NVIDIA Linux Driver 580.105.08: What’s New, Bug Fixes, and Installation Guide

If you are running Linux with an NVIDIA GPU, keeping your driver up to date can help a lot, especially if you play games, run CUDA workloads, or use high resolution and high refresh rate monitors. NVIDIA has released the Linux x64 (AMD64/EM64T) driver version 580.105.08, which focuses on fixing several real issues users have reported and adds a simple but useful control for CUDA power behavior.

In this guide, the Linux Simplify team will go through what changes in this release, what bugs it fixes, how it behaves on different types of hardware, and how you can install it on your Linux system using the official installer from NVIDIA.

Overview of NVIDIA Linux Driver 580.105.08

Here is a quick summary of this driver release:

• Driver version: 580.105.08
• Platform: Linux 64 bit (x86_64 / AMD64 / EM64T)
• Release date: November 4, 2025
• Installer type: standalone .run file from NVIDIA
• File size: about 396 MB

NVIDIA recommends that users update to the latest available driver for their GPU, especially if you are hitting any of the issues that this version claims to fix.

New CUDA Control: Disable Automatic Performance Boost

The main new feature in this release is an environment variable that lets you control how the GPU behaves when running CUDA applications:

CUDA_DISABLE_PERF_BOOST 

By default, when you run CUDA workloads, the driver may push the GPU into a higher power state to squeeze out more performance. That can be good for raw speed, but it is not always ideal if you:

• want to save power
• are trying to keep GPU temperatures lower
• need consistent benchmark results without automatic boosting
• run on a system with strict power or cooling limits

If you set the variable to 1 before starting your CUDA application:

export CUDA_DISABLE_PERF_BOOST=1 

the driver will no longer push the GPU into a higher power state just because CUDA is running. This gives you more predictable behavior, especially on systems where thermals and power draw matter.

Bug Fixes in NVIDIA Driver 580.105.08

This release is not just a version bump. NVIDIA lists several specific issues that have been addressed, which is important if you use GPU passthrough, gaming under Linux, or modern displays with HDMI FRL and VRR.

1. VFIO GPU Passthrough Soft Lockup

If you are using vfio-pci for GPU passthrough with KVM or QEMU, you might have seen the host system hang or hit a soft lockup after shutting down a virtual machine that uses an NVIDIA GPU.

Driver 580.105.08 includes a fix for this behavior, so shutting down a VM that uses a passed through NVIDIA GPU should be more stable on the host.

2. HDMI FRL Problems After Replugging the Monitor

Some users reported that HDMI FRL stopped working after unplugging and replugging a display. In practice, this could look like high bandwidth HDMI modes no longer being available after you reconnect the monitor.

This driver release addresses that issue and improves HDMI FRL behavior when you frequently unplug and replug your display.

3. Game Crashes: Rage 2 and Metro Exodus Enhanced Edition

On Linux gaming setups, especially with Proton or similar layers, there have been issues with certain titles:

• Rage 2 crashing when opening or navigating the in game map or menu
• Metro Exodus Enhanced Edition crashing under certain conditions

These crashes are specifically mentioned as fixed in this driver, so if you play either of these games on Linux with an NVIDIA GPU, it is worth trying this version.

4. VRR Being Enabled on Unsupported Modes

Another bug involved Variable Refresh Rate (VRR) being active on modes that do not really support it. When that happens, the result can be:

• a black screen
• a display that looks like it has no signal, even though the system is still running

The 580.105.08 release tightens up how VRR is handled so that it will not be enabled on modes where it should not run, which helps avoid random black screen behavior.

5. HDMI Displays Staying Blank After Replugging

There was also a case where some HDMI displays stayed blank after unplugging and plugging the cable back in, even though they were detected by the system.

This driver improves re initialization of displays in those situations, so the monitor is more likely to light up correctly without needing a reboot or manual fixes.

6. Missing Very High Resolution or High Refresh Rate Modes

If you are using very demanding display modes, such as:

7680x2160 at 240 Hz 

you might have noticed that certain modes simply did not show up when using HDMI FRL or DisplayPort.

Driver 580.105.08 includes a fix that helps ensure these high resolution and high refresh rate modes are available when both your GPU and your monitor support them.

Using Distro Packages vs NVIDIA's Official Installer

Before you decide to install the official .run file from NVIDIA, it is important to know that many Linux distributions already provide their own NVIDIA driver packages in the native package format, such as .deb or .rpm.

Using the distribution package often gives you:

• better integration with your package manager, like apt, dnf, zypper, or pacman
• automatic handling of kernel updates and module rebuilds
• cleaner integration with the rest of the system framework

If your distribution already ships a recent NVIDIA driver version that matches what you need, it can be simpler and safer to stay with the distro provided package rather than the standalone installer.

Note for SUSE and openSUSE Users

If you are using SUSE or openSUSE and plan to install the official NVIDIA driver instead of the distribution package, NVIDIA recommends that you read the SuSE NVIDIA Installer HOWTO first. That document explains how to avoid conflicts with the distribution's own tools and driver packages.

How To Install the Official NVIDIA Driver on Linux

If you prefer to use NVIDIA's official installer, either because your distribution does not provide this exact version yet or you want to track the official releases directly, the installation process is fairly simple once you get used to it.

In many cases, it is recommended to stop your display manager and run the installer from a TTY, but here we will focus on the basic commands you will use.

1. Download the Driver File

First, download the driver from NVIDIA's website https://www.nvidia.com/en-us/drivers/details/257493/ . The file name should look like this:

NVIDIA-Linux-x86_64-580.105.08.run 

Save it to a directory that you can reach easily, for example:

~/Downloads 

2. Open a Terminal and Go to the Driver Directory

In your terminal, move into the directory where you saved the driver file:

cd ~/Downloads 

3. Run the Installer as Root

To start the installation, run:

sudo sh ./NVIDIA-Linux-x86_64-580.105.08.run 

Follow the prompts that appear. The installer will handle:

• building and installing the kernel module
• copying the necessary libraries and binaries
• setting up the driver files under the correct system paths

4. Update Your X Configuration

Near the end of the installation, you will see an option that relates to your X configuration file. At this point, you can:

• let the installer update the X configuration automatically
• edit the X configuration file yourself later
• use the nvidia-xconfig tool

A simple way to make sure the NVIDIA X driver is used is to run:

sudo nvidia-xconfig 

After that, you can reboot your system so that the new driver is loaded and used by your display manager.

GPU Compatibility and Hybrid Graphics

The download page for the driver lists which GPU models are supported by this specific version. That list tells you if your GPU is officially supported, but system design also matters, especially on laptops and all in one machines.

On systems with hybrid graphics or Optimus style setups, the NVIDIA GPU may be connected behind an integrated GPU. If the hardware or firmware does not provide a way to disable the integrated graphics or switch to a dedicated only mode, the standard NVIDIA Linux driver may not be able to control the display directly.

This is common on:

• notebooks with switchable or hybrid graphics
• all in one desktops that combine an integrated GPU with an NVIDIA GPU

Because each manufacturer designs their hardware differently, NVIDIA suggests that you check with the system vendor if you are not sure whether your specific model is compatible with the Linux driver.

Documentation and Support Resources

If you want more detailed information, there are two main places you can look after installing this driver.

1. The README File Included with the Driver

The driver package includes a README that covers:

• supported features and options
• configuration examples
• known issues and limitations
• advanced setup scenarios

This file is useful if you are doing anything more advanced than a simple single GPU desktop setup.

2. The NVIDIA Linux Graphics Forum

For real world feedback, troubleshooting tips, and discussions around new releases, you can visit the NVIDIA developer forum for Linux graphics:

NVIDIA Linux Graphics Forum

Here you can find other users' experiences with the 580.xx series, including any new quirks or workarounds that appear in the wild.

Final Notes from Linux Simplify

If you are using CUDA, gaming on Linux, pushing very high resolution or high refresh rate displays, or doing GPU passthrough with virtual machines, driver 580.105.08 is a release that is worth looking at. It does not introduce flashy new features, but it fixes several real world problems and gives you a bit more control over how the GPU behaves under load.

In future guides, the Linux Simplify team can also break this down for specific distributions, for example a focused tutorial for Ubuntu, Fedora, or Arch based systems. If you want a distro specific version of this guide, you can tell us which distribution you are using and we will tailor the steps to match your environment.