Tensorflow Model training slower on better CPU

I’m just starting to get into ML and I have the problem that I have significantly longer training times for my model (mnist) on a Ryzen 7 PRO 4750U than on my old system with Intel i5 8th Gen. Does anyone have an idea what this could be fundamentally?


Just because the CPU is “better” or “newer” doesn’t mean it will lead to faster training times. My guess would be your old system might have used MKL which could explain the performance differences.

1 Like

You can try also AMD ROCM wheels:

1 Like

Hello, although tensorflow is universal, trained its models with CPUs, blah, I would advise you to orient yourself on a GPU, limit made for that