Can´t follow ubuntu process to install not working

Installation & Getting Started
#1

Hi,
I´m triying to install in a ubuntu 20.04 the GPU

following this instructions. GPU

My Hardware is a R730 with a tesla K80.
Software.- Vmware with esxci and passtrg. Tesla is working well with a Win10 VM i test it.

So.following the process i just download. the drivers as mention in the instructions.

get https://uk.download.nvidia.com/tesla/470.103.01/nvidia-driver-local-repo-rhel7-470.103.01-1.0-1.x86_64.rpm

I go to the next step sudo apt install nvidia-XXXX in my case will be
sudo apt install nvidia-driver-local-repo-rhel7-470.103.01-1.0-1.x86_64.rpm.1

Response is can´t localize the packg.

I iterate with rpm -i nvidia-driver-local-repo-rhel7-470.103.01-1.0-1.x86_64.rpm
with no success.

rpm: RPM should not be used directly install RPM packages, use Alien instead!
rpm: However assuming you know what you are doing…
warning: nvidia-driver-local-repo-rhel7-470.103.01-1.0-1.x86_64.rpm: Header V3 RSA/SHA512 Signature, key ID 7fa2af80: NOKEY

error: unpacking of archive failed on file /etc/yum.repos.d/nvidia-driver-local-rhel7-470.103.01.repo;62136264: cpio: open failed - File exists
error: nvidia-driver-local-repo-rhel7-470.103.01-1.0-1.x86_64: install failed

So at this point a don´t know how to continue wit it.
Thanks for the help!

#2

Hi,

I noted a discrepancy on the download link on our docs, so you are redirected to the driver download site. The correct link to download cuda libraries and the driver would be

wget https://developer.download.nvidia.com/compute/cuda/11.4.1/local_installers/cuda_11.4.1_470.57.02_linux.runsudo 
sh cuda_11.4.1_470.57.02_linux.run

If you prefer running DEB packages, change the option from run file local, and follow the provided instructions.

Anyway, you don’t need to install the cuda package to run out database, but you need just the driver.

you can try to use this command

ubuntu-drivers devices

on my Ubuntu 20.04 the output is like that

== /sys/devices/pci0000:00/0000:00:03.1/0000:09:00.0 ==
modalias : pci:v000010DEd00001E07sv000010DEsd000012FAbc03sc00i00
vendor   : NVIDIA Corporation
model    : TU102 [GeForce RTX 2080 Ti Rev. A]
driver   : nvidia-driver-418-server - distro non-free
driver   : nvidia-driver-450-server - distro non-free
driver   : nvidia-driver-510 - distro non-free recommended
driver   : nvidia-driver-470-server - distro non-free
driver   : nvidia-driver-470 - distro non-free
driver   : xserver-xorg-video-nouveau - distro free builtin

to check which drivers are available, then install a driver with apt

sudo apt install nvidia-driver-470

then reboot, and check if everything is ok with the nvidia-smi command, as suggested by the docs

Nvidia dropped the support of K80 since 495 (?), so pay attention to which driver you are installing; anyway, the Kepler arch has some limitations so that some queries would run slower than Pascal, Volta, or Turing.

If you want to install the driver downloaded from the Nvidia website, please choose the Linux-64 install from the site https://www.nvidia-com/downloads

Select Linux 64 bit, search, download the file, change the permission, and run the file.

The file name in your case should be NVIDIA-Linux-x86_64-470.103.01.run.

Let me know if you have succeeded installing the driver

Candido