Lc0 will hardly ever win TCEC or CCCC?

[Raphexon]

Everything is a moving target (all the mentioned engines) and so it is very hard to make predictions.

I guess the ultimate winner will be ...
Something new. At some point anyway.
It quite often turns out that way.

We get surprised and we get surprised and somehow, even after being surprised over and over and over, we get surprised again. It's pretty strange when you think about it.

Watch Olithink become world champion. Of course by then it might be bloated to nearly 100k.

Well it'll be over for quite some period. MCTS and big nets are not the way to go as it has been known for a while. And A/B with handcrafted eval is not using hardware in the best possible way. It's been clear for a while that A/B with fast small NN eval is a way to go. We lacked good implementation of it (we can now see how bad and "noob" was Giraffe NN for example). Now it's there. Once main SF devs embrace it and start tuning search together with eval, sky is the limit. This is a real paradigm shift, not A0 or Lc0.
Plus there is another aspect. New CPUs will be having more and more support for vector INT8 and BFLOAT16 multiplication operations. So running things on GPUs is not gonna make advantage.

They're already working on it to make it compatible with Fishtest.
But this won't be done in a week.

[Dann Corbit]

Plus there is another aspect. New CPUs will be having more and more support for vector INT8 and BFLOAT16 multiplication operations. So running things on GPUs is not gonna make advantage.

I think your analysis is pretty interesting, but there is something to think about with this last part. The AMD team is working on transparent memory access, including direct access to video ram by the CPU. So it is much more like a real co-processor than the current horrible paradigm of make a huge, slow copy to video RAM, have the GPU churn like mad, and then make a huge slow copy from video RAM.

And a second pause for thought is that CPUs do billions of operations per second and GPUs do trillions of operations per second. If we can eliminate the memory copy bottleneck and have the CPU and GPU do true cooperative processing that would really be something and perhaps allow a sort of synergy that has not been possible.

[Milos]

Everything is a moving target (all the mentioned engines) and so it is very hard to make predictions.

I guess the ultimate winner will be ...
Something new. At some point anyway.
It quite often turns out that way.

We get surprised and we get surprised and somehow, even after being surprised over and over and over, we get surprised again. It's pretty strange when you think about it.

Watch Olithink become world champion. Of course by then it might be bloated to nearly 100k.

Well it'll be over for quite some period. MCTS and big nets are not the way to go as it has been known for a while. And A/B with handcrafted eval is not using hardware in the best possible way. It's been clear for a while that A/B with fast small NN eval is a way to go. We lacked good implementation of it (we can now see how bad and "noob" was Giraffe NN for example). Now it's there. Once main SF devs embrace it and start tuning search together with eval, sky is the limit. This is a real paradigm shift, not A0 or Lc0.
Plus there is another aspect. New CPUs will be having more and more support for vector INT8 and BFLOAT16 multiplication operations. So running things on GPUs is not gonna make advantage.

They're already working on it to make it compatible with Fishtest.
But this won't be done in a week.

That's really good news, didn't have a clue they'd so quickly adopt it (I don't follow SFdev discord).

[Milos]

Plus there is another aspect. New CPUs will be having more and more support for vector INT8 and BFLOAT16 multiplication operations. So running things on GPUs is not gonna make advantage.

I think your analysis is pretty interesting, but there is something to think about with this last part. The AMD team is working on transparent memory access, including direct access to video ram by the CPU. So it is much more like a real co-processor than the current horrible paradigm of make a huge, slow copy to video RAM, have the GPU churn like mad, and then make a huge slow copy from video RAM.

And a second pause for thought is that CPUs do billions of operations per second and GPUs do trillions of operations per second. If we can eliminate the memory copy bottleneck and have the CPU and GPU do true cooperative processing that would really be something and perhaps allow a sort of synergy that has not been possible.

Yes, what AMD is working on is indeed very promising technology. One extra thing that is really required beside transparent RAM is cache coherence between CPU and GPU cache. However, that one is much more challenging to achieve.

[Dr.Wael Deeb]

Plus there is another aspect. New CPUs will be having more and more support for vector INT8 and BFLOAT16 multiplication operations. So running things on GPUs is not gonna make advantage.

I think your analysis is pretty interesting, but there is something to think about with this last part. The AMD team is working on transparent memory access, including direct access to video ram by the CPU. So it is much more like a real co-processor than the current horrible paradigm of make a huge, slow copy to video RAM, have the GPU churn like mad, and then make a huge slow copy from video RAM.

And a second pause for thought is that CPUs do billions of operations per second and GPUs do trillions of operations per second. If we can eliminate the memory copy bottleneck and have the CPU and GPU do true cooperative processing that would really be something and perhaps allow a sort of synergy that has not been possible.

Interesting concept indeed ....

Cheers,
Dr.D