I have not deployed an instance of gpu yet, but I can say that the OS image already has driver settings for you.
Now, from the point of view of installing CUDA, another element was used to run your code. If you have not tried EC2 at all, I can tell you about a regular instance - I can install gcc / g ++ and svn; set up the repository and run my code in 5-10 minutes.
EDIT: I looked through the documentation and found this: http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/Cluster_GPUs_Install_Driver.html#d0e18924 . This indicates reinstalling or updating the NVIDIA drivers.
source
share