Cuda 10.0, cudnn 7.6.x, win10 only uses GPU memory, but no workload like 3D usage

I am new to cuda acceleration. I’m trying to run a small project using tensorflow-gpu 1.14.0 on a laptop with Geforce MX450.
| NVIDIA-SMI 457.63 Driver Version: 457.63 CUDA Version: 11.1 |
| GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
| 0 GeForce MX450 WDDM | 00000000:2B:00.0 Off | N/A |
| N/A 67C P0 N/A / N/A | 1862MiB / 2048MiB | 39% Default |
| | | N/A |

(P.S. is WDDM usage/Cap → N/A / NA normal?)

when I ran the training program, it occupied all GPU memory, but no other workload like 3D .

is there some configuration or installation wrong?

because I have another desktop with Quadro P620, running the same program, GPU workload is about 50% stably.

Thanks for any advice.

I think you are mis-reading this, due to the cramped presentation. The “TCC/WDDM” field indicates the card is in WDDM mode. The N/A / N/A results are in the "Pwr:Usage/Cap fields and just mean the card does not support monitoring these parameters.

If the output above from nvidia-smi was taken during the tensor-flow run, then it does indicate that the GPU workload is 39%.

1 Like

thanks for answer, glad to know the GPU is useful ! ^_^