3 ms·Certain chips can work with useful local models, but compatibility is far behind CUDA.by trenchpilgrim 1y agoCertain chips can work with useful local models, but compatibility is far behind CUDA.