Hello,
I’ve been trying to install CUDA to do some parallel programing on a Tesla K40m GPU with a Ubuntu 18.04 host computer. I’ve been trying to follow the “Linux Installation Guide” and it seems like I’ve run into a roadblock.
I’ve downloaded the CUDA toolkit and I can see new directories such as:
“/opt/nvidia”
“/usr/local/cuda”
and i’ve compiled the code in the “/usr/local/cuda-11.0/samples” directory and installed it with:
sudo ./cuda-install-samples-11.0.sh /home/accmeint/CUDA
I can run “lspci | grep -i nvidia” and see my GPU…but this is where my luck ends.
When I try to run the “deviceQuery” command I get the following:
////////////////////////////////////////////////////////////////////////////////////////////////////
./deviceQuery
./deviceQuery Starting…
CUDA Device Query (Runtime API) version (CUDART static linking)
cudaGetDeviceCount returned 100
→ no CUDA-capable device is detected
Result = FAIL
////////////////////////////////////////////////////////////////////////////////////////////////////
…Looking in “/dev/” for something “nvidia” was a dead end so I tried installing the K40m driver.
by downloading the K40m driver for Ubuntu and running:
sudo ./NVIDIA-Linux-x86_64-450.51.06.run
This fails because it complains about conflicting with a driver that already exists…but there is nothing in /dev/nvidia??
I don’t know how to proceed…any insight would be greatly appreciated!
Thank you,
Ed