Hello,
The nvidia-smi command on my Jetson Nano isn’t working, telling me the command is not found. Do I have to install some driver or how can I get it to work?
And, when I’m already asking: Is the Cuda Toolkit and cuDNN on the Jetson Nano already installed and configured or do I have to do it myself?
Thanks for the help in advance!
Hi @nemestomi2,
I think nvidia-smi is only available for Desktop GPUs. For Jetson boards you can use tegrastats, for example:
sudo ./tegrastats
Did you use the pre-built SD card image and followed this Getting Started with Jetson Nano guide from NVIDIA? https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit
You can also use the Nvidia SDK Manager to install CUDA and its dependencies.
Regards,
Fabian
Thanks @fabian.solano for the suggestion!
The only file I found was in /usr/bin
and does not print the info I’m looking for. I’m looking for the architecture version of the GPU.
Yes, I use the prebuilt image exactly as described in the Getting Started guide and have CUDA 10.0 already installed. I just wanted to be sure I have those packages, too, because I don’t know wether they come together with CUDA or I have to install them manually. So are they installed on Jetson Nano already?
You can get the architecture version of the GPU from the following page CUDA GPUs - Compute Capability | NVIDIA Developer in the CUDA-Enabled Jetson Products tab
Nano is “-gencode arch=compute_53,code=sm_53
”.