Geek Answers Handbook

Recommended Tensorflow GPUs

I understand that Tensorflow requires (for GPU calculation) a GPU with Nvidia Compute Capability> = 3.0. There are many such GPUs to choose from. Game-oriented GPUs, for example. GeForce is much cheaper than computationally oriented models, for example. Tesla. My limited work is that computationally-oriented models may lack video output (not required for computation) and that game models can do 32-bit math instead of 64. Assuming that Tensorflow uses (or prefers) 64-bit Does this mean that game models will not work or will produce missing results when used with Tensorflow? What attributes should I look for when choosing a GPU to use with Tensorflow?

+4
source share
2 answers

Gaming GPUs can work quite well. You need a very recent GPU with lots of memory and CUDA cores. Most people working on neural networks these days on the GPU use 32-bit floats.

+2
source

A version of TensorFlow that supports the GPU has the following requirements:

  • Linux 64-bit
  • Python 2.7
  • NVIDIA CUDA® 7.5 (requires CUDA 8.0 for Pascal GPUs)
  • NVIDIA cuDNN v4.0 (minimum) or v5.1 (recommended)

TensorFlow GPU support requires a graphics processor board with NVidia Compute> = 3.0 capabilities. Supported cards include, but are not limited to:

  • NVidia Titan
  • NVidia Titan X
  • NVidia K20
  • NVidia K40

Tensorflow

+3

More articles:


All Articles