Please, any information or help would be appreciated. It feels like I have tried everything and I have already spent three days trying to fix it. I do not know exactly why it is happening because it started suddenly. I upgraded my PC: I bought the GPU and used it for a short time, then bought and installed a new CPU. Everything was fine for a few days and then the issue began. I do not remember making any specific changes; maybe I just ran an update and the problems started.
System Information:
Kernel: 7.0.2-2-cachyos (also tested: 6.18.25-2-cachyos-lts)
CPU: AMD Ryzen 9 5950X
Motherboard: ASUS Rog Strix B550-E Gaming
Driver: nvidia-open-dkms 595.71.05
Bootloader: Limine
Display Server: Tested on both Wayland(Niri/Hyprland) and X11/Xlibre(XFCE)
Problem Description:
GPU randomly locks up during normal use (gaming, desktop, even idle).
Symptoms:
-
Screen goes completely black
-
GPU fans ramp up to 100%
-
System becomes unresponsive — cannot switch TTY
-
Only a hard reset (hold power) recovers the system
After reboot, dmesg shows GSP firmware crash and GPU lost from PCIe bus.
Occurs randomly during gaming, normal desktop use, and while idle.
This happens anywhere from 15 minutes to a few hours after boot. The crash is more frequent under GPU load (gaming) but can also happen at idle.
Root Cause
The crash is caused by GSP (GPU System Processor) firmware crash during P-state transitions. The GPU rapidly switches between P0 (full power) and P3 (low power), and the GSP firmware fails, resulting in NV_ERR_GPU_IS_LOST — the GPU disappears from the PCIe bus entirely.
This is a known bug tracked at:
https://github.com/NVIDIA/open-gpu-kernel-modules/issues/739
https://github.com/CachyOS/distribution/issues/109
Captured Crash Logs:
dmesg excerpt (from monitoring script)
NVRM: Xid (PCI:0000:01:00): 79, pid=XXXX, GPU has fallen off the bus
NVRM: GPU 0000:01:00.0: GPU is lost
nvidia 0000:01:00.0: PCI Bus Error
GPU was stable in P0 state (90-100W power draw) before the crash.
After crash: nvidia-smi fails entirely — “Unable to determine the device handle for GPU”
nvidia-powerd errors:
nvidia-powerd: Failed to get SysPwrLimitGetInfo
nvidia-powerd: DC power limits table is not supported on GPU
What I have already tried (ALL FAILED)
| # | Workaround | Result |
|---|---|---|
| 1 | NVreg_DynamicPowerManagement=0x00 in /etc/modprobe.d/nvidia.conf |  No effect — GSP still crashes |
| 2 | Fixed /usr/lib/modprobe.d/nvidia.conf DynamicPowerManagement 0x02 → 0x00 | No effect |
| 3 | Kernel params: pci=noaer pcie_aspm=off pcie_port_pm=off | No effect |
| 4 | NVreg_RegistryDwords=RMIntrLockingMode=0 (disable CachyOS high-refresh opt) | No effect |
| 5 | Switch Wayland/Hyprland → X11/XFCE | Same crash |
| 6 | Replace xorg-server with XLibre | Same crash |
| 7 | nvidia-smi -pm 1 + PowerMizer locked to P0 | No effect |
| 8 | Downgrade to nvidia-open-dkms 570.124.04 | DKMS build fails on kernels 7.0.2 & 6.18.25 (os-interface.h not found) |
| 9 | Downgrade to nvidia-open-dkms 580.76.05 | DKMS build fails on kernels 7.0.2 & 6.18.25 |
| 10 | Proprietary nvidia-dkms driver | Not available for Blackwell (pacman resolves to nvidia-open-dkms) |
| 11 | NVreg_EnableGpuFirmware=0 (disable GSP) | No effect — GSP mandatory for nvidia-open on Blackwell (NVIDIA: open modules require GSP) |
Why downgrading drivers is not possible (my kernels)
| Driver Version | Kernel support | My kernels | Result |
|---|---|---|---|
| 570.124.04 | ≤ 6.12 | 7.0.2, 6.18.25 | Won’t compile |
| 580.76.05 | ≤ 6.17 | 7.0.2, 6.18.25 | Won’t compile |
| 595.71.05 | ≤ 7.x | 7.0.2, 6.18.25 |  Compiles but GSP crashes |
Questions:
-
Has anyone with an RTX 5060 Ti / Blackwell GPU on CachyOS managed stable operation without GSP crashes?
-
Is there a practical way to run an older kernel (e.g., 6.12.x) on CachyOS to use 570.x drivers?
-
Any CachyOS-specific patches or kernel config suggestions that might affect GSP stability?