Nvidia smi linux install

How to install nvidia-smi?

How can I install nvidia-smi? I installed CUDA and nvidia-352 driver but unfortunately nvidia-smi is not installed.

sudo apt-get update sudo apt-get install nvidia-smi E: Unable to locate package nvidia-smi 

@markkirby, I installed it like what nvidia site said. ‘sudo dpkg -i cuda-repo-ubuntu1404_7.5-18_amd64.deb’ ‘sudo apt-get update’ ‘sudo apt-get install cuda’

Did you not also install the driver ? You must install both CUDA and the driver as nvidia-smi is packaged with the driver.

@markkirby no, I completely installed nvidia-352, Even I installed nvidia-340 but there is no nvidia-smi. My laptop has Intel 3000 and Nvidia 525M, is it important?

Yes you have a hybrid GPU right ? Intel and Nvidia ? First remove everything you installed with sudo apt-get purge nvidia* Then see here for the driver instructions askubuntu.com/questions/452556/…, then reinstall cuda, please see here for bumblebee docs, bumblebee-project.org/install.html

1 Answer 1

nvidia-smi binary symlink created when you install the nvidia driver, as nvidia-340 in my case. Take a look:

darlene ➜ ls -la `which nvidia-smi` lrwxrwxrwx 1 root root 34 Mai 24 20:52 /usr/bin/nvidia-smi -> /etc/alternatives/x86_64-linux-gnu_nvidia_smi darlene ➜ ls -la /etc/alternatives/x86_64-linux-gnu_nvidia_smi lrwxrwxrwx 1 root root 34 Mai 24 20:52 /etc/alternatives/x86_64-linux-gnu_nvidia_smi -> /usr/lib/nvidia-340/bin/nvidia-smi 

And the final binary comes from:

darlene ➜ dpkg -S /usr/lib/nvidia-340/bin/nvidia-smi nvidia-340: /usr/lib/nvidia-340/bin/nvidia-smi 

Источник

How to Install Nvidia Drivers on Debian Linux 11 or 10

Most users always seek to get more performance out of their graphics cards. One of the better ways is to install Nvidia Drivers, which depending on your installation method, comes with an open-source variant. The following tutorial will teach you how to install Nvidia Graphic Drivers on Debian 11 Buster or Debian 11 Bullseye repositories and to install Nvidia drivers directly from Nvidia repositories, with open-source and proprietary drivers available.

Update Debian

Before proceeding with the tutorial, ensuring your system is up-to-date with all existing packages is good.

Optionally, you can list the updates for users who require review or are curious.

Proceed to upgrade any outdated packages using the following command.

Previous Nvidia Installations

Please note installing Nvidia drivers over the standard packages can result in unexpected issues, especially if you install the Nvidia drivers from Debian’s repository, then import the Nvidia Cuda repository and try to re-install or upgrade directly over those packages. If you use one method, stick to it, and if you want to swap, ensure you remove the previous installation to avoid conflicts.

Below I will briefly go through some options for removing the previous Nvidia Drivers.

First, use a blanket command to remove any traces of Nvidia installation on your system.

sudo apt autoremove nvidia* --purge

The second method is for anyone that installed the Nvidia drivers using the “.run” file, a method I use suggest as an installation method but is shared amongst other Linux blogs. Given the alternatives using the Nvidia Cuda repository, I recommend using this over runfile installations.

To remove the runfile type of installation, use the following command.

sudo /usr/bin/nvidia-uninstall

Lastly, remove the CUDA toolkit runfile installation using the following command.

sudo /usr/local/cuda-X.Y/bin/cuda-uninstall

These few methods may remove the Nvidia Drivers.

#1st Method: Install NVIDIA Drivers: Debian Repository

Installation Pre setup

The first task is enabling the “contrib” and “non-free” repositories to your Debian repositories. This can be quickly done using the following commands.

First, add the contrib repository.

sudo add-apt-repository contrib

Now add the non-free repository.

sudo add-apt-repository non-free

Once added, update your repository list.

Before installing any Nvidia drivers, you will need to install the proper kernel headers for the NVIDIA driver to build with. You can install either 64bit or 32bit, and you can find this out by running the lscpu command:

CPU op-mode(s): 32-bit, 64-bit CPU(s): 2 On-line CPU(s) list: 0,1 CPU family: 23 CPU MHz: 3900.025 NUMA node0 CPU(s): 0,1

From the output, the system is 64bit, so the following headers need to be installed.

sudo apt install linux-headers-amd64

32-bit Debian systems, use the following command instead.

DO NOT USE THE BELOW COMMANDS IF YOU ARE ON A 64-BIT SYSTEM.

32-bit non-PAE kernel:

sudo apt install linux-headers-686

32-bit PAE kernel:

sudo apt install linux-headers-686-pae

Install Nvidia Drivers

Next, you can install the “nvidia-detect,” which will automatically suggest the best package for your graphics card.

To do this, use the following command:

sudo apt install nvidia-detect

Now that Nvidia-detect is installed run the command to check your current Nvidia Graphics Card:

The output has shown the machine in the example has a GeForce GTX 1650 card and that it is recommended to install the nvidia-driver package , but this is just a recommendation.

Next, install the recommended package:

sudo apt install nvidia-driver linux-image-amd64

You will see a note about the free Nouveau graphics driver conflicting with the installed new driver. As per the message, you need to reboot to correct this; for now, press the “ENTER KEY” to proceed.

Once installed, reboot your system.

Once logged back in, run the “nvidia-smi” command to confirm the new Nvidia Drivers have been installed.

Install Nvidia Legacy Drivers

If your Nvidia Graphics card is quite old from 400 Series downwards, you will need to install the legacy drivers. The process is the same, just with a new install command:

sudo apt install nvidia-legacy-390xx-driver firmware-misc-nonfree -y

Once complete, do not forget to reboot your system.

Next, run the “nvidia-smi” command to verify the installation.

The 340 series can technically be installed; however, it is no longer supported, and several major security flaws exist in those drivers. It is not advised to install these drivers if you need to replace 390xx with 340xx, but Debian does not recommend this.

#2nd Method: Install Nvidia Drivers – Nvidia Repository

The second method will install the latest drivers for your desktop or server using the Cuda repository. I would recommend using these drivers since when an update is released, you will receive it straight away. This is good for security and performance since Nvidia often does new releases, and now you can install the open-source option.

Determine your Graphics Card

The first step for users with aging NVIDIA Graphics cards is to determine what it is and if it is supported, users with brand new cards can skip this part as there is no doubt they will be supported.

First, find your graphics card module.

03:00.0 VGA compatible controller: NVIDIA Corporation TU117 [GeForce GTX 1650] (rev a1)

Importing Nvidia Repository

First, ensure you install the following packages, which may be already installed.

sudo apt install dirmngr ca-certificates software-properties-common apt-transport-https dkms curl -y

Next, import the GPG key, and ensure you match the import to your distribution version.

Import GPG key for Debian 11 Bullseye:

curl -fSsL https://developer.download.nvidia.com/compute/cuda/repos/debian11/x86_64/3bf863cc.pub | sudo gpg --dearmor | sudo tee /usr/share/keyrings/nvidia-drivers.gpg > /dev/null 2>&1

Import GPG key for Debian 10 Buster:

curl -fSsL https://developer.download.nvidia.com/compute/cuda/repos/debian10/x86_64/3bf863cc.pub | sudo gpg --dearmor | sudo tee /usr/share/keyrings/nvidia-drivers.gpg > /dev/null 2>&1

Second, import the repository for your Debian system, and again ensure you import the correct one to match your Debian distribution version.

Import Nvidia Repository for Debian 11 Bullseye:

echo 'deb [signed-by=/usr/share/keyrings/nvidia-drivers.gpg] https://developer.download.nvidia.com/compute/cuda/repos/debian11/x86_64/ /' | sudo tee /etc/apt/sources.list.d/nvidia-drivers.list

Import Nvidia Repository for Debian 10 Buster:

echo 'deb [signed-by=/usr/share/keyrings/nvidia-drivers.gpg] https://developer.download.nvidia.com/compute/cuda/repos/debian10/x86_64/ /' | sudo tee /etc/apt/sources.list.d/nvidia-drivers.list

Enable the CONTRIB repository. Run this even if you already have it enabled to be safe.

sudo add-apt-repository contrib

Install Nvidia Drivers – Proprietary or Opensource options

Update your sources list to reflect the newly added repository.

Next, install the latest NVIDIA drivers.

Install NVIDIA Drivers Without Cuda Support (Proprietary):

sudo apt install nvidia-driver nvidia-smi nvidia-settings

Install NVIDIA Drivers With Cuda Support (Proprietary):

sudo apt install nvidia-driver cuda nvidia-smi nvidia-settings

Install NVIDIA Drivers Without Cuda Support (Opensource):

sudo apt install nvidia-driver nvidia-kernel-open-dkms nvidia-smi nvidia-settings

Install NVIDIA Drivers With Cuda Support (Opensource):

sudo apt install nvidia-driver nvidia-kernel-open-dkms cuda nvidia-smi nvidia-settings

Note that the download is large and can take a while; given this is a general release, it is recommended to install all the drivers to avoid any issues or inconsistencies on your Debian desktop.

One good thing about the Nvidia repository installation is that if you forgot to remove the previous one, you would see a prompt for users with pre-existing installations before the installation continues. Select Yes to remove any existing drivers, failure to do this may cause system instability. Once done, the installation will be complete.

Once the installation is complete, you will be prompted to reboot your PC to enable the NVIDIA drivers.

Reboot your system once done.

Verify the installation by running NVIDIA-SMI as the manual installation steps showed beforehand.

Alternatively, you can open the GUI for desktop users using the following command.

Or open the application icon located in the following path.

Activities > Show Applications > Nvidia X Server

Install 32-bit Support for Nvidia Drivers

32-bit support can be easily enabled and installed first. Install the 64-bit drivers above, then proceed with the following steps.

First, enable 32-bit architecture.

sudo dpkg --add-architecture i386

Update the APT-CACHE to reflect the changes to the architecture.

Install 32-bit support, and remove the Cuda package “libcuda1-i386” for those users not requiring it.

sudo apt install libcuda1-i386 nvidia-driver-libs-i386

Conclusion

The tutorial has demonstrated two methods of installing the Nvidia Drivers on your Debian system. Please take care doing this; like with custom kernel, custom driver installs can be troublesome; anyone on Linux should utilize the vast amount of free backup utilities in case things go pair shaped.

Источник