1 core can be better than 6 cores (Stockfish)

[Werewolf]

I don't think that works, but it should work as a *.pgn file. Create a new database (select file type pgn on the bottom), copy all games from cbh game list (ctlr a, ctrl c), then paste it into pgn file.

But maybe Vincent knows more about it.

Thanks

Yes, but how do you attach a pgn database to this website? I can only see a way for single positions.

[Werewolf]

This is valid for 3 days

https://we.tl/t-O3xq3aNpJg?utm_campaign ... TRN_TDL_05

[Paloma]

Thanks Carl

[Werewolf]

This is for anyone who is interested in the relationship between Stockfish's tactical ability and threads.

https://youtu.be/LqrRmXWxoMY

[sarona]

This is valid for 3 days

https://we.tl/t-O3xq3aNpJg?utm_campaign ... TRN_TDL_05

Could you post another link?

I was away from the forum and could not get the file in time.

Thanks!

[Uri Blass]

This is for anyone who is interested in the relationship between Stockfish's tactical ability and threads.

https://youtu.be/LqrRmXWxoMY

Surprising result that more threads does not give better result when more time give better results.

I wonder what is the first stockfish version when it happens(something between 11 and 17).

[Werewolf]

This is for anyone who is interested in the relationship between Stockfish's tactical ability and threads.

https://youtu.be/LqrRmXWxoMY

Surprising result that more threads does not give better result when more time give better results.

I wonder what is the first stockfish version when it happens(something between 11 and 17).

Yes, that's a great question.

[Jouni]

Wow excellent video! Clearly current SF don't benefit practically when using over 8 cores . Cluster SF gets impressive results. I have always wondered, that what is difference to normal SF? Can You use cluster with one 64 core CPU?!

[Werewolf]

Wow excellent video! Clearly current SF don't benefit practically when using over 8 cores .

For Elo/playing strength it will. For tactics I can't see any benefit over 2 cores!

[gordonr]

This is for anyone who is interested in the relationship between Stockfish's tactical ability and threads.

https://youtu.be/LqrRmXWxoMY

This is a very interesting video and I thank you for all the effort you put in with your testing and documenting of it.

However, I'm not convinced that your testing with the Stockfish Cluster refutes the first bullet point of the explanations based on a single thread. As you said yourself, the cluster involves multiple computers/nodes. This is a very different hardware architecture with a distributed memory so it's not comparing like-with-like in terms of your table.

We're agreed that 2 threads are non-determistic (as you pointed out in the video) and therefore you have to do multiple runs and average the test results because you will have some runs better than others. So you're noting the average and not the best, as you should do.

Now hypothetically, just like a type of cluster, I could run a test on 10 computers at the same time (each computer using 2 threads). That setup has more chance of hitting the solution quicker than one computer. It's like taking the best of 10 runs rather than the average.

If you want to disprove the "single thread" theory, I think it needs to be done with the same test hardware. I appreaciate that this may not be feasible in practice but then you have to just accept that you can only conclude up to 64 threads or whatever the single node limit is.

To put it another way, your table is showing no difference from 2 to 64 threads. But a difference between 64 threads and 720. If this is purely to do with thread count and nothing to do with the change in architecture, are you suggesting there is some number between 64 and 720 where Stockfish suddenly starts to get better again at solving tactical problems? That would require quite an explanation in itself

Once again, I really appreciate your efforts. I just wanted to discuss what is an interesting topic. Maybe I'm talking nonsense