Of course, the FPGA cards are quite a different story, and I would guess that many major changes would be needed to take advantage of these.
What a waste of time ,space and computer power...
Rybka 2.3.2a is bad choice here because Rybka 2.3.2a mp is very inefficient on > 8 cores computers , Zappa is very well design on >8 cores.
BTW, 120/40 is 3 minutes, not 5--or were you taking into consideration the fact that the first 15 moves tend to be booked?
>Even if someone had 20 overclocked quads for 20 simultaneous CC games, I would not recommend just leaving them to run. >Human guidance is necessary.
a) I have to sleep
b) I have family
c) sometimes I have social life (not very often, unfortunately because I am living abroad in new city and I don't have many new friends yet)
d) I like to do weekend trips
-> when I don't have time to think my games I just let my computer think. I run engine tournaments (Zappa, Shredder, Fritz, Rybka double round robin with 120/40). I run infinite analysis (quite often in 3- variation mode). When I have time then I start to work with computer output and analyse position with interactive methods. I test some opening positions by letting my computer to play over the weekend at engineroom to find out if others have interesting ideas about these opening positions.
When I and some others are telling that we do very long and cpu time intensive analysis runs it does not mean that we don't do others analysis. It means that we do these long runs in addition our interactive analysis.
>BTW, 120/40 is 3 minutes, not 5--or were you taking into consideration the fact that the first 15 moves tend to be booked?
Yes. I assumed that the 15 first moves are from book.
And finally I would like to give a hyphothetical example:
Player A is fide GM and he plays CC game assisted by rybka. He has old computer and uses 5 min cpu time per move (fast quad equivalent). from freestyle we know that best centaus can add something between 150-200 elo points to rybka.
Player B is postman. He has a fast quad and her runs 2 days per move using rybka.
CPU time ratio between these two players is around 600. If we assume that one additional ply doubles time then player B has 9 plys more than player A. I guess that this 9 ply is worth more than 150 elo points, but
the situation is different if we assume that player A does overnigth analysis (we all sleep at night?) using fast computer.
>Even if someone had 20 overclocked quads for 20 simultaneous CC games,
I am almost sure that van Oosterom have many computers because he is so rich. If I were he I would have 5-6 computers per game + Fide GM-level private secretary :-).
I am sure that van Oosterom also guides his analysis quite a bit--or else there would be no use in having 5-6 computers per move.
And rybka is somewhat faster on less threads, Rybka 232a: 29.3 ply; 3,101kN/s Intel(R) Xeon(R) CPU X7482 @ 3.00GHz 2993MHz, (32 threads)
> already seen 1024 on playchess. Something like this: Rybka 2.3.2a mp: 28.0 ply; 2561kN/s Quad Core AMD Opteron(tm) Processor 8354 2212MHz, (1024 threads)
All these machine u have listed here belong to Suj on the playchess server and he does not enter Freestyle any more since they stopped auto engine mode.
If u are lucky like me he may lend u a machine via remote but his larger machines are not available via remote i know i asked.
He can run Rybka on 1024 and zappa on 512 cores but only via a array of switches that will not take remote access so u will not see these machines in freestyle action sorry.
>>Rybka 2.3.2a mp: 28.0 ply; 2561kN/s Quad Core AMD Opteron(tm) Processor 8354 2212MHz, (1024 threads)
That must be fake of course.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill