I researched this online and found this, someone managed to compile their own libopenblas and got some x10 times speedup. But I tried their dll and Leela is still slow.
This is what I see:
Code: Select all
OpenBLAS [DYNAMIC_ARCH NO_AFFINITY Penryn].
OpenBLAS found 4 Penryn core(s).
OpenBLAS using 1 core(s) for this backend.
BLAS max batch size is 256.