between the 2 of these:
a page fault typically is caused by attempting to access memory well outside of alloc: either it exceeds the possible values alloc could ever give (typically a value greater then the amount of RAM physically in the system), or is registered to a program the current program does not have access to read from or write to.
an MMU fault is technically the same thing, but for GPU.
in escalating severity response:
this can be caused in one of 2 ways either what the driver responded with: one of your shaders is incorrectly compiled for the GPU, or corrupted
- if you are rolling your own rendering or shader compilation you might have some work to do to ensure the translation to at least nVidia GPU is correct
- if you are using purely out of the box material shader functions/nodes then you might need to create an MVP that experiences this issue (figuring out which is doing it) and submit the material as part of a bug report. You might be able to get in contact with some of the engine graphics programmers, they supposedly hang out in some publicly facing discord channels, but this could be spoty.
, your driver might have become corrupted so it is not processing the commands correctly (run DDU on your system to remove all nVidia graphics drivers, and AMD or intel Graphics drivers that exist) then do a fresh driver install.
The most sever it could be your GPU has a faulty memory controller, or memory chip, you might be able to spot this with running something like FurMark at as high resolution as you can for like 30-40 minutes seeing if you experience a crash with that. with Task manager open make sure the card is at high memory usage.