Hallo,
does someone maybe know if the RTX 4070ti got the CUDA compability that u can train tensorflow models on this gpu?
Id be thankful for any advices! 
Welcome to the Tensorflow Forum!
The RTX 4070Ti GPU card support starts with cuDNN 8.7.0. At the moment the latest stable version TF 2.11 is not compatible.
Thank you!
2 Likes
hey, ty for ur response!
ive downloaded cuda 12 and cuDNN 8.8 and added these to my path variables:
C:\tools\cudnn-windows-x86_64-8.8.0.121_cuda12-archive\bin
C:\tools\cudnn-windows-x86_64-8.8.0.121_cuda12-archive\include
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.0\extras\CUPTI\lib64
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.0\extras\CUPTI\include
then ive created a conda virtual environment to install tf-nighty versions there, and use it afterwards. It automatically installed this one “2.13.0-dev20230212”. Im not sure if this is able to work, but in my case it didnt:
Do you maybe have an idea why tensorflow doesnt detect my gpu? 
greetings
Could you please let us know the steps that you have followed to install Tensorflow?
Thank you!
In a pycharm-project ive created a new environment using the terminal:
1.: conda create --name myEnv
2.: conda activate myEnv
3.: pip install tf-nightly
4.: python -c “import tensorflow as tf; print(tf.version)”
→ 2.13.0-dev20230214
5.: python -c “import tensorflow as tf; print(len(tf.config.experimental.list_physical_devices(‘GPU’)))”
→ 0
Before that ive downloaded cuDNN 8.8 for Cuda 12.0 here: cuDNN 9.2.1 Downloads | NVIDIA Developer
and added these 2 paths to my path variables:
C:\tools\cudnn-windows-x86_64-8.8.0.121_cuda12-archive\bin
C:\tools\cudnn-windows-x86_64-8.8.0.121_cuda12-archive\include
and Cuda Toolkit 12.0 here: CUDA Toolkit 12.5 Update 1 Downloads | NVIDIA Developer
and added these 2 paths to my path variables:
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.0\extras\CUPTI\lib64
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.0\extras\CUPTI\include
This prints false as well if that helps:
python -c “import tensorflow as tf; print(tf.test.is_built_with_cuda())”
but i think it should print true, no? Thanks in advance for helping me out!
Here it doesn’t print True, because tensorFlow 2.10 was the last Tensorflow release that supported GPU on native-Windows.
Starting with TensorFlow 2.11 you will need to install TensorFlow in WSL2.
Thank you!
1 Like
So i can downgrade to tensorflow 2.10 to use my GPU in tf? Or should i wait for tf 2.12?
Because you are using Windows, i advise you to use TF 2.10 to utilize the GPU.
Thank you!
Okay ty ima try this. Can you btw tell me approximately when tf-2.12 will be released?
Greetings
Last week, TF 2.12.0 rc0 was released. The final version (TF 2.12) is scheduled for release at the end of this month.
Thank You.
Sorry I have question of it , I use the 4070ti on windows 11. However, I cannot use the tensorflow with 2.10.0. Is it must download to cuda 11.2 and cudnn 8.2? But my GPU is 4070ti that you say starting with 8.7 version. Hoever, the advanced version 2.11.0 will not support it. That what should I download if I must use the tensorflow to train my GPU deep learning…