Tensorflow r1.14 c++ API, very slow on RTX 2080


I was able to build and run C++ API r1.14 tensorflow version.

In Configure.py configuration (before build) I specified GPU 7.0.

I use RTX 2080 GeForce card, and expect to get a high Frame rate using a video from camera of my laptop.

When I use my PC internal NVIDIA GPU QUADRO P1000, the frame rate is 12 FPS.

However, the frame rate using RTX 2080 is about 6 FPS. Why is it so low ? How to speed it up ?

I disabled my PC internal NVIDIA GPU (P1000) , and verified that just RTX 2080 is running.

I also tried to speed up the GPU using the following code, but it doesn’t help:

     tensorflow::SessionOptions session_options;

tensorflow::ConfigProto* config = &session_options.config;
tensorflow::GPUOptions* gpu_options = config->mutable_gpu_options();


Do I understand correctly that the RTX 2080 is attached to your notebook? If so, please compare achieved Host2Device memory bandwidth for each of your GPUs using the CUDA bandwidth test sample. https://github.com/NVIDIA/cuda-samples/tree/master/Samples/bandwidthTest