CPU mode faster than GPU mode??

[Timo-hab]

I am running the yolo_deploy.prototxt and get a runtime of 0.36 s for the net.forward_all. On GPU it takes 0.7 s for the same image (500x250px). How can it be that the runtime takes longer in GPU mode (2 times)? Can this experience anyone confirm? Thank you in advance!

Edit: Do i need a certain version of cudnn?

[xingwangsfu]

Hi, could you provide more information, such as the type of CPU and GPU you used? cuDNN would greatly accelerate the gpu computation.

[Timo-hab]

Hi CPU: Intel Xeon E5 GPU: Titan X CUDNN: v5.1 i guess I do not think it is up to the hardware. Other CNN-Models are running a lot faster on GPU related to CPU.

[xingwangsfu]

Hi, I tested it by myself. Given an input image of size 768x576, the running time on Intel i7 CPU is about 2 s for forward_all. On GPU Titan X with Cuda 7.0 and Cudnn 4, the running time is about 0.2 s. I updated the yolo_main.py file if you want to test it by yourself.