-1

I am interested in getting a 2nd hand Nvidia Tesla P40 GPU. Before I buy it, I want to check if TensorFlow version 2.x supports it.

I use tensorflow version 2.x on Python 3.6.6

My question is this: Does TensorFlow version 2.x support this card? (Tesla P40).

I did an online search and found the following:

"TensorFlow 2.x does support Compute Capability 6.1, which includes the Nvidia Tesla P40 GPU" https://docs.nvidia.com/deeplearning/deeplearning/frameworks/tensorflow-release-notes/rel_19.06.html

"if you are running on Tesla (Tesla V100, Tesla P4, Tesla P40, or Tesla P100), you may use NVIDIA driver release 384.111+ or 410"

"Release 19.06 supports CUDA compute capability 6.0 and higher. This corresponds to GPUs in the Pascal, Volta, and Turing families. "

"Up to release 19.08 supports this card." https://docs.nvidia.com/deeplearning/frameworks/support-matrix/index.html

On further reading into what Release 19.08 means, I discovered it is contained in a docker with TensorFlow version 1.x (which is undesirable to me. I prefer TensorFlow version 2.x).

Just wondering if anyone has this GPU and has it running OK with TensorFlow version 2.x please? Thanks a lot. Do you like this GPU?

I looked also at Nvidia Geforce 3060, but is more expensive and has fewer CUDA and smaller RAM than the P40.

Additional question is this: When you accessed the P40 GPU from Python, did you need to install tensorflow-gpu separately? Or was pip install tensorflow alone enough?

Thank you.

George
  • 121
  • 7
  • This is no code-related question, and you answered your own question in the text. When installing TensorFlow, just follow the official guideline. Also, the 3060 has Tensor Cores and is cheaper where I looked.. – mhenning Aug 14 '23 at 09:24
  • There's big green box on the nvidia website saying: "Starting with the 23.06 release, containers will no longer be tested on Pascal GPU architecture.". The P40 is Pascal arch – PeterM Aug 28 '23 at 23:23

0 Answers0