which gpu arch begin to fp8 support

> FP8_E4M3 and FP8_E5M2 input, output, and intermediate tensor data type is only available with compute capability 8.9 and 9.0. https://docs.nvidia.com/deeplearning/cudnn/backend/v9.3.0/developer/graph-api.html <img width="1039" height="705" alt="Image" src="https://github.com/user-attachments/assets/3ea5474e-a842-4618-8e71-d08517780f20" /> https://docs.nvidia.com/cuda/cuda-programming-guide/05-appendices/compute-capabilities.html