AMD GPU: Difference between revisions

Revision as of 01:11, 19 March 2025

This guide is about setting up NixOS to correctly use your AMD Graphics card if it is relatively new (aka, after the GCN architecture).

Basic Setup

For ordinary desktop / gaming usage, AMD GPUs are expected to work out of the box. As with any desktop configuration though, graphics acceleration does need to be enabled.

# as of 24.11
hardware.graphics = {
  enable = true;
  enable32Bit = true;
};

# pre 24.11
hardware.opengl = {
  enable = true;
  driSupport = true;
  driSupport32Bit = true;
};

Problems

Dual Monitors

If you encounter problems having multiple monitors connected to your GPU, adding `video` parameters for each connector to the kernel command line sometimes helps.

For example:

boot.kernelParams = [
  "video=DP-1:2560x1440@144"
  "video=DP-2:2560x1440@144"
];

With the connector names (like `DP-1`), the resolution and frame rate adjusted accordingly.

To figure out the connector names, execute the following command while your monitors are connected:

head /sys/class/drm/*/status

System Hang with Vega Graphics (and select GPUs)

Currently on the latest kernel/mesa (currently 6.13 and 24.3.4 respectively), Vega integrated graphics (and other GPUs like the RX 6600^[1]) will have a possibility to hang due to context-switching between Graphics and Compute.^[2] There are currently two sets of patches to choose between stability or speed that can be applied: amdgpu-stable and amdgpu-testing.

See Linux Kernel#Patching a single In-tree kernel module, keep in mind how to make patch diffs from commits from GitHub, and consider this example configuration:

{ config, pkgs, ... }:
let
  amdgpu-kernel-module = pkgs.callPackage ./packages/amdgpu-kernel-module.nix {
    # Make sure the module targets the same kernel as your system is using.
    kernel = config.boot.kernelPackages.kernel;
  };
  # linuxPackages_latest 6.13 (or linuxPackages_zen 6.13)
  amdgpu-stability-patch = pkgs.fetchpatch {
    name = "amdgpu-stability-patch";
    url = "https://github.com/torvalds/linux/compare/ffd294d346d185b70e28b1a28abe367bbfe53c04...SeryogaBrigada:linux:4c55a12d64d769f925ef049dd6a92166f7841453.diff";
    hash = "sha256-q/gWUPmKHFBHp7V15BW4ixfUn1kaeJhgDs0okeOGG9c=";
  };
  /*
  # linuxPackages_zen 6.12
  amdgpu-stability-patch = pkgs.fetchpatch {
    name = "amdgpu-stability-patch-zen";
    url = "https://github.com/zen-kernel/zen-kernel/compare/fd00d197bb0a82b25e28d26d4937f917969012aa...WhiteHusky:zen-kernel:f4c32ca166ad55d7e2bbf9adf121113500f3b42b.diff";
    hash = "sha256-bMT5OqBCyILwspWJyZk0j0c8gbxtcsEI53cQMbhbkL8=";
  };
  */
in
{
  # amdgpu instability with context switching between compute and graphics
  # https://bbs.archlinux.org/viewtopic.php?id=301798
  # side-effects: plymouth fails to show at boot, but does not interfere with booting
  boot.extraModulePackages = [
    (amdgpu-kernel-module.overrideAttrs (_: {
      patches = [
        amdgpu-stability-patch
      ];
    }))
  ];
}

Sporadic Crashes

If getting error messages in dmesg with page fault or GCVM_L2_PROTECTION_FAULT_STATUS it might be from AMD GPU boosting too high without enough voltage

Use a tool like LACT to increase power usage limit to 15%, undervolt by moderate amount (e.g. -50mV for 7900 XTX) and optionally decrease maximum GPU clock.

Special Configuration

The following configurations are only required if you have a specific reason for needing them. They are not expected to be necessary for a typical desktop / gaming setup.

Enable Southern Islands (SI) and Sea Islands (CIK) support

The oldest architectures that AMDGPU supports are Southern Islands (SI, i.e. GCN 1) and Sea Islands (CIK, i.e. GCN 2), but support for them is disabled by default. To use AMDGPU instead of the radeon driver, you can set the kernel parameters:

# For Southern Islands (SI i.e. GCN 1) cards
boot.kernelParams = [ "radeon.si_support=0" "amdgpu.si_support=1" ];
# For Sea Islands (CIK i.e. GCN 2) cards
boot.kernelParams = [ "radeon.cik_support=0" "amdgpu.cik_support=1" ];

Doing this is required to use Vulkan on these cards, as the radeon driver doesn't support it.

HIP

Most software has the HIP libraries hard-coded. You can work around it on NixOS by using:

  systemd.tmpfiles.rules = 
  let
    rocmEnv = pkgs.symlinkJoin {
      name = "rocm-combined";
      paths = with pkgs.rocmPackages; [
        rocblas
        hipblas
        clr
      ];
    };
  in [
    "L+    /opt/rocm   -    -    -     -    ${rocmEnv}"
  ];

Blender

Hardware accelerated rendering can be achieved by using the package blender-hip.

Currently, you need to use the latest kernel for blender-hip to work.

OpenCL

hardware.graphics.extraPackages = with pkgs; [ rocmPackages.clr.icd ];

You should also install the clinfo package to verify that OpenCL is correctly setup (or check in the program you use to see if it is now available, such as in Darktable).

Radeon 500 series (aka Polaris)

As of ROCm 4.5, AMD has disabled OpenCL on Polaris-based cards. This can be re-enabled by setting the environment variable ROC_ENABLE_PRE_VEGA=1

environment.variables = {
  ROC_ENABLE_PRE_VEGA = "1";
};

Older GPUs (TeraScale)

For graphics cards older than GCN 1 — or for any GCN using the "radeon" driver — enable OpenCL by adding Clover instead of the ROCm ICD:

hardware.opengl.extraPackages = with pkgs; [
  # OpenCL support for the older Radeon R300, R400, R500,
  # R600, R700, Evergreen, Northern Islands,
  # Southern Islands (radeon), and Sea Islands (radeon)
  # GPU families
  mesa.opencl
  # NOTE: at some point GPUs in the R600 family and newer
  # may need to replace this with the "rusticl" ICD;
  # and GPUs in the R500-family and older may need to
  # pin the package version or backport Clover
  # - https://www.phoronix.com/news/Mesa-Delete-Clover-Discussion
  # - https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19385
];

Merely installing mesa.opencl with nix-shell -p will not work; it needs to be present at build-time for the OpenCL ICD loader, which only searches static paths.

Vulkan

Vulkan is already enabled by default (using Mesa RADV) on 64 bit applications. The settings to control it are:

hardware.opengl.driSupport = true; # This is already enabled by default
hardware.opengl.driSupport32Bit = true; # For 32 bit applications

AMDVLK

The AMDVLK drivers can be used in addition to the Mesa RADV drivers. The program will choose which one to use:

#24.11 
hardware.graphics.extraPackages = with pkgs; [
  amdvlk
];
# For 32 bit applications 
hardware.graphics.extraPackages32 = with pkgs; [
  driversi686Linux.amdvlk
];

#24.05 and below
hardware.opengl.extraPackages = with pkgs; [
  amdvlk
];
# For 32 bit applications 
hardware.opengl.extraPackages32 = with pkgs; [
  driversi686Linux.amdvlk
];

More information can be found here: https://nixos.org/manual/nixos/unstable/index.html#sec-gpu-accel-vulkan

GUI tools

LACT - Linux AMDGPU Controller

This application allows you to overclock, undervolt, set fans curves of AMD GPUs on a Linux system.

In order to install the daemon service you need to add the package to systemd.packages. Also the wantedBy field should be set to multi-user.target to start the service during boot.

environment.systemPackages = with pkgs; [ lact ];
systemd.packages = with pkgs; [ lact ];
systemd.services.lactd.wantedBy = ["multi-user.target"];

Links

References

[1] ttps://bbs.archlinux.org/viewtopic.php?pid=2224147#p2224147

[2] ttps://bbs.archlinux.org/viewtopic.php?id=301798

[1]

[2]