I’ve put this on the deadline forums as well, but figured I might get some assistance from the UE forums as well since now the error is specifically a UE crash, the job seems to successfully find and attempt to load the scene (which renders locally without issue) on the same computer that has picked the job up on the farm.
Here is a snippet of the deadline crash log, if any knows what this specific Fatal error is that would help a lot, I’m not sure what DXGI ERROR means? or what it failed to build/load?
The scene in question is extremely simple just a default scene, and a sequence created from “Main” level nothing complex at all, and renders on this same machine locally no issues, with the same MRQ settings that are used to submit this job. So all I can think is something is going wrong with a headless render, or something is wrong on the deadline side of things. Not sure.
There is also this warning/error a fair bit before this fatal crash: 2025-07-02 12:34:11: 0: STDOUT: [2025.07.02-19.34.10:946][ 0]LogCUDA: Display: Initialising CUDA API…
2025-07-02 12:34:11: 0: STDOUT: [2025.07.02-19.34.10:946][ 0]LogCUDA: Display: CUDA API initialised successfully.
2025-07-02 12:34:11: 0: STDOUT: [2025.07.02-19.34.10:946][ 0]LogCUDA: Display: Found 1 CUDA capable devices.
2025-07-02 12:34:11: 0: STDOUT: [2025.07.02-19.34.10:946][ 0]LogCUDA: Display: Found device 0 called “NVIDIA GeForce RTX 4090”.
2025-07-02 12:34:11: 0: STDOUT: [2025.07.02-19.34.10:946][ 0]LogCUDA: Error: CUDA module failed to create a CUDA context on the RHI selected device.
first error it says CUDA module failed, this is on a 4090 and latest studio drivers, so not sure why this failure?
but im not sure why a CUDA module would fail, is the CUDA context Error ignorable? but then the CUDA api is initialized so should this just be ignored?
Let me know if you guys need more info