RTX 4090 LAPTOP GPU and Tensorflow 2.13.0

Accelerated Computing CUDA CUDA Setup and Installation
1

It appears tensorflow 2.13.0 requires 11.8 and the earliest driver for the laptop GPU comes with 12.0

Does this mean I can’t yet use the RTX Laptop GPU with tensorflow 2.13.0?

Any workarounds to get this working locally …

Its for this laptop

2

The driver CUDA support is not what matters here. As long as the driver supports CUDA 11.8 or newer, it will be fine. You can install CUDA 11.8 toolkit with a driver for CUDA 12.0 and it will work.

  • leave your CUDA 12.0 driver alone
  • install the CUDA 11.8 toolkit, while deselecting the option to install the GPU driver
  • make sure you set your PATH and LD_LIBRARY_PATH correctly to point to the CUDA 11.8 install

from that point onward, you should be able to continue setup of TF 2.13

2 Likes
3

Thank you

Do I also need to install cudnn 8.6?

4

I got everything working wonderfully. Thanks for the clarification.

5

Does RTX 4090 Laptop GPU support CUDA ?