How to solve compatibility conflict of OS(Ubuntu 20.04) and GPU(Quadro T500)

Accelerated Computing CUDA CUDA Setup and Installation
1

Hello,
I want to use keras and cuda on my device running with Ubuntu 20.04 and with a Nvidia Quadro T500 GPU.
According to the following links, my GPU capability only support cuda toolkit 10.0 which is not available for Ubuntu 20.04, but since I am using a work laptop, it is not so easy to just downgrade to 18.04.
https://developer.nvidia.com/cuda-gpus
https://www.tensorflow.org/install/source#gpu

Can anyone tell me how to solve that conflict?

Also (but not really important) I am wondering, why on the nvidia website they talk about cuda compatibility which shows 7.5 for my T500, when on the tensorflow website 7.5 seems to refer to the cuDNN version.

2

Quadro T500 is compute capability 7.5

That number doesn’t refer to CUDA version (a software version) cuDNN version (a software version) or any other software version. It is a GPU architectural family number.

That GPU is supported by all recent CUDA toolkits up through and including CUDA 12.3.

1 Like
3

Well, that is explaining a lot and great news. I was somehow misled because it was listed here https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html under “2.1 verify that you have a CUDA-capable GPU”.

Thanks for the quick answer.