I followed the toturials on the official website to install TensorRT, converting Tensorflow graph and running inference on Nvidia GPU 1080Ti [CUDA 10.0 + CuDNN 7]. The results were disappointing as there is no speed improvements at all. I have tried 2 different models including Tensorflow version of YoloV3.
So, my question is: Does TensorRT supports Nvidia 1080Ti or not? If yes, what is the probable cause of this problem?
Thanks in advance
GeForce GTX 1080 Ti has CUDA compute capability of 6.1 and TRT support FP32 and Int8 on all the 6.1 compute capable devices.
Please refer below support matrix for more details:
Could you please share the script, model and data file so we can better help?