@zebcom I’ve tried with DLSS and DLFG disabled and the hang still happened, although this time I was actually able to get much closer to the temple before it did:
@doomperfect00 I can’t seem to be able to reproduce the hang in your spot:
Interestingly enough, mine doesn’t result in Xid 109 being printed anywhere in dmesg.
The First Descendant now seems to be broken and always crashes on load (after the main menu). I tried deleting shaders and also did a full reinstall. I also did run steam from CLI and didn’t see anything. I am running this on Bazzite, switching to the an older driver version completely fixes this. When crashing the game pops up this error message twice:
nvidia-bug-report.log.gz (386.0 KB)
I can confirm this regression in The Witcher 3. For me, it occurs on 590.44.01 and 595.45.04 for DX12, FULL RT, DLSS and Uber+ settings.
The game is only stable after reverting to version 580.142.
Elden Ring does not work with HDR anymore with 595 and GE-Proton32. Launch options: PROTON_ENABLE_WAYLAND=1 PROTON_ENABLE_HDR=1 MANGOHUD=1 gamemoderun %command%
However, Overwatch (same launch options) and the MPV player work perfectly with HDR. The package vk-hdr-layer-kwin6-git is already uninstalled.
Specs: RTX 4090, Ryzen 7 9800X3D
My experience with the issues I’ve had from previous versions:
The graphics still crashes for me when trying to enable HDR: Graphics driver hangs when enabling HDR Edit: only on a Gigabyte G27Q monitor
LFC is still broken: Wildly fluctuating refresh rate and flicker at low framerates , Gsync compatible VRR - Low Framerate Compensation (LFC) transition not seamless, unlike on Windows (though this issue has been acknowledged now and is being tracked by Nvidia in issue 5374195).
I am getting these Xid errors as well. @tayler Is it the mission where Leon drives into Raccoon City? If I have /WineDetectionEnabled:False in my launch parameters, I can reproduce the Xid error on that level every time.
@amrits Here’s my nvidia logs. I can also provide my save files if you need them.
nvidia-bug-report.log.gz (595.6 KB)
Summary: Gamescope embedded mode (DRM backend) shows persistent visual artifacts/corruption on the Steam UI and when the Steam overlay appears. The issue was previously reported against the 580 series driver here: 580 release feedback & discussion - #427 by alneor.ellyuan — the reporter confirmed the issue disappears when downgrading to 575.64.05. The bug persists in driver 595.45.04.
System:
Details: Artifacts appear as colored rectangles/blocks on the Steam Gaming Mode UI. They temporarily disappear when the mouse cursor is moved, then reappear when the cursor goes idle. The issue is not present in KDE Plasma desktop mode (KWin compositor) — only in Gamescope embedded/DRM session.
Attempted workarounds that did NOT help:
This appears to be the same nvidia-drm regression introduced in the 580 series, as described in the linked post. Would appreciate if this could be tracked and addressed.
Related gamescope issue: [DRM,NVIDIA] - Embedded Gamescope: Steam notifications/overlay cause artifacts and flickering · Issue #1964 · ValveSoftware/gamescope · GitHub
Interesting that it doesn’t crash for you, and yours isn’t Xid 109… and right! Thanks for the tip, I didn’t realise you needed vkd3d-proton debug build for breadcrumbs. Took me a while to figure out where to replace the dll files (for proton-cachyos, it turned out to be in /usr/share/steam) to get it to work.
Unfortunately AC Shadows still hangs in the exact same place with breadcrumbs enabled with the same Ctx switch timeout. Attaching my vkd3d breadcrumbed log here, in the hopes it helps someone debug…
vkd3d.tar.gz (354.4 KB)
Also attaching my savegame, in case it’s a game condition dependent crash (from <steam library>/steamapps/compatdata/3159330/pfx/drive_c/Program Files (x86)/Ubisoft/Ubisoft Game Launcher/savegames/). Just run towards the custom Kuji-Kiri marker I already set (green brackets on minimap) immediately after loading in, and it will crash.
savegames.tar.gz (487.3 KB)
Reattaching bug report also, just for completeness:
nvidia-bug-report.log.gz (626.3 KB)
I can confirm that upgrading to 595 (from 590) fixed the following issues for me: Stalker 2 freezing and crashing with CTX SWITCH TIMEOUT messages in the logs, and The Outer Worlds 2 freezing when hardware ray tracing is enabled.
could you guys also make a repo for fedora 44 at Index of /compute/cuda/repos on the next driver update?
otherwise, there won’t be packages for multiple months after its estimated April 14th release date
595 has introduced two issues for me (CachyOS, 64 GB DDR5, 9950X3D, RTX 5080, MSI B850 EDGE TI WIFI, DELL U2723QE 4K dual-monitors daisy chained):
Both issues were solved after downgrading back to 590.48.01.
There already was a bug in the previous release, and this release made it much, much worse:
While the system appears to be fine, the journal grows at a fantastic rate at 3000 entries per minute.
This driver enables the nvidia-drm modeset=1 parameter by default, which makes it set a higher-resolution framebuffer console. Maybe that’s not working for some reason. Can you please generate and attach a bug report log? You can also try adding nvidia_drm.fbdev=0 to turn off the framebuffer console behavior and see if that makes the password prompt appear.
I can reproduce the Elden Ring freeze on RTX 5090 as well, same Ctx switch timeout. Happens to me about 3-5 seconds after loading into game (worked perfectly fine on 590 drivers). Attached bug report during freeze.
As mentioned in the post below, you could try the descriptor heap branch of vkd3d-proton? Might at least get your game running while the issue is investigated by Nvidia…
[37919.357435] NVRM: GPU at PCI:0000:01:00: GPU-361d002b-e0e7-bab8-4ea1-8959234e4b25
[37919.357439] NVRM: GPU Board Serial Number: 0
[37919.357439] NVRM: Xid (PCI:0000:01:00): 109, pid=126101, name=eldenring.exe, channel 0x00000029, errorString CTX SWITCH TIMEOUT, Info 0x3c03f
nvidia-bug-report.tar.gz (566.8 KB)
Very interesting finding: if I use vkd3d-proton built from the descriptor-heap-test branch (here), both the AC Shadows and Elden Ring Ctx switch timeout freezes no longer happen for me. Changed nothing else in the configs for either, and both work now with 595.
Caveats: Elden Ring unfortunately seems to have some weird flickering and a large performance hit with that branch when ray-tracing is enabled (is fine if you disable RT), but AC Shadows seems to work pretty okay at all max settings + DLSS preset L + FG. It does now freeze occasionally, like every 30 mins or so, which it never did on 590 drivers (no Ctx switch timeout errors anymore though), and I haven’t really compared performance yet, but at least I can play without crashing now…
Can confirm that The Talos Principle: Reawakened does not crash anymore when enabling raytracing.
Unfortunately, I still seem to need VK_hdr_layer + ENABLE_HDR_WSI=1 to have working HDR output in Proton. However, it’s worth noting that I have a slightly non-standard setup: dual GPU laptop running in hybrid mode, plus EDR is used in KDE Plasma instead of a true HDR display.
nvidia-bug-report.log.gz (893.9 KB)
Can you try Ghost of Tsushima by any chance? For my RTX 5080, aside from Elden Ring, it’s another game where 595 produces a Ctx switch timeout (see my report above) while it works on 590. And for this one, the descriptor-heap-test branch doesn’t seem to work for me. Also, are you on GNOME? I’m wondering if the compositor is also relevant for this problem.
Seems to me there is some more general problem in 595 with either just the 5080 or the whole 5000 series, when multiple older games that usually ran well before consistently produce the Ctx switch timeout quickly after the rendering starts.
Sorry, I totally missed reading your report before!
I don’t have Armored Core, but I tried Ghost of Tsushima just now on Proton-GE on whatever vkd3d-proton it ships with; same as you, it froze with Ctx switch timeout as soon as the main menu rendered. But my descriptor-heap-test build (with proton-cachyos) seems to solve that issue and work pretty well for me, I just played for about 20 mins around the Hiyoshi Springs area without any issues. Playing on all max settings.
I am on hyprland, not GNOME, and using a 5090. I guess the compositor/rendering backend could matter here, but given that swapping vkd3d out makes a difference and we both have the freezes without descriptor heaps, I’m inclined to agree it’s a more fundamental driver issue with Vulkan (and maybe with the 5000 series, specifically). I guess you could try to swap compositors and see if it helps with Ghost of Tsushima?
I think it’s some timing issue, perhaps also on the Linux side. As I said, when I switch the monitor off and on again, I see the prompt and it’s also in a high resolution / different font.
How would I generate a report log at that stage? Could be difficult because the hard drive is not even available at that stage.
Can you try Ghost of Tsushima by any chance? For my RTX 5080, aside from Elden Ring, it’s another game where 595 produces a Ctx switch timeout (see my report above) while it works on 590. (…) Also, are you on GNOME? I’m wondering if the compositor is also relevant for this problem.
FWIW, I’ve also noticed the Xid 109 error with Ghost of Tsushima on a 5080 under GNOME.
I also noticed SpaceEngine often, but inconsistently, failing with Xid 13 and 31, but didn’t have time to explore it further. It does appear to be a regression in 595, but I have a dual GPU setup with a 3080 12 GB, so I don’t know if this or something else could not be introducing noise. I could try testing more if there is interest in pursuing those problems for the 595.