System stalled. Can't launch programs and Plasma becomes unresponsive.

seth · 2025-07-13 06:58:31

It's unlikely but it absolutely does impact the performance a lot

Post system journal, xorg log and "glxinfo -B" output. It's way more likely that you end up w/ software rendering than that the GSP (which has not been used up until very recently anyway) causes a 4-fold performance difference.
"Unlikely" being a euphemism for "impossible" here…

I forgot to tell you that downgrading to 570xx made the gpu completely disappear from the system and only upgrading back fixed the issue

Changing the driver will not have made it disappear from the bus.
There's either a secondary issue or the dkms module didn't build…

techmanwalker · 2025-07-13 08:32:17

seth wrote:

It's way more likely that you end up w/ software rendering than that the GSP (which has not been used up until very recently anyway) causes a 4-fold performance difference.

About that, I found that when the battery is almost full (85%+) the game does run at the aforementioned 100 FPS, but as soon as the level goes below that, the GPU performance is heavily tanked down to deliver 30 fps despite having the battery in performance mode, slightly more than before but still so little for a cubic shooter game.

Changing the driver will not have made it disappear from the bus.
There's either a secondary issue or the dkms module didn't build…

I did rebuild it afair

glxinfo -B
prime-run glxinfo -B
journal for this boot
xorg log- I use Plasma Wayland btw

seth · 2025-07-13 13:19:18

Sounds like there's some powersaving daemon active that clocks down the system once the battery drops?

You're running on

Device: AMD Radeon Graphics (radeonsi, rembrandt, LLVM 20.1.7, DRM 3.63, 6.15.6-zen1-1-zen) (0x1681)

OpenGL core profile version string: 4.6.0 NVIDIA 575.64.03

is available via prime-run, looks unsuspicious.

jul 12 20:35:30 archlinux kernel: nvidia-modeset: Loading NVIDIA UNIX Open Kernel Mode Setting Driver for x86_64  575.64.03  Release Build  (root@malasdecisiones)

Unlike nvidia, nvidia-open will under no circumstances work w/o the gsp.
nvidia-open, nvidia w/o GSP and nvidia 570xx are three mutually exclusive things to test to narrow down the cause/mitigate the problem

Remove all xf86-video-* packages but maybe xf86-video-amdgpu.

The only things that are remotely suspicious in the latest journal are

jul 13 00:58:47 malasdecisiones kernel: nvidia-modeset: WARNING: GPU:0: Correcting number of heads for current head configuration (0x00)

(but I doubt that there's a problem, you have no output attached ot the nvidia GPU) and

jul 13 00:47:27 malasdecisiones kernel: ACPI: EC: interrupt blocked
jul 13 00:47:27 malasdecisiones kernel: ACPI: EC: interrupt unblocked
jul 13 00:47:27 malasdecisiones kernel: NVRM: RmHandleDNotifierEvent: RmHandleDNotifierEvent: Failed to handle ACPI D-Notifier event, status=0x11

but those strictly occur in relation to s2idle sleeps.

Are you currently encountering the freezes that were the original problem?
If yes, we shall try the 570xx driver again.
Install it, ifff you've trouble booting, only boot the multi-user.target (2nd link below) and in doubt along "nomodeset" and post a journal from the console and the output of

dkms status; pacman -Qs 'nvidia|kernel'

techmanwalker · 2025-07-14 17:07:15

seth wrote:

nvidia 570xx

The open drivers made my GPU disappear again actually, so it did not hang but just vanished from existence again.
You linked to the normal nvidia570 package so that's why it was not working- I needed the nvidia-dkms version to make it work on zen kernel with GSP disabled and so far so good, although my uptime is barely 9 hours so I can't say it's completely hang free and vanish free just yet (steamwebhelper hanging seems normal at this point)

Sounds like there's some powersaving daemon active that clocks down the system once the battery drops?

I just did a test and the performance tanked down right after hitting 50% battery while in-game despite the laptop being doing a reasonable 60 fps on balanced mode- I'm using power-profiles-daemon due to its integration with Plasma. Anyway, do you know any fix or setting for setting this throttle limit to something around 30% or so?

tanked logs
dkms status; pacman -Qs 'nvidia|kernel' # despite having booted successfully

seth · 2025-07-14 20:13:42

"pacman: la opción requiere un argumento -- 'b'"

https://gitlab.freedesktop.org/upower/p … on-changes ?

techmanwalker · 2025-07-14 20:35:36

lol I just copied that over the journal command- actual dkms status; pacman -Qs 'nvidia|kernel' # despite having booted successfully

seth wrote:

?

THANKS [prayingemojibecauseforumdoesnotseemtosupportit]
I'll try it now

Last edited by techmanwalker (2025-07-14 20:36:15)

techmanwalker · 2025-07-14 23:50:41

FINALLY SOMETHING NEW. The computer also hanged in 570xx drivers.

At this point, I'm beginning to get into desperation.

Things that don't work until now:

Disabling GSP firmware- the GPU, instead of disappearing, now hangs the whole system
nvidia-open drivers- they made the GPU disappear again
nvidia570xx drivers- the whole system hangs again

Now I regret not buying a laptop with an AMDGPU card. This bug is taking to absolutely nowhere and absolutely killing the whole system stack. It really sucks

And disabling the battery aware profile also didn't work, the laptop does the same impressive 20 FPS anyway

hanged right after closing a game- hang type: new programs don't launch
logs from another hang that happened right after when I didn't even use the nvidia card for anything, just compiling a program on battery- this is a complete halt-freeze- are some hangs not even nvidia related?

EDIT 4: Wait... I can reproduce that last halt: while performing a CPU intensive task (compiling a program) on low battery (<50%) in performance mode, the system completely freezes and doesn't go on anymore- is this some sort of BIOS bug or power delivery issue?

EDIT 5: I've just disconnected the power plug while being in TTY and it triggered some kworker hangs and call traces, and to confirm I connected it back and removed again and some more got triggered and it ended up halting the system, so I think power delivery is definitely involved here

Last edited by techmanwalker (2025-07-15 02:36:24)

seth · 2025-07-15 07:12:24

And disabling the battery aware profile also didn't work, the laptop does the same impressive 20 FPS anyway

Completely disable power-profiles-daemon. If it still happens, there's a good chance the firmware powers down the system - do you have any power management related options in your "Fisher-Price coded" UEFI menu?
'cause

I connected it back and removed again and some more got triggered and it ended up halting the system, so I think power delivery is definitely involved here

Otherwise, is the notebook supported by asusctl/rog-control-center? (Afaiu asus might simply have delegated the entire FW control to userspace tools like myasus on windows)

techmanwalker · 2025-07-17 02:33:07

Even with no power daemon my laptop still hanged thrice both on battery and on AC and gives me the 20 fps as well. I had to reboot quite a lot these days to work on some programs, but right now the Nvidia gpu disappeared as well with the GSP firmware disabled, which has never happened until now: it didn't halt-freeze like other times and now the errors are different. (I was playing Cyberpunk a bit while)

jul 16 19:00:09 malasdecisiones kernel: nvidia-modeset: WARNING: GPU:0: Correcting number of heads for current head configuration (0x00)
jul 16 19:00:18 malasdecisiones kernel: NVRM: nvCheckOkFailedNoLog: Check failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from _memdescAllocInternal(pMemDesc) @ mem_desc.c:1353
jul 16 19:00:18 malasdecisiones kernel: NVRM: nvAssertOkFailedNoLog: Assertion failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from status @ kernel_gsp.c:4615
jul 16 19:00:18 malasdecisiones kernel: NVRM: nvAssertOkFailedNoLog: Assertion failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from kgspCreateRadix3(pGpu, pKernelGsp, &pKernelGsp->pSRRadix3Desc>
jul 16 19:00:18 malasdecisiones kernel: nvidia 0000:01:00.0: can't suspend (nv_pmops_runtime_suspend [nvidia] returned -5)
ksystemstats[24518]: Could not retrieve information for NVidia GPU "0000:01:00.0"

My laptop seems to not be supported by asusctl/rog-control-center as it's a Zenbook and the keyboard is not on lsusb as the manual says, but it is actually supported by MyASUS, so should I change some setting in it on Windows?

logs

Last edited by techmanwalker (2025-07-17 03:15:05)

seth · 2025-07-17 06:58:27

jul 16 16:15:37 archlinux kernel: pci 0000:01:00.0: [10de:25a2] type 00 class 0x030200 PCIe Legacy Endpoint

GPU is there…

jul 16 16:15:37 archlinux kernel: NVRM: loading NVIDIA UNIX Open Kernel Module for x86_64  575.64.03  Release Build  (root@malasdecisiones)  
jul 16 16:15:38 archlinux kernel: NVRM: testIfDsmSubFunctionEnabled: GPS ACPI DSM called before _acpiDsmSupportedFuncCacheInit subfunction = 11.
jul 16 18:48:33 malasdecisiones kernel: NVRM: RmHandleDNotifierEvent: RmHandleDNotifierEvent: Failed to handle ACPI D-Notifier event, status=0x11
jul 16 19:00:18 malasdecisiones kernel: NVRM: nvCheckOkFailedNoLog: Check failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from _memdescAllocInternal(pMemDesc) @ mem_desc.c:1353
jul 16 19:00:18 malasdecisiones kernel: NVRM: nvAssertOkFailedNoLog: Assertion failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from status @ kernel_gsp.c:4615
jul 16 19:00:18 malasdecisiones kernel: NVRM: nvAssertOkFailedNoLog: Assertion failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from kgspCreateRadix3(pGpu, pKernelGsp, &pKernelGsp->pSRRadix3Descriptor, NULL, NULL, gspfwSRMeta.sizeOfSuspendResumeData) @ kernel_gsp_tu102.c:1303
jul 16 19:00:24 malasdecisiones kernel: NVRM: Error in service of callback

You're running on nviida-open, the GSP is not disabled and I'm pretty sure I pointed out that you CANNOT disable the GSP w/ nvidia-open. It critically relies on it. Don't even try.

This is

jul 16 18:48:34 malasdecisiones systemd[1]: Stopped target Suspend.
jul 16 18:48:34 malasdecisiones suspend[16717]: nvidia-resume.service

somehow 12 minutes after a suspend and you get

jul 16 19:00:18 malasdecisiones kernel: NVRM: nvAssertOkFailedNoLog: Assertion failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from kgspCreateRadix3(pGpu, pKernelGsp, &pKernelGsp->pSRRadix3Descriptor, NULL, NULL, gspfwSRMeta.sizeOfSuspendResumeData) @ kernel_gsp_tu102.c:1303
jul 16 19:00:18 malasdecisiones kernel: nvidia 0000:01:00.0: can't suspend (nv_pmops_runtime_suspend [nvidia] returned -5)

and OOM on the GPU (probably from efforts to restore the VRAM) and that seems to make

jul 16 19:00:24 malasdecisiones kernel: NVRM: Error in service of callback 
jul 16 19:00:26 malasdecisiones steam[21489]: [2025-07-16 19:00:26.848] [MANGOHUD] [error] [nvidia.cpp:42] Getting device handle by PCI bus ID failed: Unknown Error
jul 16 19:00:26 malasdecisiones steam[21489]: [2025-07-16 19:00:26.848] [MANGOHUD] [info] [nvidia.cpp:75] NVML and NVCTRL are unavailable. Unable to get NVIDIA info. User is on DFSG version of mangohud?
jul 16 19:23:00 malasdecisiones ksystemstats[24518]: Could not retrieve information for NVidia GPU "0000:01:00.0"

*steam* believe the GPU isn't there because it currently (?) doesn't respond.
Have you checked nvidia-smi whether steam holds an outlier position here?

Also, I don't think we've seen any credible attempt to use the 570xx drivers as of yet?

actually supported by MyASUS, so should I change some setting in it on Windows?

You can at least look at it's power config whether there's anything battery dependent.

techmanwalker · 2025-07-20 07:32:36

seth wrote:

You're running on nviida-open, the GSP is not disabled

I was until #29 when the GPU vanished again. Then I inmediately switched to 570, disabled the GSP and the system soft-hanged again, which is why I'm posting here again (I'm not doing that frequently to gather more info on a larger timeframe)

*steam* believe the GPU isn't there because it currently (?) doesn't respond.

It's not just steam, it's nvtop and the games that run on way lower framerate or outright don't run (bc of the prime-run argument). Haven't checked directly on nvidia-smi, but any program that tries to use the NVIDIA gpu either fails or fallbacks to the iGPU.

You can at least look at it's power config whether there's anything battery dependent.

I'll check

Arch Linux

#26 2025-07-13 06:58:31

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#27 2025-07-13 08:32:17

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#28 2025-07-13 13:19:18

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#29 2025-07-14 17:07:15

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#30 2025-07-14 20:13:42

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#31 2025-07-14 20:35:36

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#32 2025-07-14 23:50:41

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#33 2025-07-15 07:12:24

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#34 2025-07-17 02:33:07

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#35 2025-07-17 06:58:27

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

#36 2025-07-20 07:32:36

Re: System stalled. Can't launch programs and Plasma becomes unresponsive.

Board footer