4070Ti doesn't work with tensorflow2.15

AI & Data Science Deep Learning (Training & Inference) TensorRT
,
1

I am experiencing a issue with TensorFlow 2.15 on my system. I have been unable to get TensorFlow to recognize my GPU, and I thought sharing my setup and steps I’ve taken might contribute to finding a solution.

Here are the specifics of my setup:

  • Operating System: Windows 11 Home
  • Python Version: 3.10.1
  • TensorFlow Version: 2.15
  • CUDA Version: 12.3 (also tried 12.2)
  • cuDNN Version: 8.9
  • GPU: NVIDIA GeForce RTX 4070 Ti
  • GPU Driver Version: (546.33-desktop-win10-win11-64bit-international-nsd-dch-whql)

Despite correctly setting up the environment variables and restarting my computer after each configuration, tf.config.list_physical_devices('GPU') returns an empty list, and testing cuDNN installation returns false.

It’s interesting to note that this problem persists with TensorFlow 2.15&2.14, as I have taken all recommended steps to ensure proper installation and configuration. If anyone has insights or suggestions on what might be going wrong, it would be greatly appreciated.

Thank you for your time and help!

2

Hi @31675931,
I always had bad luck using default packages for ML stuff, for me what always works best is to compile locally, check here for more details. However checking that link at the end it states the following:

image
image1275×113 22.5 KB

So I would suggest to move to WSL and try there, first you need to install it and then install CUDA there. Check this links for more details.

Also check the limitations to see if it might be a deal breaker for you use case:

image
image1135×945 138 KB

Regards,
Andres
Embedded SW Engineer at RidgeRun
Contact us: support@ridgerun.com
Developers wiki: https://developer.ridgerun.com/
Website: www.ridgerun.com