CUDA Toolkit Installation Issue with CentOS7.

Accelerated Computing CUDA CUDA Setup and Installation
1

I successfully installed NVidia Graphics Drivers - NVIDIA-Linux-x86_64-390.87.run
Then I tried installing CUDA Toolkit from - cuda-repo-rhel7-9-2-local-9.2.148-1.x86_64.rpm

Most of the install went well, but I got the following errors -WARNING: Your driver installation has been altered since it was initially
installed; this may happen, for example, if you have since
installed the NVIDIA driver through a mechanism other than
nvidia-installer (such as your distribution’s native package
management system). nvidia-installer will attempt to uninstall as
best it can. Please see the file ‘/var/log/nvidia-uninstall.log’
for details.

Installing : 1:xorg-x11-drv-nvidia-396.37-1.el7.x86_64 31/70
Error unpacking rpm package 1:xorg-x11-drv-nvidia-396.37-1.el7.x86_64
error: unpacking of archive failed on file /etc/vulkan/icd.d/nvidia_icd.json;5b87486f: cpio: rename
Installing : libICE-devel-1.0.9-9.el7.x86_64 32/70
error: xorg-x11-drv-nvidia-1:396.37-1.el7.x86_64: install failed
Installing : libSM-devel-1.2.2-2.el7.x86_64 33/70
Installing : libXt-devel-1.1.5-3.el7.x86_64 34/70
Installing : libXmu-devel-1.1.2-2.el7.x86_64

Towards the end I noticed -

Verifying : 1:xorg-x11-drv-nvidia-396.37-1.el7.x86_64

Dependency Installed:
…xorg-x11-drv-nvidia-gl.x86_64 1:396.37-1.el7

However at the ver last line -
Failed:
xorg-x11-drv-nvidia.x86_64 1:396.37-1.el7

Complete!

I am not sure how to interpret this. Did the xorg-x11-drv-nvidia get installed or not. If not how do I fix this??
Please advise…

Thanks,
Raj

2

you’ve mixed runfile install:

NVIDIA-Linux-x86_64-390.87.run

with package manager install:

cuda-repo-rhel7-9-2-local-9.2.148-1.x86_64.rpm

this is a no-no.

Read the linux install guide.

3

Can you point me the the install guide
Thanks,
–Raj

4

If I google:

cuda linux install guide

it is the very first hit

note:

I’m not trying to be a jerk here. I could easily give you the link with less time than I’m spending now writing this.

But in the future, if you need to use it again, you’ll either have to find your way back to this thread or ask again. This way, hopefully you will learn that it is easy to find, and this will save you time in the future when you need to find it again. Because I bet you can remember to google that.

5

Aha… got it… Thanks !!!
–Raj

6

by the way, you may still run into the same issue. But even if you got past that issue, you probably wouldn’t have a working setup due to the runfile/package manager issue.

If you still run into the same issue, you could try a runfile install for the CUDA toolkit.

read the linux install guide.

7

Thanks, I understand… I should know better than to have to be spoon fed ;-)
I am trying to remove the broken installation and notice that the uninstall script -
/usr/local/cuda-9.2/bin/uninstall_cuda_9.2.pl
is missing. Plan is to completely remove the current install and walk through the install guide (meticulously adhering to the procedures to create a working install).
Can you send me the script?

Thanks,
Raj

8

The uninstall script only comes with the CUDA runfile installer. You installed CUDA via package manager.

There is no script for that. You should use package manager commands to remove CUDA.

(hint: read the linux install guide, and/or google, hint: “package manager remove cuda”)

For example, read this section of the linux install guide, carefully:

[url]Installation Guide Linux :: CUDA Toolkit Documentation

(the thing you installed via runfile installer is not CUDA, it is a GPU driver)

9

@Robert_Crovella
Yes, you are being a jerk here.