OpenBLAS is faster than Intel MKL on AMD Hardware (Ryzen)

There are a lot if scientific computing benchmarks here:

[On Newegg] the Intel process costs $1980 and the AMD $900, under half as much.

It has Numpy, Scikit-Learn, Matrix multiplication, GMP, FFTW, glibc (eg, tanh, sqrt), and lots of other things doing some sort of numerical computing like “MrBayes” primate phylogeny.

But no OpenBLAS with AMD vs MKL with Intel. I agree, it would be cool.
Of course, with most applications (eg, Julia) the majority of software will be the same between them.