With ML workloads being so ram intensive, more than a few times I wished AMD properly supported the right ML packages. Their cost/GB is such a good deal vs Nvidia.