Thank you, @vrachatte for this news! Kindly share with us when the fix for this issue is resolved in future driver release?
Testing each release only to have to downgrade and version lock again is very time consuming and frustrating, as I’m sure you can understand. Many thanks again and in advance!! 🙌🏻
Any news on this? As it’s been an entire month already since you were able to reproduce it.
Hi @opisalwaysgreat , sorry for the delay.
A fix for this issue has been identified, it should be present in the next driver release.
Sorry to bother, may I ask when would the next release be.
Can someone else try out the 580.94.17 beta drivers? Not getting the GSP-CrashCat error anymore, though it still doesn’t work.
nvidia-bug-report.log.gz (1.2 MB)
It has now been over 15 weeks since this CRITICAL issue was reported. Since then, NVIDIA has released 3 new driver versions (if I’m not wrong), yet this specific bug remains unresolved.
For those of us on rolling release Linux distributions, this situation is becoming UNSUSTAINABLE. Every time there is a system update, we are forced to manually revert and pin older drivers just to maintain basic functionality. This is TEDIOUS, TIME-CONSUMING, and UNACCEPTABLE for hardware that is supposedly still supported.
Speaking of hardware—the RTX 3080 Mobile is (was) one of the few mobile cards equipped with 16GB of VRAM. This makes it perfectly suited for running smaller local LLMs and for gaming in a Linux environment.
Should we interpret this lack of support and the 15-week delays as a sign that NVIDIA is quietly abandoning support for these cards—even though they are still widely used (and sold)?
When can we realistically expect a full working driver or a fix?
+1, the cat&mouse game with the NVIDIA employees to know when we will be able to actually use our hardware is becoming very annoying. It’s been four months already.
Sorry for the delay, please try with the 580.126.18 driver release.
No luck with 580.126.18 on a 16ach6h (5800h, 3060), tried both a proprietary and a MIT/GPL installation. 580.82.09 remains the last usable version, which will be even more of a bother as distros move to 6.19. It would’ve warranted a mention in the change log for 580.126.18 if an attempt to fix this was made…
So as previously stated by @nthrootofnera , it’s still broken; still the same GSP task exception: load access page fault as soon as I plug in an HDMI cable.
Seriously, your employees were able to reproduce the bug locally. Why not trying it out before shipping?
nvidia-bug-report.log.gz (2.2 MB)
I’m just reading the first post in the thread, and the errors look like normal errors (that can be ignored)(and that were made out to be critically important) and it seems like a normal situation, attempting to connect to an external monitor with an hdmi cable, BUT without actually using the external monitor settings to change the input settings.
Using an external monitor can be a little frustrating, otherwise, because the output of the desktop will immediately switch to the misconfigured monitor, and you’ll be looking at two black screens. There’s a solid handful of little annoying problems that can occur trying to get an external monitor to work.
Again, it isn’t clear from the initial report, that the issue is actually an nvidia problem, or simply user error.
(is it a wayland problem? Nvidia, please get it together with external monitors on Wayland )
now if we all had rtx laptops, and oled tvs lying around, and were using flashy wayland, maybe we could have a shared experience! but unfortunately we all have disparate hardware/software on gnu/linux
why
how
we do
Will we have to wait another trimester before we can try a new driver release, just to notice it’s still broken?
LGTM on 595, cheers!
Using the nvidia-beta-dkms-595.45.04 on Arch, I can confirm the issue is now fixed.
Very good news it was released this day, as Linux 6.19.X released also that same day on Arch.
Sorry for the delay, I think the latest 580.142 should also have the fix.
Using 580.142 fixes the issue. My external monitor works!
As for 595, those drivers have issues for me on Nobara, so I can’t say if the issue is fixed on 595.