AWS Thinkbox Discussion Forums

Install Nvidia drivers on an linux instance

Hey,

Im using a p2.xlarge instance, with a k80 gpu
When i try and install these drivers:
https://us.download.nvidia.com/tesla/470.129.06/NVIDIA-Linux-x86_64-470.129.06.run

It just complains that i dont have the kernel source files installed. I have tried installing various packages with out luck (eg kernel-devel), but it’s still not happy.

I can’t find a AMI with drivers that are new enough for Redshift (495 or higher)

Any ideas?

These are really old, do you think RS will run on them with an old unsupported driver too? Don’t you have access to newer GPU instances?

It’s just a cheap instance while im testing. Planning on using p3.16xlarge which I think is x8 a100s.

Anyone got Redshift running on an linux instance? What did you do?

what about using the CPU option to test a simple render works as it supposed to, and check UBL check out etc, then prep a more powerful machine with a single card and test that with a proper benchmark scene?
https://help.maxon.net/r3d/maya/en-us/Content/html/Redshift+CPU+Rendering.html?TocPath=Redshift%20CPU|_____0

1 Like

Yeah, but ultimately i want to render with a GPU, and I need to work out how to install the latest GPU drivers for that. Im assuming this issue will be the same regardless of the the gpu.

I find that the aws docs are straightforward:

We usually use the g4 or g5 instance types, so I don’t have any first-hand experience with the p3 instances. In general, the installation procedure should be similar.

nvidia recommends installing first (e.g. centos):
sudo yum install kernel-devel-$(uname -r) kernel-headers-$(uname -r)

I seem to remember that you may need to install libglvnd-devel first before running the nvidia run file.

You did not mention if you made your own AMI or are using one of Deadline’s AMIs, so other required libs/deps might already be installed.

Privacy | Site terms | Cookie preferences