The existing thread was locked without reason given:
It’s a persistent issue that still exists and affects many games using both dxvk and vkd3d. It has been over three years since it was first reported.
Is this being looked at?
I can semi-reliably make this occur in quite a few games if there was somehow I could help debug this.
The only thing available in the log is this, which isn’t of much help:
NVRM: Xid (PCI:0000:0a:00): 31, pid=5595, Ch 0000002b, intr 10000000. MMU Fault: ENGINE GRAPHICS GPCCLIENT_T1_0 faulted @ 0x5a_00600000. Fault is of type FAULT_PDE ACCESS_TYPE_READ
This isn’t an issue with DXVK or VKD3D on AMD or Intel products, and developers of said libraries have implicated it as being driver bugs.
Further testing seems to suggest this is possibly actually a regression so it may have been fixed at some point?
After trying many different driver versions, I’m not seeing the issue appear using 465.31.
This is still an issue with Ubuntu 22.10 (development edition) using the 5.15 driver. In the previous thread, an Nvidia employee hinted at an internal bug number (2432712), which we can’t see, but since we can’t see it, there’s no way to know if it’s for sure the same issue… however this bug impacts Vulkan only, so it seems entirely possible it could be present on Windows and no one has noticed… it’s a real shame as I will have to ditch my Nvidia card for AMD as this has been ongoing for almost 4 years now… loads of my games work just fine on my AMD laptop with Integrated GPU, but on my nice “gaming” rig some stuff just crashes out randomly, it’s a real pain in network play.
Come on Nvidia, game breaking bugs open for 4 years is not cool.
Ran into this issue as well with Starship Simulator Demo and driver 535.54.03
NVRM: GPU at PCI:0000:26:00: GPU-6f98b267-20cc-5347-51dc-8bad07fd2ad0
NVRM: Xid (PCI:0000:26:00): 31, pid=1000, name=GameThread, Ch 0000006e, intr 00000000. MMU Fault: ENGINE HOST0 HUBCLIENT_ESC faulted @ 0x0_19d00000. Fault is of type FAULT_PTE ACCESS_TYPE_VIRT_READ
Hopes where high it was also fixed by this issue which seems to be solved now, but the above error is still there using VKD3D in version 2.8.
With Proton Experimental using VKD3D 2.9 this issue seems to be circumvented by the compatibility layer as the game ran stable for several hours straight.
I recommend re-testing your games with VKD3D 2.9 and see if the issue is also solved this way.
Maybe it’s not a driver bug but an application error doing something illegal.
nvidia-bug-report.log.gz (664.2 KB)
Exactly my case, will this isn’t promising for a bug with this effect to be unsolved, my system slows, and I’ve to do so mental gymnastics, and the gift of ungraceful reboot/shutdown. you are not alone, bro.
Jun 26 16:46:59 Korialo-Arch kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module 535.54.03 Tue Jun 6 22:20:39 UTC 2023
Jun 26 16:53:49 Korialo-Arch kernel: NVRM: GPU at PCI:0000:01:00: GPU-b35c4bab-d26b-64e4-1fb3-f76cc2b2f16a
Jun 26 16:53:49 Korialo-Arch kernel: NVRM: Xid (PCI:0000:01:00): 31, pid=15212, name=mpv/vo, Ch 0000001e, intr 00000000. MMU Fault: ENGINE GRAPHICS GPCCLIENT_PROP_0 faulted @ 0x1_01a12000. Fault is of type FAULT_PDE ACCESS_TYPE_VIRT_READ