Is windows TCC performance the same as native linux?

I’ve been trying to find more benchmarks on TCC performance. I reached out to sales as well. With small models like yolov5 nano, inference on windows is like 10-20ms vs 3-6ms. Interestingly enough the gap is still there with larger models, just less.

I’m specifically wondering if the quadro cards can perform on windows natively vs their consumer rtx variants on linux.