Hi everybody,
i want to enable GPU to use tensorflow on my notebook but it seems not be recognized by jupyter.
I have:
NVIDIA GeForce GT 540M
Windows 7-64 bit
CUDA 9.0
cuDNN 7.6.5
Anaconda
tensorflow-gpu 1.14.0
keras 2.4.3
could anyone help me? i found a topic that say the minimum needed GPU CUDA capability for tensorflow is 3.0
I think this is kind of the same issue. But the problem seems that the GPU is not supported as per people in that thread. I have no problem believing that considering the GPU is over 10 years old now. I would recommend 
using Google Colaboratory for a very similar experience to jupyter notebook. Colab Pro is also a viable option when you are ready to do things fast. Even I use it although I have a powerful GPU locally.
Also there are tons of tutorials and getting started videos that show you how to use it.
Hope this helps
@Luca_Zanetti , I think the problem here is with CUDA & cuDNN version.
You can see here the corresponding supported CUDA & cuDNN versions for each version of tensorflow.
After installing CUDA & cuDNN, you can check whether your tensorflow uses GPU or not with below command.
python -c "import tensorflow as tf; print(assert tf.test.is_gpu_available())"
I’ve already seen this compatibility table but the GPU doesn’t work with some configuration of cuda and cuDNN.
Python doesn’t recognize it.
Hello, a class 5xx GPU is more than obsolete. I know that we are suffering the full force of the silicon crisis. But this is a bit exaggerated, think about two things: a desktop and Linux
the desktop is stable and updates easily
a desktop GPU costs much less than a laptop has lasted a limited life
Linux is stable and updates easily and the command line is easy to learn, not to mention the millions of tutorials on youtube