[SOLVED] Unusable system after Nvidia 615 updates

ttofano · 2026-10-24 20:51:00

Hello all!

I am running into several issues with the latest Nvidia driver update on my system rendering it virtually unusable. I have a MSI Nvidia RTX 3060 (not the ti version, the one with capped bitcoin hash rates) and since the 6xx nvidia-open drivers release, nothing works as it would before.

First, I had black screens during boot. System was starting, but the screen would hang at the "Loading console fonts" (or something of the sorts) and never get to a login prompt. Reinstalling the system only worked until installing the Nvidia drivers again, which would bring me back to the same problem.

After a lot of recovery attempts with no success, I got it fixed by: 1) migrating to linux-zen latest kernel before installing the nvidia drivers; and 2) adding

"module_blacklist=simpledrm initcall_blacklist=simpledrm_platform_driver_init"

to my kernel parameters (using GRUB here) (source: this thread. Now I can boot back into the system and 'startx' into an i3 session.

However, now I have a horrid performance even on desktop. Just moving the mouse around causes severe stutters and freezes. Just watching a YT video became impossible: the video pauses constantly and eventually it crashes completely, forcing me to reload (shift+F5, as a simple F5 refresh doesn't even work) the page to make it run again for another minute or so. After my third full OS reinstall in less than a week, I have not even installed Steam to test games yet. To add insult to injury, DP/HDMI audio is just awful, with a constant low-volume buzzing/crackling in background. Works ok on my headphones.

On my attempts, I tried installing M$ Windows 11 to test if it was a hardware issue, and it doesn't seem so, as it all works fine there. Also tried live-isos and they all seem ok (tried Pop!_OS and CachyOS, had some lags with Cachy, but given I was running from a usb stick, I guess its fine?). I tried installing Cachy and had the same experience with the black screens on boot, so I just reinstalled Arch - back to zero.

I have tried to use downgrade (from AUR) to install previous driver versions: 610.xx had the same issues - black screen on boot, terrible performance when booted with the kernel options workaround; 590.xx wouldn't even load nvidia modules - nvidia MODULES (namely, "nvidia nvidia_modeset nvidia_uvm nvidia_drm") throws "module not found" errors with "mkinitcpio -P".

Reading the Nvidia's official release post for the latest driver I see loads of issues, but nothing as severe as mine, so I am lost. Any suggestions? Every help is much appreciated!

Some (possibly) useful info about my system:

Any other info needed, please let me know.
[ps.: sorry for the long post, its been a hell of a month since all of this started...]
Thanks in advance, guys!

Last edited by ttofano (Today 10:17:53)

seth · 2026-10-25 07:13:53

out 24 16:25:14 horizons kernel: NVRM: nvAssertFailedNoLog: Assertion failed: GPPut < WATCHDOG_GPFIFO_ENTRIES @ kernel_rc_watchdog.c:1608
out 24 16:25:21 horizons kernel: NVRM: krcWatchdog_IMPL: RC watchdog: GPU is probably locked!  Notify Timeout Seconds: 7
out 24 16:25:21 horizons kernel: NVRM: GPU at PCI:0000:01:00: GPU-4ec3f08f-ae40-5200-4508-025f793f7178
out 24 16:25:21 horizons kernel: NVRM: Xid (PCI:0000:01:00): 8, pid=655, name=Xorg, channel 0x00000002
out 24 16:25:21 horizons kernel: NVRM: krcCheckBusError_KERNEL: PCI-E corelogic: FATAL_ERROR_DETECTED

Which driver are you using on the other distros?
nouveau? Some older Nvidia one (tried the 580xx drivers)?

Edit please also post xorg log and your xinitrc.

Last edited by seth (2026-10-25 07:16:26)

ttofano · 2026-10-25 15:56:12

Hey seth! Thank you for your time!

Since I posted this, I reinstalled Windows to finish some work, and sadly enough, the issues affect this OS as well (just takes a little longer, for some reason). I've also found the same type of "PCI-E fatal error" on Windows logs, so... Yea.... "in this economy"

Anyways, I am keeping my PC shut down for the moment, to avoid further damage, while I arrange with a friend (since I do not own a second PC) to perform swap tests. Now I just have to narrow down if this is a mobo or GPU (PSU even?) issue, I'm afraid.

Again, I greatly appreciate your efforts around here! I'll mark the post as "solved" and come back if I need further assistance.
Thank you!

seth · 2026-10-26 21:36:32

out 24 16:16:22 horizons kernel: nvidia-modeset: ERROR: GPU:0: nvHandleCpEventDeferredWork: displayId 0x8000 invalid

https://github.com/NVIDIA/open-gpu-kern … ssues/1167

See the 3rd link below. Mandatory.
Disable it (it's NOT the BIOS setting!) and reboot windows and linux twice for voodo reasons.

Can you test https://aur.archlinux.org/packages/nvidia-580xx-dkms (since the OP suggested a regression) and maybe a different output/monitor?

ttofano · 2026-10-27 20:10:51

Hey seth. Oh boy... Thank you for keeping up with this.

To keep you up-to-date: I did perform the swap tests. Surely enough, my GC had some hiccups on a different rig with Win11 and drivers-only. However, there were no PCI-e errors, so I decided to do a full nvidia-suite installation and, after installing the nvidia-app (that fancy control panel for windows), the hiccups were gone! So I thought "ok, its my mobo then" and installed an AMD card on my rig as a test (expecting to find the same PCI-e errors), but nothing, the Radeon card works flawlessly. I was coming back here to re-open the thread when I saw your message, and decided to do as you said first.

So, despite the fact that I didn't have Windows installed on my rig when it all started (I installed it on a second partition as a test), I did the mambo-jambo tricks mentioned on your sig. It didn't really do much after going back to Linux, latest 615-open-dkms drivers still reproduce the same (black screen on boot, low perf if booted).

As for the 580xx-dkms, this might be tricky. I did try to install it before and, as I mentioned, I get a black screen on reboot, but with this version changing the boot parameters pose no difference (mind you, with the latest 615-open-dkms I have to add the parameters mentioned on my 1st post so I can get to the login prompt). I do not have a second display to test a different output, and my TV has had issues with PC resolutions in the past, so I don't trust it for this task. I'll try to get a display from my friend as well, but might take a little while - because week days...

In the meantime, any other suggestions that I may try while I'm at it? Especially if it involves HW swapping lol. This way I can arrange it all at once with my friend XD
Don't mind if I take a little to reply, I am reading this from my phone and doing as you say, so I'll come back when I have any news.

Thanks again mate! Talk again soon.

seth · 2026-10-27 21:41:47

Can you get a log from the run w/ the 580xx drivers (either via ssh or by rebooting w/ https://wiki.archlinux.org/title/Keyboa … el_(SysRq) + REISUB or frenetically pressing ctrl+alt+del and "sudo journalctl -b -1" for the journal of the previous boot?

ttofano · Yesterday 02:50:35

Hey again, @seth. So, this was weirder than expected, but I managed to make it work with 580xx-dkms.

So, when I got the second display to test a different output, I had my card plugged to my friend's PC. So I swapped it back to my PC and now the card wasn't working at all - I couldn't even see the BIOS screen. Since it was working on my friend's PC, i swapped back and voilà, working. So I had to start a little dance. Bear with me, please lol.

First, I realized that the Windows PC (my friend's) had hibernated or went into sleep mode and, given it was "off", I took off the card and plugged it back to my system. Plugging it back to his PC "woke up" the card. I then disabled the hibernation on his system as per the link on your signature, rebooted Windows twice, shut off the system properly and only then moved the card to my PC. Then I was back to the initial issues. Now, given the Github discussion you linked on your 2nd reply, I started believing that the sleep/hibernation thing was the root of my evils, so instead of trying to downgrade the drivers, I decided to wipe the system once more and ONLY install the 580xx-dkms drivers. Whaddyaknow, it is all working again, including sleep/hibernation (which pisses me off because I hate this and disable them every time!). I can safely put out the candles and stop the drums; the voodoo worked

However, my dumb arse didn't remember you wanted a journalctl with a downgraded 580xx-dkms, not a fresh one. I can still do it if we want to find the reason for all this. What I gathered so far is that if I install the 615xx drivers, it will trigger the problems and just downgrading it won't help (no matter the version I downgrade it to). Whatever changes on the latest verion remains changed even after the downgrade, but I have no idea what it could be. Another test that crossed my mind was to trigger the problem and try to "wake up" the card from my friend's Windows PC - I don't know, maybe the difference on the drivers could bring the card back as it did before?

Anyways, I have the journalctls from my fresh install, both "driverless" (just barebones system with nothing installed) and after installing the 580xx-dkms drivers, if they help with anything.

I'll wait for your reply before marking this as solved. As I said, I'm willing to trigger the issue again so we can try to diagnose it properly, but maybe this would be more suitable to NVidia's github, no? Please let me know.
And thank you so so much for all the help!

seth · Yesterday 14:24:36

What I gathered so far is that if I install the 615xx drivers, it will trigger the problems and just downgrading it won't help

Have you actually attempted 615xx after all the voodoo?
If the GPU travels into and out of hibernating systems there could be some hiccup in the nvrm, since you're not(?) even disabling https://wiki.archlinux.org/title/NVIDIA … P_firmware as of now that might be the problem rather than any particular driver.

Otherwise and since your GPU is covered by the 580xx drivers and they're still maintained for now there's currently no downside in using them (except the dkms part) until the nvidia-open drivers improve coverage of ampere chips

ttofano · Yesterday 18:15:12

I decided to give it a shot so I disabled the GSPf with the 580xx (you are correct, I did not disable it at any moment after the first system wipe; only time I tried it was when this all started and it might have been after a load of failed attempts), but I could find no difference. Here are the journals:

-- journalctl-nvidia580xx_gsp_Enabled
-- journalctl-nvidia580xx_gsp_Disabled

As for trying 615xx, I'll wait for my night time as I have some work to do, but one little question: is it at all possible to disable the GSP on the Nvidia-open drivers? I've read somewhere that this parameter is ignored on the open drivers, but cannot remember nor find where (sorry, I've read a lot lately). If possible, can you point me the directions, please? If not, disabling the GSP firmware from a previous version would pose a difference before the update to 615xx? (I mean, after all the voodoo, "I want to believe" it will even if logic says otherwise lmao)

Thanks mate!

seth · Yesterday 19:46:02

The nvidia-open drivers critically rely on the GSP and will just ignore the parameter, that's correct.
The point however was that since you didn't even have to disable the GSP there might be a chance that the previous situation had only been caused by some inconsistency in the nvrm (similar to the windows situation) and that clearing that up might allow you to use the nvidia-open drivers (incl 615xx) again.

ttofano · Yesterday 22:42:47

Alrighty, thanks for the explanation! Sooo, weirdly enough... it worked? o.O

I tried the current nvidia-open driver and still had the boot issues and horrid performance (here is the journalctl -b for this specific boot). Then I just installed the nvidia-open-dkms: I did not remove anything, did not change any files, just "sudo pacman -S nvidia-open-dkms", fixed the conflicts, rebooted, and it is working like a charm???? O_O (here the current journalctl)... The only "issue" I had was the pipewire HDMI audio sink that vanished, but reinstalling all the wireplumber+pipewire packages fixed it all.

I'm sorry if I sound (read) weird, but I am weirded out. Should I worry about my GC's health? Could whatever happened be caused by a HW fault/malfunction? I see no indication of such, but at this point, I couldn't be any more surprised, honestly.

Really, you have no idea how much I appreciate your help here, @seth! Thank you so much. If you have any suggestions (other than "buy an AMD card and ditch those bloodsuckers from Nvidia already" lol), or if I can help with any other tests, please let me know. Wish you the best, mate!

ttofano · Yesterday 23:32:35

PS.: please disregard my awe. Took me a while to fully understand your explanation there, and once I fully understood, it was a little "eureka" moment lol. My guess now is that either I was mid-render for some complex 3D scene when I applied the update that opened Hell's gates, or I might have updated the system once and never rebooted before doing another update (damn me, I'm getting old...). I'm used to take long render times to update the system without issues, so I didn't connect these dots. Either way, this time I apparently screwed the nvrm and the rest is here... Both ways, lesson learned, never again hahaha.
--
(Edit: removed double sentence)

Last edited by ttofano (Yesterday 23:36:19)

seth · Today 09:00:23

I tried the current nvidia-open driver and still had the boot issues and horrid performance (here is the journalctl -b for this specific boot). Then I just installed the nvidia-open-dkms

out 30 18:33:47 horizons kernel: Linux version 7.2.11-zen1-1-zen (linux-zen@archlinux) (gcc (GCC) 16.2.1 20260810, GNU ld (GNU Binutils) 2.47) #1 ZEN SMP PREEMPT_DYNAMIC Fri, 16 Oct 2026 00:28:42 +0000

Because you're using the zen kernel (meaning you need to use dkms) - nvidia-open is for the main kernel only and you can have that in parallel to the dkms package (at the end of the day you need to get a module for whatever used kernel somehow™)

Please always remember to mark resolved threads by editing your initial posts subject - so others will know that there's no task left, but maybe a solution to find.
Thanks.

ttofano · Today 10:17:34

seth wrote:

Because you're using the zen kernel (meaning you need to use dkms) - nvidia-open is for the main kernel only

True fact. This time it went overhead. Anyways, marked as solved. Thanks again!

Arch Linux

#1 2026-10-24 20:51:00

[SOLVED] Unusable system after Nvidia 615 updates

#2 2026-10-25 07:13:53

Re: [SOLVED] Unusable system after Nvidia 615 updates

#3 2026-10-25 15:56:12

Re: [SOLVED] Unusable system after Nvidia 615 updates

#4 2026-10-26 21:36:32

Re: [SOLVED] Unusable system after Nvidia 615 updates

#5 2026-10-27 20:10:51

Re: [SOLVED] Unusable system after Nvidia 615 updates

#6 2026-10-27 21:41:47

Re: [SOLVED] Unusable system after Nvidia 615 updates

#7 Yesterday 02:50:35

Re: [SOLVED] Unusable system after Nvidia 615 updates

#8 Yesterday 14:24:36

Re: [SOLVED] Unusable system after Nvidia 615 updates

#9 Yesterday 18:15:12

Re: [SOLVED] Unusable system after Nvidia 615 updates

#10 Yesterday 19:46:02

Re: [SOLVED] Unusable system after Nvidia 615 updates

#11 Yesterday 22:42:47

Re: [SOLVED] Unusable system after Nvidia 615 updates

#12 Yesterday 23:32:35

Re: [SOLVED] Unusable system after Nvidia 615 updates

#13 Today 09:00:23

Re: [SOLVED] Unusable system after Nvidia 615 updates

#14 Today 10:17:34

Re: [SOLVED] Unusable system after Nvidia 615 updates

Board footer