CUDA 6.5 RC doesn't work on Ubuntu 14.04, AWS EC2, g2.2xlarge instance

I wanted to use CUDA for computing acceleration. I launched a g2.2xlarge instance (with NVIDIA GRID K520) on AWS EC2. I installed the Ubuntu 14.04 (hvm) AMI and updated to the latest.

As CUDA 6.0 doesn’t provide an installer for Ubuntu 14.04, I downloaded CUDA 6.5 RC and installed it following the instructions. Everything went OK.

I rebooted. For whatever I try to do afterwards I always get this error

modprobe: ERROR: could not insert ‘nvidia_340’: Unknown symbol in module, or unknown parameter (see dmesg)

There’s nothing like /dev/nvidia*. Apparently the module is never started by the Linux kernel.

Did I miss anything? Do I need to do sudo apt-get install ubuntu-desktop beforehand? I thought CUDA would install all the packages it depends on, e.g. xorg-server, etc?

Thanks in advance.

Are you doing this as an ordinary user?
What happens if you run

nvidia-smi -a

as root user, then re-run your tests?

@txbob I already removed entirely the cuda package so I have no way to tell you. Sorry…

However CUDA 6.0 works perfectly for Ubuntu 12.04.5 on g2.2xlarge.

I am having the same issue.
While logged in as root, the error given when issuing the aforementioned is as follows

modprobe: error: Could not insert ‘nvidia_340’: Unknown symbol in module, or unknown parameter (see dmesg)
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

try installing the cuda toolkit (including driver) from a .run file rather than via a repo

I followed this verbatim. couple errors, but nothing that really stood out.