Issues with nvidia-powerd.service and Optimus PRIME offloading on NixOS

[miskatul-anwar]

I’m encountering an issue where the nvidia-powerd.service fails to start, and I am unsure whether my Optimus PRIME offloading configuration is working properly on NixOS. I have an Intel and Nvidia GPU setup (Intel UHD 630 and Nvidia GTX 1050 Ti), and I want to enable GPU offloading for performance tasks while using the Intel GPU for less demanding tasks.

System Information:

• Nvidia GPU: GTX 1050 Ti Mobile (PCI:1:0:0)
• Intel GPU: UHD Graphics 630 (PCI:0:2:0)
• Kernel Version: 6.10.9-zen1
• NixOS Version: 24.11.20240926.1925c60 (Vicuna)

Steps Taken:

1. I configured my nvidia.nix file to enable Nvidia offloading through Optimus PRIME, using the following setup:

   services.xserver.videoDrivers = ["nvidia"];
   
   hardware.nvidia-container-toolkit.enable = true;
   
   hardware.nvidia = {
     modesetting.enable = true;
     powerManagement.enable = true;
     powerManagement.finegrained = true;
     dynamicBoost.enable = lib.mkForce true;
     nvidiaSettings = true;
     package = config.boot.kernelPackages.nvidiaPackages.production;
   
     prime = {
       offload = {
         enable = true;
         enableOffloadCmd = true;
       };
       nvidiaBusId = "PCI:1:0:0";
       intelBusId = "PCI:0:2:0";
     };
   };
   

2. I successfully rebuilt my system with sudo nixos-rebuild switch.

3. After rebooting, I noticed that the nvidia-powerd.service fails to start with the following error in the logs:

   nvidia-powerd.service - nvidia-powerd service
      Loaded: loaded (/etc/systemd/system/nvidia-powerd.service; enabled; preset: enabled)
      Active: failed (Result: exit-code) since Mon 2024-12-02 18:57:41 EET; 268ms ago
      Invocation: 3407735c2e43471e9a1c29c4209082a6
     Process: 149160 ExecStart=/nix/store/xdxyd05p3brlj7s1nsij58kivf7q9lc1-nvidia-x11-550.120-6.10.9-bin/bin/nvidia-powerd (code=exited, status=1/FAILURE)
   

   The logs show:

   Dec 02 18:57:41 nixos /nix/store/xdxyd05p3brlj7s1nsij58kivf7q9lc1-nvidia-x11-550.120-6.10.9-bin/bin/nvidia-powerd[149160]: No matching GPU found
   Dec 02 18:57:41 nixos /nix/store/xdxyd05p3brlj7s1nsij58kivf7q9lc1-nvidia-x11-550.120-6.10.9-bin/bin/nvidia-powerd[149160]: Failed to initialize Dynamic Boost
   Dec 02 18:57:41 nixos systemd[1]: nvidia-powerd.service: Main process exited, code=exited, status=1/FAILURE
   

4. I have verified the bus IDs for both the Intel and Nvidia GPUs using lspci:

   00:02.0 VGA compatible controller: Intel Corporation CoffeeLake-H GT2 [UHD Graphics 630]
   01:00.0 VGA compatible controller: NVIDIA Corporation GP107M [GeForce GTX 1050 Ti Mobile]
   

   The correct bus IDs are:

   • Intel GPU: PCI:0:2:0
   • Nvidia GPU: PCI:1:0:0

Expected Behavior:

• The nvidia-powerd.service should start successfully and manage GPU power states without errors.
• Optimus PRIME offloading should allow me to offload rendering tasks to the Nvidia GPU as needed, while the Intel GPU is used for less demanding tasks.

Actual Behavior:

• The nvidia-powerd.service fails to start with an error: No matching GPU found.
• The Optimus PRIME offloading might not be functioning as expected, as I am unsure if the Nvidia GPU is being used for rendering tasks.

What I Have Tried:

• Disabled nvidia-powerd by setting systemd.services.nvidia-powerd.enable = false; in the configuration, but the issue persists.
• Verified the correctness of the Bus IDs using lspci, and the configuration matches the actual system.
• Attempted to test Optimus offloading using prime-run, but I am unsure if it is working as expected.

Logs:

Here are the relevant logs from journalctl:

Dec 02 18:57:41 nixos /nix/store/xdxyd05p3brlj7s1nsij58kivf7q9lc1-nvidia-x11-550.120-6.10.9-bin/bin/nvidia-powerd[149160]: No matching GPU found
Dec 02 18:57:41 nixos systemd[1]: nvidia-powerd.service: Main process exited, code=exited, status=1/FAILURE

Questions:

1. Has anyone encountered a similar issue with nvidia-powerd.service failing to start?
2. Does my Optimus PRIME offloading configuration look correct, and how can I test it effectively?
3. Are there any additional steps I should take to resolve the issue, especially with Nvidia power management?

Any help or insights would be greatly appreciated!

---

[eljamm]

According to the Nvidia docs - Chapter 23. Dynamic Boost on Linux, among the hardware requirements for dynamic boost, the GPU architecture must be Ampere or newer, but your GPU uses Pascal, which is older, so I guess your hardware simply doesn’t support this.

@miskatul-anwar

Disabled nvidia-powerd by setting systemd.services.nvidia-powerd.enable = false; in the configuration, but the issue persists.

To disable nvidia-powered, you need to remove hardware.nvidia.dynamicBoost.enable.

@miskatul-anwar

Attempted to test Optimus offloading using prime-run, but I am unsure if it is working as expected.

I don’t think there is a prime-run command. Instead, you should use nvidia-offload, which should be installed in your system since you’ve enabled hardware.nvidia.prime.offload.enableOffloadCmd (See NVIDIA - NixOS Wiki).

Aside from this, your config looks fine to me.

[marmar]

The command is nvidia-offload as already mentioned

You can also try this (seems to be a Xorg syntax…)

    nvidiaBusId = "1@0:0:0";
    intelBusId = "0@2:0:0";

Struggling with nvidia prime

For your bus ids try: intelBusId = "0@0:2:0"; nvidiaBusId = "1@0:0:0"; I have no idea where I got that syntax from, but it works on my AMD/NVIDIA laptop. nvidia-smi should show Xorg as a process.

[miskatul-anwar]

Thank you so much, @marmar and @eljamm!

After disabling DynamicBoost, I used the following configuration:

• intelBusId = "0@0:2:0"
• nvidiaBusId = "1@0:0:0"

This adjustment worked flawlessly, and the setup is now running perfectly without any errors. I truly appreciate the assistance!