TalkChess.com

Is it the Hardware speed or some breakthrough programming innovation? So far what is the score between Komodo and Leela ? https://www.youtube.com/watch?v=ymDA7-hdkKA My answer is a bit different than others. Leela (following AlphaZero) has found a way to use a powerful GPU for chess, which is itsel...

I just noticed the following sentence on your page: "Move Overhead is set to 500ms, if an engine allows to do so." This would be half of the increment, unless it is a typo (perhaps you meant 50 ms?). We (Komodo) have never used more than 125 ms in any tournament for Overhead, and that was only on a ...

Thanks for explanation. But how do You know, that Komodo then uses real cores and no virtual ones, if computer has HT on? It seems that the truth is somewhere between these two extreme possibilities. When the thread count goes beyond the core count, total NPS continues to climb, but nps per thread ...

Houdini 6.03 (64 threads, 16 GB hash) Komodo 2242 (46 threads, 6 GB hash) Stockfish 19010616 (90 threads, 16 GB hash) SF is too weak and get more power :?: Since the machine has 48 cores, we don't believe it is helpful for Komodo to use more than that number of threads (minus 1 or 2 for other thing...

It seems that with multiple threads (and perhaps also with ponder?) 12.3 MCTS is subject to occasional crashes and time forfeits, with the frequency increasing with thread count. We have made progress towards a fix, but we want to avoid multiple bugfixes so we want to be reasonably sure we have fixe...

Thanks. This tells us that the crashes in CCC and TCEC were probably not due to any changes since 12.3. We're trying to find the problem. Have there been any crashes in CEGT non-ponder testing? Are there any common elements in the positions where it crashes, or any illegal moves displayed?

Larry

Hi Larry, why MCTS search in Komodo is non-determinicstic despite using 1 thread only? I see some randomness during the search. Each run gives slightly different output. For example Leela's search is deterministic on single thread. Here is Komodo 12.3 on 1 CPU in starting position: First run: 0.02:...

:arrow: Has splitting the cores between MCS and ordinary alpha-beta been tried? It looks to me like the MCS version is fast at finding a stable, good move and the alpha-beta version solves tricky positions faster. I've been keen on that idea for years, long before we even started on MCTS, but it's ...

Using another core to access the tablebase won't save any time if the normal search has to wait for the result of the probe to decide what to do next. This is the fundamental problem of MP search in chess; the alpha-beta algorithm is basically sequential; two cores aren't as good as one core at twic...

:arrow: Has splitting the cores between MCS and ordinary alpha-beta been tried? It looks to me like the MCS version is fast at finding a stable, good move and the alpha-beta version solves tricky positions faster. I've been keen on that idea for years, long before we even started on MCTS, but it's n...