I’m not very familiar with BOINC so I’m having trouble testing this. I’m trying to run the einstein_O3AS_1.25_x86_64-pc-linux-gnu__GW-cuda-4 binary by itself with the parameters that the project says it launched it with in client_state.xml, but it prints “Adding --GPUDevice=-1” to stderr.txt and then complains that --GPUDevice is specified twice. Do you know how to run this in a debuggable way?
1 Like
Thanks for testing this out. With 580 drivers I was not able to even make it into the game as the freeze happens regardless around 5 seconds after landing on the main menu. I can get about as far as clicking “Continue” in the menu, I’ll see the options for Open/Group/Solo but never any further.
I’ll install the driver again and try again. Are there any other details that I can provide that may help?
Many thanks.
EDIT: I just tried again with the new drivers, same result. Elite Dangerous freezes 5-10 seconds after landing on the main menu.
❯ pacman -Q | grep -i 'nvidia'
lib32-nvidia-utils 580.76.05-2
lib32-opencl-nvidia 580.76.05-2
libva-nvidia-driver 0.0.14-1.1
linux-cachyos-nvidia-open 6.16.4-2
linux-firmware-nvidia 1:20250808-1
nvidia-settings 580.76.05-2
nvidia-utils 580.76.05-4
opencl-nvidia 580.76.05-4
~
❯ sudo dmesg | grep -i 'nvrm'
[ 5.692775] NVRM: loading NVIDIA UNIX Open Kernel Module for x86_64 580.76.05 Release Build (notroot@) Fri Aug 29 18:35:29 UTC 2025
[ 244.439010] NVRM: GPU at PCI:0000:01:00: GPU-9e6dc988-ab46-8294-6c4a-5f8fafa84c15
[ 244.439013] NVRM: Xid (PCI:0000:01:00): 13, pid=9392, name=EliteDangerous6, Graphics Exception: channel 0x00000053, Class 0000c9c0, Offset 00000000, Data 00000000
EDIT 2: I tried moving from Proton 10 to Proton Experimental. I was able to load into the game, fly to the nearest station, but after de-activating supercruise and approaching the station it crashed with the same error after about 6-8 seconds.
1 Like
This is still an issue with 580.82.07
2 Likes
Never mind, I managed to track down the problem and filed internal bug #5497358.
2 Likes
Driver 580.82.07. Ghost of Tsushima. NVMR Xid 13 still not fixed. Not reproducible on 575 driver.
modinfo nvidia
filename: /lib/modules/6.16.4-arch1-1/extramodules/nvidia.ko.zst
alias: char-major-195-*
version: 580.82.07
supported: external
license: NVIDIA
firmware: nvidia/580.82.07/gsp_tu10x.bin
firmware: nvidia/580.82.07/gsp_ga10x.bin
srcversion: 3B7BEF3266B8498664D3B57
alias: pci:v000010DEd*sv*sd*bc06sc80i00*
alias: pci:v000010DEd*sv*sd*bc03sc02i00*
alias: pci:v000010DEd*sv*sd*bc03sc00i00*
alias: of:N*T*Cnvidia,tegra264-displayC*
alias: of:N*T*Cnvidia,tegra264-display
alias: of:N*T*Cnvidia,tegra234-displayC*
alias: of:N*T*Cnvidia,tegra234-display
depends:
name: nvidia
retpoline: Y
vermagic: 6.16.4-arch1-1 SMP preempt mod_unload
parm: NvSwitchRegDwords:NvSwitch regkey (charp)
parm: NvSwitchBlacklist:NvSwitchBlacklist=uuid[,uuid...] (charp)
parm: NVreg_ResmanDebugLevel:int
parm: NVreg_RmLogonRC:int
parm: NVreg_ModifyDeviceFiles:int
parm: NVreg_DeviceFileUID:int
parm: NVreg_DeviceFileGID:int
parm: NVreg_DeviceFileMode:int
parm: NVreg_InitializeSystemMemoryAllocations:int
parm: NVreg_UsePageAttributeTable:int
parm: NVreg_EnablePCIeGen3:int
parm: NVreg_EnableMSI:int
parm: NVreg_EnableStreamMemOPs:int
parm: NVreg_RestrictProfilingToAdminUsers:int
parm: NVreg_PreserveVideoMemoryAllocations:int
parm: NVreg_EnableS0ixPowerManagement:int
parm: NVreg_S0ixPowerManagementVideoMemoryThreshold:int
parm: NVreg_DynamicPowerManagement:int
parm: NVreg_DynamicPowerManagementVideoMemoryThreshold:int
parm: NVreg_EnableGpuFirmware:int
parm: NVreg_EnableGpuFirmwareLogs:int
parm: NVreg_OpenRmEnableUnsupportedGpus:int
parm: NVreg_EnableUserNUMAManagement:int
parm: NVreg_MemoryPoolSize:int
parm: NVreg_KMallocHeapMaxSize:int
parm: NVreg_VMallocHeapMaxSize:int
parm: NVreg_IgnoreMMIOCheck:int
parm: NVreg_NvLinkDisable:int
parm: NVreg_EnablePCIERelaxedOrderingMode:int
parm: NVreg_RegisterPCIDriver:int
parm: NVreg_RegisterPlatformDeviceDriver:int
parm: NVreg_EnableResizableBar:int
parm: NVreg_EnableDbgBreakpoint:int
parm: NVreg_EnableNonblockingOpen:int
parm: NVreg_CoherentGPUMemoryMode:charp
parm: NVreg_RegistryDwords:charp
parm: NVreg_RegistryDwordsPerDevice:charp
parm: NVreg_RmMsg:charp
parm: NVreg_GpuBlacklist:charp
parm: NVreg_TemporaryFilePath:charp
parm: NVreg_ExcludedGpus:charp
parm: NVreg_DmaRemapPeerMmio:int
parm: NVreg_RmNvlinkBandwidth:charp
parm: NVreg_RmNvlinkBandwidthLinkCount:int
parm: NVreg_ImexChannelCount:int
parm: NVreg_CreateImexChannel0:int
parm: NVreg_GrdmaPciTopoCheckOverride:int
parm: rm_firmware_active:charp
Error:
[ 1143.592614] NVRM: GPU at PCI:0000:01:00: GPU-71a75b40-96ff-20fd-1903-bc22f816be2b
[ 1143.592619] NVRM: Xid (PCI:0000:01:00): 13, pid=16620, name=GhostOfTsushima, Graphics Exception: channel 0x00000083, Class 0000c9c0, Offset 00000000, Data
00000000
All ther same on X and Wayland and different settings, etc. Works only on 575 and previous drivers. I believe it is different from announced fix for hanging Wayland applications.
EDIT:
4070 Ti SUPER. Not reproducible on 5070 Ti by the way, so newer cards seem unaffected.
1 Like
So Metro Exodus EE is probably still broke too. Beta versions, why do they even exist?
gnome-shell hang on login bug still exists. Nice.
1 Like
NVRM: Xid (PCI:0000:0c:00): 13, pid=7728, name=MetroExodus.exe, Graphics Exception: channel 0x00000051, Class 0000c9c0, Offset 00000000, Data 00000000
Very nice.
Edit: tried Smooth motion on GTA 5: Enhanced. It breaks the Rockstar launcher.
Tried it on The Finals.
No Smooth Motion: 138 FPS.
With Smooth Motion: 90 FPS.
So now there is 3 broken forms of frame gen on Linux: DLSS3, DLSS4, and Smooth Motion.
Amazing. 10/10 quality control. RTX: It’s ON.
id like to report back that the latest driver resolved my issues, cyberpunk2077 runs as it should now within expectations and no other glitches from the driver in my other games have been found as of now
1 Like
Sure, Thanks for the update. I would like to know which build version you are using now.
1 Like
We gotta start somewhere. DLSS on Linux is close to perfect. Huge thanks NVIDIA for bringing Smooth Motion too!
In fact, your result (138 FPS without smooth motion, 90 FPS with smooth motion) is probably working completely correctly. Smooth Motion works by doubling the framerate (interpolating every 2nd frame) but not exceeding your monitor output’s maximum refresh rate. So the 90 FPS rendered by the GPU is then doubled with smooth motion and is probably 180 FPS on the output port. I do not believe for a second that Linux framerate counters are able to understand DLSS/Frame-Gen/Smooth Motion framerates yet. On Windows, such “frame generation FPS” measuring was invented recently by Valve and required special analysis hooks.
PS: Smooth Motion works with any game. It works by restricting the game’s own renderer’s framerate (forced sync) and then generating interpolated frames between each engine-rendered frame. It’s meant as a tool for frame generation in all games that don’t support DLSS/FSR natively. I am so grateful that NVIDIA brought it to Linux too! <3 <3 <3
1 Like
Thank you very much!
1 Like
@vrachatte anything you need from me in terms of debugging? I’m not sure what else I can provide, but I think the amount of NVRM Xid 13 failures for the DX12 games is good enough already. They all have the same class 0000c9c0 in the messages,
1 Like
Wtf are you talking about? DLSS works perfectly in every game for me on 2070 Super and 3090.
2 Likes
FRAME GENERATION
You’re quite unpleasant, eh?
6 Likes
DLLS and Frame Generation have been working without issues for me on Linux
2 Likes
4090 here, dlss and fg work fine for me on Arch. What problems should i have?
1 Like
To be fair, 3x and 4x frame generation have issues on my system too, but I’m inclined to believe that this is an implementation issue with the games. For me they’re really really stuttery and mangohud shows the frametimes in a sawtooth pattern with the minimum frametime being 0,1ms and the maximum frametime being ~12ms
When I then change the settings in the game around 20 times, tab in and out several times at some point it fixes itself and then the frametime graph is smooth and the frametimes are between 4ms and 7ms.
It’s particularly noticeable in the Cyberpunk 2077 benchmark run. With 3x or 4x FG when this sawtooth pattern appears in Mangohud it looks to my untrained eyes as if the frames were drawn out of order. Areas where this can be seen very clearly are the big columns in the bar, on one of them there’s a newspaper that says “on a roll” or something like that, this jitters heavily when the sawtooth pattern is present, then when the camera moves past the bar and barkeeper the bottles on the shelf in the back start to jitter strongly just when you move past, and so does the bottle on the billiards table. Then when the camera moves outside there are some bushes on the left side of the alley and they too jitter heavily while the sawtooth pattern is present in mangohud.
After the 30 minutes of settings fiddling when the frametime graph turns smooth then gone is all the jitter from all of those described places.
Unfortunately I don’t know if this is even noticeable on a lower-refresh rate monitor. When I am testing this on my 5090 FE I get an average fps of ~290 and a minimum fps of ~260 with frame generation x4, and my monitor is a 240hz monitor.
Given that Cyberpunk is the only game I have with frame generation and given that I don’t even play cyberpunk and only use the benchmark as a stress test, this is not really a big issue for me personally.
Having no other games to test this with I can’t say for sure that it’s an implementation issue with Cyberpunk, but I heavily lean in that direction.
Edit: Added a screenshot, tho this screenshot is 1½ months old. That big spike is very likely caused by the screenshot utility, so ignore that as an outlier.
Edit2: The screenshot is 1½ months old and the issue is probably older, but it’s still present in the 580 driver
1 Like