Cannot use GPU with Tensorflow

0
votes

I've tensorflow installed with CUDA 7.5 and cuDNN 5.0. My graphics card is NVIDIA Geforce 820M with capability 2.1. However, I get this error.

Ignoring visible gpu device (device: 0, name: GeForce 820M, pci bus id: 0000:08:00.0) with Cuda compute capability 2.1. The minimum required Cuda capability is 3.0.
Device mapping: no known devices.

Is there any way to run GPU on a 2.1 capability? I scoured online to find that it is cuDNN that requires this capability, so will installing an earlier version of cuDNN enable me to use GPU?

2
tensorflowgpucompute-capability
You are correct, Geforce 820M is a GPU with compute capability 2.1. I am reasonably sure that all versions of cuDNN require compute capability >= 3.0, so the answer to your question would appear to be "no". Based on the specification at Wikipedia this GPU is low-end and likely wouldn't provide much acceleration relative to a modern CPU if it were supported (I am surprised that NVIDIA launched a Fermi-based GPU as late as 2014). - njuffa
I have the same problem! How did you manage?? - MrRobot9
Install tensorflow-cpu - Keerthana Gopalakrishnan

2 Answers

6
votes

tensorflow-gpu requires GPUs of compute capability 3.0 or higher for GPU acceleration and this has been true since the very first release of tensorflow.

cuDNN has also required GPUs of compute capability 3.0 or higher but since the very first release of cuDNN.

-1
votes
config = tf.compat.v1.ConfigProto(gpu_options = tf.compat.v1.GPUOptions(per_process_gpu_memory_fraction=0.8))
config.gpu_options.allow_growth = True
session = tf.compat.v1.Session(config=config)
tf.compat.v1.keras.backend.set_session(session)

Try placing this code before your python code. Also make sure you have all the cuDNN dlls installed into your CUDA toolkit (if on Windows then you should install the dlls to Program Files > NVIDIA GPU computing toolkit > CUDA > version x.x directory). The documentation can be found on the NVIDIA website for developers: https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html