Elden Ring freezes in Limgrave

[jnack95]

Game plays normally (launched from steam) until I get to Limgrave and then it always freezes the screen, the music continues and I have to force quit. My system is as follows:

Alienware aurora r10 w/ AMD Ryzen 7 5800, Nvidia RTX 3070 (OEM Asus)
Dist kernel 6.6.62
Nvidia-drivers 550.135 and also happens with new 565.77.


I had a similar problem in Lies of P but I found that setting the max FPS to unlimited solved the problem. Limiting the framerate in Lies of P always caused a freeze similar to the one in Elden Ring. In Elden Ring I have tried the following with no luck:
-Trying all available protons, experimental and latest GE
- Enabling/disabling steam overlay, shader precompile, background vulkan processing in all sorts of configurations.
- Disabling v-sync by passing PROTON_NO_FSYNC=1 %command%
- Passing __GL_SHADER_DISK_CACHE_SKIP_CLEANUP=1 %command%
- emerging game mode and passing gamemoderun %command%
- sysctl -w vm.max_map_count=1048576
- Flatpak steam vs native


I've even disassebled my video card and applied new thermal paste. I have no idea what to try next, other than get a new video card or try a different dist. Anybody encountered something similar? Thanks

[Hu]

A full freeze in a graphics heavy program suggests to me a graphics driver problem. Is this reproducible without the proprietary nVidia drivers? Is there anything logged in dmesg when the game hangs?

[jnack95]

I agree seems like a driver problem but since nobody else seems to have the same problem, I'm not sure. I haven't tried the nouveau drivers but I have tried the nvidia-drivers with the kernel-open flag set and it didn't seem to change anything. Interestingly I did check the dmsg and here is the output from when i start the game until it crashes :

[ 92.303974] nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.
[ 92.329270] nvidia-uvm: Loaded the UVM driver, major device number 511.
[ 92.610831] Disabled but requested by user space. You should enable Landlock at boot time: https://docs.kernel.org/userspace-api/l ... figuration
[ 148.313539] input: Xbox Wireless Controller as /devices/virtual/misc/uhid/0005:045E:0B13.0005/input/input19
[ 148.313721] hid-generic 0005:045E:0B13.0005: input,hidraw4: BLUETOOTH HID v5.20 Gamepad [Xbox Wireless Controller] on e8:9c:25:5d:ce:ca
[ 148.405120] loaded hid-xpadneo v0.9.6
[ 148.405149] xpadneo 0005:045E:0B13.0005: BLE firmware version 5.20
[ 148.405151] xpadneo 0005:045E:0B13.0005: pretending XB1S Windows wireless mode (changed PID from 0x0B13 to 0x028E)
[ 148.405153] xpadneo 0005:045E:0B13.0005: working around wrong SDL2 mappings (changed version from 0x00000520 to 0x00001130)
[ 148.405156] xpadneo 0005:045E:0B13.0005: report descriptor size: 283 bytes
[ 148.405158] xpadneo 0005:045E:0B13.0005: fixing up Rx axis
[ 148.405159] xpadneo 0005:045E:0B13.0005: fixing up Ry axis
[ 148.405160] xpadneo 0005:045E:0B13.0005: fixing up Z axis
[ 148.405161] xpadneo 0005:045E:0B13.0005: fixing up Rz axis
[ 148.405163] xpadneo 0005:045E:0B13.0005: fixing up button mapping
[ 148.405213] xpadneo 0005:045E:0B13.0005: enabling compliance with Linux Gamepad Specification
[ 148.405271] input: Xbox Wireless Controller as /devices/virtual/misc/uhid/0005:045E:0B13.0005/input/input20
[ 148.405379] xpadneo 0005:045E:0B13.0005: input,hidraw4: BLUETOOTH HID v11.30 Gamepad [Xbox Wireless Controller] on e8:9c:25:5d:ce:ca
[ 148.405383] xpadneo 0005:045E:0B13.0005: controller quirks: 0x00000050
[ 149.396142] xpadneo 0005:045E:0B13.0005: Xbox Wireless Controller [40:8e:2c:3e:a0:28] connected
[ 284.217358] NVRM: GPU at PCI:0000:0b:00: GPU-084af9c9-4830-49dc-03da-f0c329a2220b
[ 284.217364] NVRM: Xid (PCI:0000:0b:00): 109, pid=3532, name=eldenring.exe, Ch 0000002b, errorString CTX SWITCH TIMEOUT, Info 0x14c02d

Let me know if you see anything interesting

[Hu]

None of that looks suspicious to me. A kernel log message on failure isn't guaranteed, but it would've made things easier if there had been one.

[jnack95]

Actually, the last line is a big clue. the Xid:109 CTX SWITCH TIMEOUT error is well known and the symptoms are the exact same as mine and after two years, there is still no solution. bummer
More info here....

https://forums.developer.nvidia.com/t/m ... 235459/450

[Ralphred]

NVRM: Xid (PCI:0000:0b:00): 109

Xid 109 errors have been plaguing nvidia for a couple of years, including "fixes" that pushed a regression causing them in games that never had them before...

www.protondb.com used to be a really good resource for environment variable workarounds, but a quick look at a couple of recent releases and the skew towards AMD card users is noticeable.
PROTON_HIDE_NVIDIA_GPU=0 PROTON_ENABLE_NVAPI=1 shows up a lot and VKD3D_CONFIG=dxr11 VKD3D_FEATURE_LEVEL=12_2 is mentioned a couple of times.

- Disabling v-sync by passing PROTON_NO_FSYNC=1 %command%

fsync doesn't do that, it's to do with file handling at the kernel level to reduce wine overhead.

I had a similar problem in Lies of P but I found that setting the max FPS to unlimited solved the problem. ~~~ In Elden Ring I have tried the following with no luck

The fps cap in Elden Ring is a hard coded studio thing, the Dark Souls games are capped at 60fps too. Just searching for "elden ring frame cap patch" will give a list of ways to unlock it, though they will require you to sidestep EAC and give up on co-op play.

I always get better results using mangohud to cap fps and turn on/off vsync, vsync instructions are near the bottom of the readme. Check with vkcube that what you are asking for is being done and not overridden by some default X setting.

Maybe a 55fps cap in mangohud undermining the 60fps from the game might help too?

The whole nvidia thread is here.

[jnack95]

Thanks for the reply.

The fps cap in Elden Ring is a hard coded studio thing, the Dark Souls games are capped at 60fps too.

That makes sense....was wondering why it pegged at 60. I'm sure this is the crux of my problem. Anytime a game tries to throttle the fps, it crashes with Xid 109. It seems like it has to be able to constantly vary the framerate for whatever reason.

PROTON_HIDE_NVIDIA_GPU=0 PROTON_ENABLE_NVAPI=1 shows up a lot and VKD3D_CONFIG=dxr11 VKD3D_FEATURE_LEVEL=12_2

Unfortunately, those launch options didn't have any effect. I haven't ventured into the mangohud world yet and I'm not sure how much more time I want waste on this. Might just be easier to get a new AMD card. Thanks for your help

[jnack95]

Im thinking this problem has something to do with shader caching and, specifically, Elden ring shader caching. If I play Lies of P with shader pre-caching enabled I get XiD 109 errors all day. When I turn of shader pre-caching and clear the cache, the game play amazing...dlss,1440p 144fps. If I then go to elden ring and it crashes, when I go back to Lies of P, it also crashes again (even with shader pre-caching turned off) until I clear the cache and restart steam. Something weird going on with how nvidia clears it shader cache.... and what gets left in memory.

This article seems to kind of get at it a bit too.
https://www.ign.com/articles/valve-expl ... steam-deck

[Ralphred]

I haven't ventured into the mangohud world yet

The reason I use it's fps cap is it doesn't interfere with freesync: Back when the shader caches were all "locally generated" the frame spikes were palpable, forcing something capable of running at 90 fps down to 50 freed up enough resources that the frametime jitter was "liveable" (after a couple of hours of gameplay you got one frametime spike per ~15 minutes, but games vary so...) and as long as I didn't drop below 48fps, my monitors lower freesync limit, it was buttery smooth.

Might just be easier to get a new AMD card.

I understand the sentiment, I picked up a 6600 for under £220 for my lad, included the firmware in the kernel and he's happy as Larry with his "not potato mode" graphics, but cards from the 20x0 and 30x0 era were so expensive I just feel like your getting ripped off...