Author: Sandro Necchi
Date: 04:58:32 02/27/05
Go up one level in this thread
On February 27, 2005 at 07:51:30, Uri Blass wrote: >On February 27, 2005 at 07:11:35, Sandro Necchi wrote: > >>On February 27, 2005 at 06:22:42, George Tsavdaris wrote: >> >>>On February 27, 2005 at 06:09:15, Sandro Necchi wrote: >>> >>>>On February 27, 2005 at 06:00:06, Kurt Utzinger wrote: >>>> >>>>>On February 27, 2005 at 05:54:35, Rex wrote: >>>>> >>>>>>With all that was learned out of IPCCC I sure would like a "free" update. >>>>>> >>>>>>Enjoy >>>>> >>>>> I am almost sure that there will be no such update. >>>>> At Paderborn an experimental version of Shredder 9 >>>>> is participating and nobody knows about the real >>>>> playing strength of it. >>>>> Kurt >>>> >>>> >>>>No, we have used the same 9.0 version. >>>> >>>>We only used a modified book for the tournament...an intermediate version for >>>>the next WCCC or match against Hydra hoping someone can arrange it. >>>> >>> >>> I really hope a match between Shredder and Hydra, now that you know the huge >>>strength of the second and you will be prepared properly. >> >>Yes, last year we could not prepare properly also because I was very busy with >>my work and to move to Lucca. >>Now we are more and I am planning to make more work too even if I am moving to >>another house this summer, still in Lucca. >>Also the bug problem, which I discovered, "made me crazy" trying to change the >>book to avoid problems without having the time to do it. >>With a reliable version is different as I can use the material which is ready >>and add more to it. >> >>>But i would like a 14 >>>or more games match, for having a good indication of the strength..... >>> >>> Since many believe that hardware differences is not so important as we think, >>>while i believe that in this high level even small hardware differences can make >>>the difference, can you give an approximation of the expected ELO difference >>>between: >>>"Shredder+Dual" , "Shredder+Quad" , "Shredder+8 CPU's" , "Shredder+16 >>>CPU's"....? >> >>Well, we need to improve this as only 2 processors are used fully and the others >>partially only. >>Stefan did not have time enough to work on this, so he can improve this a lot >>spending time enough. >> >>I believe what follows, comparing with one processor only: >> >>1. Dual = +30 Elo >>2. Quad = +40 Elo >>3. 8 cpus = +50 Elo (we did not used this harware yet) >>4. 16 cpus = are not supported yet as the max is 8. >> >>P.N. >> >>1. these are estimates by me as I have no data to support this rather than my >>impression looking the games. >>2. These data are referred to a 32 bit program version, and not fully optimized >>to use more than 2 processors. This means that there is a lot of room for >>improvements. >>3. The program strenght does not improve very much because if Shredder does not >>find a solution in a reasonable time it will take very long to find it. This can >>be improved too. > >Does 3 mean that the program is relatively weaker in correspondence games >relative to what you expect or the problem is only in case of using more than >one processor? No, I think it is based on chess knowledge connected with search. Every programmer has his own way. MChess was able to solve almost everything if a much higher time was given, but it would be much slower than Shredder on most positions. What amaze me is that Stefan is able to improve his program on this with each version. This gives me the impression that he can still make strong improvements. > >Is 3 specific problem of shredder9 or also a problem of older versions? No, see above. > > >It seems to me that there is a problem with shredder9 search based on the >following analysis(without KBP vs KP but with all of the 4 piece tablebase and 5 >piece tablebases with no pawns and KPP vs KP) > >http://www.talkchess.com/forums/1/message.html?414099 Pls. inform Stefan. He will take a look on this. > >I wonder if there is general problem with recovering from fail low with >shredder9 or this is a problem only in that specific endgame. Stefan can answer you on this. > >Thanks for your information. > >Uri 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.