u have to code it for your self .... nvidia sell's u more work, not a complete solution, use rather ROCm, install driver and up u go with parallel processing .... if u re not coder, CUDA is useless (performance only on paper for over-overpriced GPU) ... throw away money for something that u need to use but something that u cant use anyway, designed that way to the core .... why u think they sell complete proprietary AI solutions for obnoxious amount of money (closed architecture so u re dependent on it) ? Think again ! They don't want people to run what so ever parallel processing on home workstations or powerhouse home PC's ! They only pretend for doing so, that's why people have so much headaches with it ! And open source is crap and junk at supporting parallel processing, i have try to use it, instead of dong data mining and testing models its all about the code and about code and not end user end results ! YES its not a RANT, its a TRUTH !