RTX 3060 Mobile on HP Omen 15-en1xxx: Xid 79 ⁄ “GPU has fallen off the bus” on CachyOS

Issues & Assistance
1

Hi,

I am trying to investigate a repeatable NVIDIA GPU crash on my laptop and I would appreciate any help.

Hardware / system

Laptop: HP Omen 15-en1xxx

CPU: AMD Ryzen 7 5800H

iGPU: AMD Cezanne / Radeon Graphics

dGPU: NVIDIA GA106M / GeForce RTX 3060 Mobile / Max-Q

BIOS: updated to the latest version

OS: CachyOS

Desktop: KDE Plasma

sudo cachyos-bugreport.sh
Right after the GPU crash https://paste.cachyos.org/p/388210f.log
After reboot https://paste.cachyos.org/p/c776b68.log

Problem
The system was stable before, but after NVIDIA driver updates the discrete GPU started crashing under load.The main symptom is that during gaming the NVIDIA GPU eventually crashes and becomes unavailable. In logs I see errors like:

Xid 79, GPU has fallen off the bus

GPU lost from the bus

rpcSendMessage failed with status 0x0000000f

Sometimes after a clean reinstall the problem temporarily improves and I can play for quite a while, but then it comes back.

A very similar issue also appeared earlier on Windows after a driver update, and at that time it seemed to be resolved by a later NVIDIA driver update. Because of that, I suspect this may be a driver / firmware / power-management issue rather than immediate hardware failure.

On Linux the system was fine until the move from the older kernel (not sure if that was an exact moment when it started). At some point, the issue became frequent and difficult to eliminate.

What I already tried

Clean reinstall of NVIDIA drivers multiple times

Tested both nvidia-open-dkms and closed nvidia-580xx-dkms

Removed custom hp-omen-wmi-dkms

Updated BIOS from F.20 to F.30

Reset BIOS to defaults

2

Isn’t there a way to rollback to a working condition? I see there is a “Limine-snapshot-restore” under Utilities.

3

Thank you for looking into my problem and fast response.

I tried to rollback but had a problem with the kernel loading on an earlier snapshot, so had to load back to the current state

4

I tried to rollback to an older snapshot again and had to restore the system from a liveUSB afterwards. It looks like some update changed partition type of /boot and made some other changes that make cachyOS to load into emergency state.

5

GPU has fallen off the bus error usually indicates hardware problems such as:

  • Power delivery issue where GPU doesn’t reveive enough power due to power supply or other component failure.
  • Thermal issues (Overheating).
  • Unstable overclock.
  • Potential physical damage to the GPU silicon.

This is rarely a software issue and a driver update is likely just a coinsidence.
But you can still try installing older driver such as nvidia-550xx-dkms or nvidia-535xx-dkms and see if that helps.

6

Thank you for the suggestions and for taking the time to look into this.

To be honest, until the very end I was sure that in case of overheating the GPU would just throttle, not completely disconnect from the bus. But it now looks like overheating really may have been the problem.

I cleaned the laptop and removed a large amount of dust from the cooling system, replaced all thermal pads as well, and after that I did get a small improvement in stability. However, while checking the cooling system more closely, I noticed that one of the cooling fans is not working properly at all. It occasionally twitches, but in general it does not spin, even though power is present.

So at this point it seems quite possible that this faulty fan was the real cause, or at least a major contributing factor.

I have already ordered replacement fans and I will write back once I replace them and do more testing.

7

I replaced fans and also repasted the laptop with Arctic MX-6.

The repaste definitely helped. From my temperature logs, before the repaste the CPU was going up to roughly 100–101°C and the NVIDIA GPU was around 86–87°C before the crash. After the repaste the CPU temperatures are lower than 97, and in one of the new crashes the NVIDIA GPU was only around 63–73°C.

Unfortunately, the issue is still there. The GPU still disappears with Xid 79 / “GPU has fallen off the bus”.

So the difficult part is that I also cannot clearly say “the GPU is dead”, because the laptop can work normally under load and even at higher temperatures for some time. The crash does not seem to happen exactly at one specific temperature or load.

At this point I am trying to understand how to debug the real trigger behind the Xid 79 event. Any ideas what else I should log or test would be very welcome.