Hi, can you look into your "dmesg" when it crashes ?
It should look like this:
[16052.214603] NVRM: Xid (PCI:0000:01:00): 31, pid=68326, name=GameThread, Ch 0000005e, intr 00000000. MMU Fault: ENGINE GRAPHICS GPCCLIENT_GCC faulted @ 0xe40_00000000. Fault is of type FAULT_PDE ACCESS_TYPE_VIRT_READ
because this issue can be dupe of https://github.com/HansKristian-Work/vkd3d-proton/issues/1510
@Dahipppo1555 You are right , I have :
[ 2466.946427] NVRM: Xid (PCI:0000:01:00): 31, pid=8479, name=GameThread, Ch 0000011f, intr 00000000. MMU Fault: ENGINE GRAPHICS GPCCLIENT_GCC faulted @ 0x6330_00000000. Fault is of type FAULT_PDE ACCESS_TYPE_VIRT_READ
What is the cause of this bug ?
What is the cause of this bug ?
Impossible to say without lots of painful debugging.
Maybe the post about Witcher 3 freezing a few minutes into gameplay is related.
I've narrowed it down to commit https://github.com/HansKristian-Work/vkd3d-proton/commit/21a920d529e6644f94443e439f3a8e453f6171cb -> vkd3d: Re-enable KHR_present_wait on NV 535.xx+.
https://github.com/HansKristian-Work/vkd3d-proton/issues/1638
NVRM error: Xid (PCI:XXXX:XX:XX): 109 appears most often:
[ 2025.929023] NVRM: Xid (PCI:0000:26:00): 109, pid=11323, name=GameThread, Ch 0000006e, errorString CTX SWITCH TIMEOUT, Info 0x26c03d
I see no improvement on a Tesla T4 from disabling KHR_present_wait. A fast way to reproduce the bug is open and close the map. If it doesn't crash within a few seconds of closing the map, repeat. You must be moving either on foot or in a car, but on foot doesn't result in the Vulkan device lost error -4, while in a car usually does.
yeah disabling KHR_present_wait doesn't do anything.
but it's not just cars , crash happens in braindances and conversations with npcs
Bump. Same error in dmesg: [11226.556628] NVRM: Xid (PCI:0000:0a:00): 31, pid=30942, name=GameThread, Ch 0000005f, intr 00000000. MMU Fault: ENGINE GRAPHICS GPCCLIENT_GCC faulted @ 0x6960_00000000. Fault is of type FAULT_PDE ACCESS_TYPE_VIRT_READ .
Cyberpunk version: 2.0
GPU: RTX 2060 Super
Nvidia driver: 535.113.01
Proton: GE 8-16
Happens mostly with map interaction or when going fast in a car.
I can't prove this is true, but the game lasts longer if I bottleneck the CPU, limiting frames. When I use higher quality settings and the bottleneck rests on the GPU, the game crashes instantly.
Game plays fine until you drive a car and open up any kind of menu with a hotkey (inventory, map, quests). The game crashes hard within 2 seconds, requiring it to be killed.
Dmesg shows:
NVRM: Xid (PCI:0000:0a:00): 31, pid=37878, name=GameThread, Ch 00000057, intr 00000000. MMU Fault: ENGINE GRAPHICS GPCCLIENT_GCC faulted @ 0x4df0_00000000. Fault is of type FAULT_PDE ACCESS_TYPE_VIRT_READ
The vkd3d errors in the proton log show:
`err:d3d12_command_queue_execute: Failed to submit queue(s), vr -4
Driver: 535.43.09
GPU: RTX 2080Ti
Kernel: 6.5.5
VKD3D: vkd3d-proton@089df9c
Proton: GE 8-15
Vulkan: 1.3.262
Distro: Gentoo
DE: i3 version 4.21-126-g30131ed6+
CPU: Ryzen 9 5950x
someone has found a workaround by using an older driver version :
https://forums.developer.nvidia.com/t/cyberpunk-2077-1-62-1-63-crashes/260799/14?u=bensalem.amine321
I haven't tested this claim since my kernel version doesn't allow me to install older versions of nvidia drivers , but it seems the version 510.108.03 prevents the Xid 31 crashes , at the cost of not being able to run DLSS and a 10% perf decrease.
I have been playing for a few days now using VKD3D_DISABLE_EXTENSIONS=VK_KHR_present_wait on driver 535.43.09 and the frequency of my crashes has gone down. The game still has a tendency to crash when mixing menus with driving, but I am able to get anywhere from 10 minutes to 3 hours of gameplay before I get the Xid error, making it difficult to diagnose. I am unsure if disabling VK_KHR_present_wait had an actual effect on stability, or if the game just stabilized after a bit of the shader cache had accumulated.
hmm , I don't think so , I remember disabling present_wait and it didn't really change anything.
Despite it not being directly related, I did a quick test with the nv_dgcc_enable branch that enables VK_NV_device_generated_commands_compute with the latest vulkan beta driver 535.43.10. This shows no significant improvement to the crashing. However, it does solve all of the crashing problems in Starfield, but not the performance problems.
Driver: 535.43.10
GPU: RTX 2080Ti
Kernel: 6.5.5
VKD3D: https://github.com/HansKristian-Work/vkd3d-proton/commit/01f62c874ca5f580f8c6c3e202189f2eb55aaaa9
Proton: GE 8-15
Vulkan: 1.3.262
Distro: Gentoo
DE: i3 version 4.21-126-g30131ed6+
CPU: Ryzen 9 5950x
The crash is now resolved with the 545 beta driver . closing the issue .
Since 1.62 , the game crashes in 5 min to 10 min . No matter which version of proton , wich version of GE proton , experimental , hotfix , the game crashes.
It is the only game that crashes in this manner , I do not have any problem on other games , on any settings.
If you check the logs , there's this error that is shared between the two :
Also , on protonDB , you'll notice that most people who experience this crash have a 20xx RTX...
Software information
Cyberpunk 2077 , any setting.
System information
Log files
GE proton 8-1 :
steam-1091500.log
Proton 8.0-2 :
steam-1091500.log