[kjs3]

At this point the better answer is invest in CUDA parity on other platforms, not throwing out the ecosystem to spite nVidia.