What exactly is disappointing?
Using dual-Xeon for chess analysis
Sort:
To me, the fact I've understood more or less 0.01% of the post.
If this is about the stockfish engine, it's probably a better idea to ask on the stockfish forum or support. http://support.stockfishchess.org/
Been wondering how chess engines improve with multi-core CPUs, but I have the feeling that your dual-xenon setup is a bit of an overkill... or maybe simply a waste. I don't think that taking advantage of such a high number of cores/threads is a trivial thing to program on a chess engine... could be wrong. You'll have to ask developers to know if that setup makes any sense for chess analysis.
Even if all those cores could be optimized for chess analysis, I'm skeptical that they could keep up with the exponential increase in complexity significantly better than a more mundane CPU. In a useful time frame at least. There's inevitably some point at which it's impossible to analyze the next ply quickly enough. Diminishing returns overkill.
I have a new built dual-Xeon system ( Xeon-E5-2686-V3) with Supermicro mobo. It performs well as a slave system over my LAN, providing 35 cores to Aquarium's IDeA. However, the dual-Xeon is a huge disappointment in infinite analysis.
Running i.a. from a typical middlegame position and using 35 cores as a single engine, dual-Xeon is actually slower vs. my Intel i7-4930k O/C to 4.4 GHz.
I use mostly SF versions, e.g., SF 16100909 with 8192 MB of hash.
I tried doubling the SF setting of threads from 35 to 70 but that did not help. I also use TBs.
Could this have something to do with the Supermicro mobo I'm using in the dual-Xeon system. Other possible issues might be the need for Large Pages in Windows 10;
or correct settings in BIOS?