hello, when i infer YOLOv5 for detect some object shows high CPU usage
(about 45~50% in task manager).
but GPU usage very low (“gpu 3D” about 4% and “gpu cuda” about 35~40%. in task manager)
CPU : 11th Gen Intel(R) Core™ i7-11700K @ 3.60GHz 3.60 GHz
GPU : RTX 3090 x 1
CUDA : 11.1
Im using libtorch for load YOLOv5 model and " torch::cuda::is_available() " function returns “TRUE”
Question. ===========================================
- Is it normal?
- What is the difference between “GPU 3D” and “GPU cuda” in task manager?
- Why only high GPU cuda(35~40%) > GPU 3D(4%)
- How to increase GPU usage and decrease cpu usage?
5.Judging from the 35-40% gpu cuda usage, it doesn’t seem like the gpu isn’t being used. is true?
6. When same code run different env shows “gpu 3D” usage about 70~75% and cpu usage about under 10% why its happend…?
(Diffent ENV : RTX 3060 and same cpu)
please let me know … thank you!