GPU driver crashes on flight loading [not related to Osprey]

I switched back to the v1.0 version (which crash in the same way at LPPI frequently) … and this time it worked at DRRN.

I then switched back to the v1.0.1 version … and this time it worked too at DRRN … (hmm … I did change the livery this time … not sure what impact that has)

Does DX11 vs DX12 make any difference?

The aircraft model wasn't changed, and textures are just regular DDS, texture resolution hasn't been changed.

The only thing I can think of that is different from V1.0 and computed by the GPU is VTOL physics. But again, fully stable on all beta systems (all Nvidia based)

Ok … I think I can reproduce the issue … and it is related to my custom (v1.0?) livery which I did derive from the "blank" version that I had found on flightsim.to.

v1.0.1 comes with a bundled "blank" livery now … but this should not be related to a name clash, because I did change the name of my livery.
The interesting aspect is that when I go to the "My Hangar" section … the Osprey 3D model comes up perfectly fine with my custom livery.

Only when I load into a flight … then it crashes on load … and it seems 100% reproducible.

I have no idea if developers can "bug" the sim "asset load phase" in more detail … but I can provide my livery if that might help to debug the issue … as it might affect the usage of other custom liveries (in the future)

… and just one more "idea" …

• perhaps … MSFS is loading the livery … before the Osprey aircraft is completely loaded (as it might be "stuck" in some WASM init phase)
  • that would be the "race condition"
  • and so it "crashes" because it is referencing 3D objects … which are still stuck in some other MSFS pipeline

It somehow would explain the "My hangar" anomaly

• "My hanger" maybe … first loads the 3D object
• the the livery data
• and it does not even try to compile any WASM (autopilot or flight physics) aspects of the aircraft.

Interesting. Can you please provide the livery via email to check if I can reproduce?

A naming change may just be enough. No idea what's the logic behind loading on the hangar vs the sim.

OK … I will send it to … contact@miltechsimulations.com

Correct?

Since it is 38 MB I just uploaded it to our webserver …

https://www.objectfarm.org/Public/Misc/MV-22B-Nene-nui-v20221109.zip

I basically reused the flightsim.to "Blank" and only changed the folder names and some values in the CFG files (should be easy to diff). I did not yet start changing the texture images.

Please let me know once you have downloaded it … so that I can remove it again.

I will make some tests (tomorrow) with DX11 vs DX12

(and just to reiterate … I am convinced that the "real bug" is part of the MSFS code … but only with the Osprey am I now able to reliably trigger it)

Some more background after some additional tests.

• it only seems to happen in DX12beta mode
  • … and I use 4K * 70% renderscale … TAA … with Vsync and a 30 FPS limit
  • no overclocking of my AMD Vega II GPU
    • … and the GPU runs at low average power (because the system is Main-Thread limited)
      • so the PC's PSU is only running at 50% of its peak output
• it never happens in "My Hanger" mode
• with the custom "Nene-nui" it happens 80-90% of my tests (around 10 so far)
• with bundled liveries it happens 10-20% of the time … if the "Nene-nui" livery is installed
• it also does happen with other aircraft / airports
  • e.g. the B407 at LPPI … or the A310 … both with no add-ons

So it confirms to me that …

• it is a race condition … because it is not 100% reproducible
• the primary cause is not in the Osprey
  • even when it seems very easy to trigger with my present Osprey setup