9
votes

After launching Ubuntu Server 14.04 LTS (HVM), I tried to install CUDA as following

wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_7.5-18_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1404_7.5-18_amd64.deb
sudo apt-get update
sudo apt-get install -y cuda
export PATH=/usr/local/cuda-7.5/bin:$PATH 
export LD_LIBRARY_PATH=/usr/local/cuda-7.5/lib64:$LD_LIBRARY_PATH
sudo reboot

Then I run a test

cuda-install-samples-7.5.sh .
cd ~/NVIDIA_CUDA-7.5_Samples/1_Utilities/deviceQuery 
make  
./deviceQuery

got the error:

modprobe: ERROR: could not insert 'nvidia_352': Unknown symbol in module, or unknown parameter (see dmesg) cudaGetDeviceCount returned 38 -> no CUDA-capable device is detected

What could be wrong here?

2
Did you found a solution ? - Anoracx
There is an issue with CUDA 7.5 and AWS gpu instances. Seems like Nvidia is looking into it. CUDA 7 should work just fine for now. - osondoar
@osondoar: You could add that as an answer to get this question off the unanswered list. - talonmies
cuda 7.5 on EC2 g2.2xlarge should now be usable with a 352.63 or later driver. - Robert Crovella
@RobertCrovella: Do you want upvote the Community wiki answer for this so it falls off the unaswered list (or add your own if you prefer and I'll reciprocate on the vote and delete mine). - talonmies

2 Answers

10
votes

Upgrade Ubuntu and install linux-image-extra before install nvidia-driver sudo apt-get update && sudo apt-get -y upgrade sudo apt-get install -y linux-image-extra-`uname -r`

2
votes

It would appear that this is a known issue within CUDA 7.5 on AWS gpu instances. The short term work around seems to be to downgrade to CUDA 7 for the moment.


Edited in December 2015 to add that Robert Crovella from NVIDIA advises that CUDA 7.5 on EC2 g2.2xlarge should now be usable with a 352.63 or later driver. So a driver update should resolve the issue.

[This answer was assembled from comments and added as a community wiki entry to get this question off the unanswered list]