Dragan wrote:I also noticed this difference in depth. It indicates that clusters computing power (slow CPUs + overhead) is much lower then the power of Clemens's 16CPU machine. Of course assuming that depth shown by the cluster is comparable to R4.1 which is more then fair assumption.
They should really pull the plug on the cluster or update its software and hardware. The way it is now, it looks pretty close to being a scam.
I think calling it a scam is harsh as the cluster is really designed more for deep analysis for professionals or using its analysis capability to search several lines deep very quickly. It really is not setup to play engine vs. engine matches especially with only 64 cores. Remember that it is using several motherboards that are communicating via network, this creates huge latencies. I actually beleive that a 64 cluster would play weaker than a 16 core single motherboard with shared memory Rybka 4.1. At the moment the cluster plays like a MPV Rybka 4.1 with 3 lines full depth. What is certain is that the core Rybka code (or chess playing code) is in serious need of upgrading, as evaluation wise and search wise ... it is one or two generations behind.
