Tuesday, 17 May 2022
Eta - v0700
If I am convinced that a neural network driven BestFirstMiniMax would make a good search algorithm, why not just implement for CPU with NNUE? Maybe I will give it a try.
If I am convinced that a neural network driven BestFirstMiniMax would make a good search algorithm, why not just implement for CPU with NNUE? Maybe I will give it a try.
Some basic benchmarks of the Zeta NNUE AB framework showed 100 Knps to max 200 Knps throughput per worker for current GPUs, definitely too slow for an Eta v0501 hybrid approach considering modern CPUs with ~2 Mnps per core.