SPCC: Testrun of SF nnue gk200627 finished

[pohl4711]

AB-testrun of Stockfish nnue gk200627 (with AVX2 binary from 200715) finished. Very impressive result!!!

https://www.sp-cc.de

(Perhaps you have to clear your browsercache or reload the website)

[Raphexon]

Amazing results

[Ron Langeveld]

You should not run 20 games at the same time on a 12 core (24 thread) laptop. The conditions won't be the same (unreliable results).

[Rebel]

You should not run 20 games at the same time on a 12 core (24 thread) laptop. The conditions won't be the same (unreliable results).

Can you elaborate why this setup is unreliable?

And what is reliable?

[chrisw]

You should not run 20 games at the same time on a 12 core (24 thread) laptop. The conditions won't be the same (unreliable results).

Can you elaborate why this setup is unreliable?

And what is reliable?

number of cores - 1

[Rebel]

You should not run 20 games at the same time on a 12 core (24 thread) laptop. The conditions won't be the same (unreliable results).

Can you elaborate why this setup is unreliable?

And what is reliable?

number of cores - 1

For the sake of the discussion :

I have heard that many times before, but on what is it based? Did someone ever tested it? Personally on my 20core/40threads I always test with concurrency=20 (task manager=50%) but in the past I have done tests with 40 (task manager=100%) and hardly noted a difference. Maybe we underestimate the (windows) schedular?

[chrisw]

You should not run 20 games at the same time on a 12 core (24 thread) laptop. The conditions won't be the same (unreliable results).

Can you elaborate why this setup is unreliable?

And what is reliable?

number of cores - 1

For the sake of the discussion :

I have heard that many times before, but on what is it based?

collapsing nps when core count gets pushed, gui failing, experiential things like that, so, rather than delve into massive testing to try and found out why (been there, done that many times) I just go with N-1 which appears to work.

Did someone ever tested it? Personally on my 20core/40threads I always test with concurrency=20 (task manager=50%) but in the past I have done tests with 40 (task manager=100%) and hardly noted a difference. Maybe we underestimate the (windows) schedular?

[Raphexon]

I'm fairly sure Pohl has 1-2 machines extra running in the background for those 20 concurrent games.

[pohl4711]

You should not run 20 games at the same time on a 12 core (24 thread) laptop. The conditions won't be the same (unreliable results).

That is not true. All engines run slower, if more than 12 threads are used. Thats clear. But this means not a distortion. Only a slowdown.
I opened 20x Stockfish 11-engine in console mode in Windows and started all of them with "go infinite". All of them ran smooth and with stable speed. As long as at least one thread is not in use (free for Windows-operations), there is no distortion. And I keep 4 threads unused.

[pohl4711]

I'm fairly sure Pohl has 1-2 machines extra running in the background for those 20 concurrent games.

No. Why should I? On 24 threads machine I could run 23 games simultaneously. But for a 24/7-usage, it is better not to use all CPU-power. With 20 games and only 3.0GHz speed (CPU reduced to 99%, that switches the 4.0GHz turbomode off), I have around 80% CPU usage. So the notebook CPU never gets hotter than 70 degrees Celsius.