Nearly 90% of MTP's speedup data has only been tested on Apple Silicon, and there is no equivalent public data for NVIDIA and CUDA scenarios.