0

I have just started working on this quite big dataset with my new and shiny GTX 1080. I have installed the drivers as well as cuda, cuDNN etc.

Observation:

When I run nvidia-smi I get the following picture:

nvidia-smi

enter image description here So, both of the GPUs are using loads of memory ( That is good, I think? ) however, the GPU utilization is very low for both of them, especially the second GPU.

Explanation: Why?

Any tips on how I can improve the performance as well of someone know why the GPU utilization is so low?

talonmies
  • 70,661
  • 34
  • 192
  • 269
Hugo
  • 135
  • 2
  • 9
  • Which "this quite big dataset" are you using? – D.Laupheimer Mar 16 '17 at 15:32
  • @D.Laupheimer the comma.ai dataset on github :) I actually ran the code as well just because I wanted to try it out and how fast my new GPUs were, since I read in the comment below that keras does not provide the use of multiple GPUs I guess I have to write something myself. – Hugo Mar 16 '17 at 19:40
  • Yep, you have to do so - or you use the mentioned solution below. – D.Laupheimer Mar 17 '17 at 07:14

1 Answers1

1

First of all, the nvidia-smi may be misleading. Have a look at the discussion: nvidia-smi Volatile GPU-Utilization explanation?

Second: Unfortunately, Keras doesn't provide an out-of-the-box solution for using multiple GPU. Keras uses automatically (using TensorFlow as backend) one GPU. You have to use TensorFlow directly if you want to make use of both GPUs. If you are using Keras version 1.x.x you can try this solution: Transparent Multi-GPU Training on TensorFlow with Keras.

Community
  • 1
  • 1
D.Laupheimer
  • 1,074
  • 1
  • 9
  • 21