2770 >>>>>Crafty Cito1.2 2557 >>>>> >>>>>This is without the original opening books but I do not see a reason to assume >>>>>the result is very different with opening book. >>>>> >>>>>Fruit performed clearly better than Crafty in WCCC inspite of clearly inferior >>>>>hardware so 200 elo is clearly logical. >>>> >>>>Clearly you use the word 'clearly' too much. >>>> >>>>What hardware is used to play those games? One of Crafty's strengths is >>>>it's scalability. If those games were played on a single processor machine, >>>>then the gap would be much smaller than 200 points. >>> >>>WCCC were not played on a single processor and Crafty used 8 processors against >>>1 processor of Fruit. >>> >>>> >>>>I'm convinced Fruit is better than Crafty, at least on a single processor >>>>box. My original point was that it isn't *so* much better than Crafty >>>>that Crafty beating it should be the biggest upset of the tournament. >>> >>>It is at least 200 elo better on a single processor. >>> >>>In the CCT blitz tournament Crafty used 2 processors against 1 of fruit but I >>>read that this was not an hardware advantage because the 1 processor of fruit >>>was faster. >> >>That is correct. A single AMD64 is faster in raw NPS than my dual xeon. Factor >>in the 30% loss to search overhead and the AMD64 is significantly faster. Not >>that this matters one bit. If you look at the game, you will find that Crafty >>reached an easily won position, but the recent king-safety changes caused it to >>basically self-destruct, and open up its own king position for no good reason. > >If you talk about the game of Crafty against Fruit then the best that I see is >one pawn advantage for Crafty against Fruit in the middle game that is evaluated >by fruit as 0.7-0.8 pawns. > >It is an advantage but I do not think that it is an easily won position and I am >not sure if it is possible to translate the advantage into a win. > >> >> >>> >>>In case that Crafty use 8 processors efficiently in blitz then the difference at >>>blitz is probably small but it never happened in tournaments and in the WCCC >>>blitz tournament Crafty did not use them efficiently. >> >> >>Care to explain what this is based on? Crafty uses them pretty efficiently in >>_all_ games. The WCCC blitz event was not very fast... > >I remember that Crafty did not get significant speed improvement from 8 >processors in the WCCC blitz event and that peter berger even hesitated if to >use them because Crafty WCCC version was not ready for blitz at the time of the >blitz tournament. > >I do not claim that Crafty cannot use them 8 processors efficiently but only >that as far as I know it did not happen in tournaments. > >Uri This is simply misinformation. What I said was "The SMP search is not as efficient in a very fast game as it is in longer games." But "not as efficient" is a +long+ way from "not very efficient." If you'd like to see some search times for 1-2 second searches, I can provide plenty, and the speedup is nearly as good as for 1-2 minute searches, within a small margin of error. I would _always_ play with 8 processors rather than 4, 2 or 1, given the chance, no matter what the time control. N is always better than 1 no matter what the time control.
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.