Time forfeits because of too few cores?

[mvanthoor]

My quadcore has hyperthreading, so I run 8 games in parallel (total CPU load 100%) using c-chess-cli. All while browsing the web, listing to music, and watching videos. No problems, neither with the engines, nor even any lags in the browser.

Are the engines that are running on the logical cores not degraded in performance; or are you going by the premise that the operating system will probably switch them around (possibly even during gameplay), and that they all get the same amount of degradation (= running time on a logical core) during the tournament?

I could start running 6-8 matches at once on my old 6700K, but I have never felt comfortable running more stuff than I have real cores, if speed actually counts.

[Ras]

Are the engines that are running on the logical cores not degraded in performance

They are because the computing time for browsing and videos has to come from somewhere.

or are you going by the premise that the operating system will probably switch them around (possibly even during gameplay), and that they all get the same amount of degradation (= running time on a logical core) during the tournament?

Yes - I'm running several 10k games, and the impact will be equal for all engines.

I could start running 6-8 matches at once on my old 6700K, but I have never felt comfortable running more stuff than I have real cores, if speed actually counts.

I've benchmarked a hyperthreaded core with two workers to be worth 1.5 physical cores. This means that compared to running only 4 games on that quad, each engine will only see 75% of the performance, but since the performance penalty is the same for all engines, that levels itself out again.

The issue with hyperthreading and engines can be different if you want to do analysis using a multi-threaded engine. You get 50% more total nodes, but split up across twice as many threads. Whether that is a good thing depends on how the analysis engine scales with the thread count. It may well happen that going from 16 threads on 16 real cores up to 32 hyperthreaded threads on 16 real cores is a net loss of strength.

[Modern Times]

or are you going by the premise that the operating system will probably switch them around (possibly even during gameplay), and that they all get the same amount of degradation (= running time on a logical core) during the tournament?

Yes - I'm running several 10k games, and the impact will be equal for all engines.

Stefan Pohl says the same, he uses 20 threads on his Ryzen 3900X (12 cores, 24 threads) for his ratings list. (20 instances of engines using 1 thread)

I am still reluctant to do that, I currently don't run any more threads than I have real cores.