http://www.msnbc.msn.com/id/33992023/ns ... of_energy/
a mere 1.759 petaflops!
Can this somputer beat Cluster Rybka?
Moderators: hgm, Rebel, chrisw
-
Vinvin
- Posts: 5228
- Joined: Thu Mar 09, 2006 9:40 am
- Full name: Vincent Lejeune
Re: Can this somputer beat Cluster Rybka?
1) Computer chess power is not about FLOPS but about MIPS.
2) Computer chess support badly so much processors (the one who perform very well recently is Rybka and supports less than 200 processors).
2) Computer chess support badly so much processors (the one who perform very well recently is Rybka and supports less than 200 processors).
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: Can this somputer beat Cluster Rybka?
I do not believe Rybka's cluster implementation is anything more than mostly hyperbole. Rybka is very strong. Which means that a cluster version will be no weaker, even if it is not much stronger. Which is a problem for everyone of course. But I hardly worry about 72 cores vs 8 cores, since their parallel (cluster) algorithm is known and is not very good.Vinvin wrote:1) Computer chess power is not about FLOPS but about MIPS.
2) Computer chess support badly so much processors (the one who perform very well recently is Rybka and supports less than 200 processors).
Rybka still makes ugly mistakes, as do we all. It made two horrible blunders against Crafty in the ACCA event this past weekend and turned a win into a simple draw.
I still think almost 100% of its "output" is utter nonsense. Bogus depth. Bogus NPS. Bogus scores (how many programs have a score of 1.29-1.30 for 10 consecutive moves in a game???
Personally I worry more about Rybka as a program, than as an SMP program. I believe that every time we give both programs 2x the number of processors, that I will see more of a gain in performance and playing skill. But there is a good way to go to catch up.