Why jetson nano not support int8

I have searched some topics and it says that the jetson nano not support int8 inference.


But I am wondering why it can not support?

Hi lixibo456, Jetson Nano is based on TX1 SoC with integrated Maxwell GPU, which doesn’t support INT8 inference in hardware (like Jetson TX1 and TX2 it does support FP16 however). Jetson Xavier with it’s integrated Volta GPU does support INT8 inference.

is this both Xavier NX and Xavier AGX ?

Yes, NX is basically a stripped down AGX (less CUDA cores, less RAM) in a Jetson Nano form factor