Reduce FPS processing multiple video concurrently gpu 30% cpu 20%

[LionelLeee]

The simultaneous processing of multiple (n) video streams is compared with the processing of a single video stream. fps=single fps/n, why? GPU utilization is less than 30%, cpu utilization is less than 20%。 GPU:1080Ti CPU：inter xean E5-2620 v4 @2.10GHZ

[AlexeyAB]

show :

1. command that you use
2. screenshots with such information

   ./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights data/dog.jpg
   CUDA-version: 10000 (10000), cuDNN: 7.4.2, CUDNN_HALF=1, GPU count: 1
   CUDNN_HALF=1
   OpenCV version: 4.2.0
   0 : compute_capability = 750, cudnn_half = 1, GPU: GeForce RTX 2070
   net.optimized_memory = 0
   mini_batch = 1, batch = 8, time_steps = 1, train = 0
   layer   filters  size/strd(dil)      input                output
   0 conv     32       3 x 3/ 1    608 x 608 x   3 ->  608 x 608 x  32 0.639 BF

[LionelLeee]

[AlexeyAB]

GPU utilization is less than 30%,

Show screenshot of GPU-utilization in GPU-Z: https://www.techpowerup.com/download/techpowerup-gpu-z/

[LionelLeee]

[LionelLeee]

pre： post

[AlexeyAB]

Open Sensor Tab:

[LionelLeee]

[LionelLeee]

[AlexeyAB]

GPU utilization is less than 30%

GPU-load is 83%, not 30%.

[LionelLeee]

sorry ， I mean graphics card memory utilization.

[AlexeyAB]

Large GPU Memory isn't necessary for detection.

Large GPU Memory is required only for training.

[LionelLeee]

But the value I wrote is a bit big, in fact the graphics card storage only accounts for more than 10%. Is the fps drop because the load of the graphics card is too large?

[LionelLeee]

@AlexeyAB Is the fps drop because the load of the graphics card is too large?

[AlexeyAB]

Because it is impossible to handle infinite streams with infinite total FPS.

[LionelLeee]

How to solve this problem? Thread or process? Or replace with a better GPU? Or other methods?

[AlexeyAB]

Or replace with a better GPU?

Yes, 2080Ti or 3080Ti.

[LionelLeee]

OK,Thank you for your reply