TalkChess.com

Nvidia seems to have come with a new GPU ... as usual we always wonder if this can be ported to a chess engine. What caught my eye is the "double precision" calculations. Can this be useful for a chess engine?

https://devblogs.nvidia.com/parallelfor ... de-pascal/

Nvidia is very busy

http://www.nvidia.com/object/deep-learning-system.html

JP.

M ANSARI wrote:Nvidia seems to have come with a new GPU ... as usual we always wonder if this can be ported to a chess engine. What caught my eye is the "double precision" calculations. Can this be useful for a chess engine?

https://devblogs.nvidia.com/parallelfor ... de-pascal/

No
https://chessprogramming.wikispaces.com/GPU
http://chessgpgpu.blogspot.com/

Isn't there ANYTHING that can be done on the GPU?

It would be good if it could off-load SOMETHING from the CPU, like for example Monte Carlo during the late middlegame onwards - that might add some elo to an engine.

I know all chess programmers are saying the same thing: it's not possible, but it would be amazing if someone could creatively find a use for these powerful cards.

I remember Zeta chess using GPU

https://chessprogramming.wikispaces.com/Zeta

https://devtalk.nvidia.com/default/topi ... r-release/

JP.

5 million nps? That's good for those old cards.

It had been said & written hundred times that the GPUs are useless for computer chess....

Sad actually regards,
Dr.D

It's possible, but when GPUs are used as compute engines it is usually for floating-point calculations, and most chess programs use very little floating point.

--Jon

GPU won't work well with chess programming because each GPU kernel thread is designed to run floating instructions in a serial way, ideally without a loop and branching. Obviously, this is not the case for chess.