I have a project using Unity 6. About half the time when I start the Editor to open this project, it BSOD’s in exactly the same spot (around 40%) loaded once it hits the “Running backend” phase. The error code is PFN_LIST_CORRUPT. This causes my machine to reboot, hang during Windows startup before the login page is displayed, then a hard power off/on allows Windows to load properly and gets me back to my desktop.
This is a brand-new Dell workstation provided by corporate IT. And to be clear, I’ve used this desktop 40+ hours a week since the beginning of November. It only crashes in this manner when starting Unity 6 and (so far) only on this project. It has never crashed loading a project in Unity 2022 or using any other app. The fact that it’s both intermittent and consistent is extremely puzzling.
Solutions already tried:
Full SMART report of SSD (no errors)
Update Unity 6 (0.26 → 0.32)
Wipe and rebuild project from clean git checkout (only Assets, Packages, and ProjectSettings folders are stored)
Delete cached Unity libraries and assets
Examined the Windows Event log. The best I can get is a report saying the machine restarted after an unexpected shutdown with no indicator of reason.
The box is running Windows 11 23H2 and has all current updates.
Update your drivers, make sure to use “stable” drivers only and for GPU prefer the Studio/Pro drivers for creative work, rather than the “game ready” drivers.
If the issue persists it is most likely a system issue, not related to Unity. Unity respectively your project just taxes the system in a way that it fails. You may want to run a full system test program from a boot disk, perhaps Dell even provides one. Alternatively, let Prime95 and Furmark run together for 1-2 hours at “max heat” settings to verify the system is otherwise stable, but if not, you should contact Dell for possible repair/replacement.
Prime + Furmark is always the first things I burn-in any new machine and so far it’s proven itself as a guarantor for a stable system.
If it’s an Intel i7/i9 system you may need a BIOS update to fix Intel’s CPU destroying bug due to excess heat caused by “performance optimized” (beyond specs) BIOS settings. My i9-14900K had stability issues until a BIOS update was made available end of last year, and this also prevents CPU degradation to the point where it may eventually fail altogether.
I had the studio drivers from mid-November already installed. However there was a mid-December update so I took your advice and did the update. I also ran Dell’s built-in system tests that run from BIOS and it detected no problems, but I did take the opportunity to update the BIOS and system drivers as well.
Afterwards I was able to start up my Unity project a dozen times in a row without a crash. So I’m cautiously optimistic this fixed the issue.
Thanks!
I was able to download Prime95 but for some reason corporate security has blocked Furmark downloads whether in executable or zip format. Not sure why. Are there good alternatives to that?
Great to hear! Could you share the Dell model number or CPU? I like to know if it was one of the affected Intel CPUs. I remember one previous instance (besides mine) where the BIOS update did fix similar instabilities.
Hmmm 3DMark comes to mind, and maybe Passmark. Not sure if they have “burn in” features which is what you would want, but even running consecutive high-end graphics tests would be okay. The thing is though that Furmark runs entirely on the GPU and thus you can cause it to heat throttle within seconds, while Prime does that at the same time to the CPU. With other tests like 3D Mark they also utilize the CPU and if the CPU is taxed, they may not tax the GPU as much.
So I would run 3D Mark without Prime in the background and loop that several times.
Corporate IT … if it costs 10 times as much it’s gotta be at least five times faster, right?
I used to work on a project where we had single-threaded Unity interact with single-threaded 30 Hz Kinect and for each Kinet, they got us a Xeon workstation with dual processors (!), each with a dozen physical cores. In fact, we got ten of these. No, make that twelve - two serving as backups. Times three, because we ran the software in three different rooms. And each room was handled by a monstrous Xeon server (dual 32 cores) with a Titan GPU for the central orchestration application … which was almost entirely single-threaded.
Well, at least I got to shave off a few ms in our image processing because it was easy to multithread and we could afford to compress the image data before sending it over the network to the central server.
But regardless, BSOD would indicate the system isn’t stable and sounds more like a ticket for the IT team.
