@kisak-valve Would it not make sense to have specific issues for specific errors? I can imagine that this issue, a full system crash, has a different cause then an issue related to rendering or controller support? You have a soul crushing pile of issues to manage so I can understand you can't make to many distinctions, but you might want to consider an ''impact' scale.
Hello @Eonfge, if you have detailed, specific feedback that affects multiple games, or issues affecting whitelisted games, we can evaluate that as a separate issue, but beyond that we're using one issue report for all feedback regarding each unofficially supported game title and you can add a note here for all quirks you're seeing with this game demo.
I'll keep that in mind for the future.
This is not actually the first time that I have a full system crash with proton 6.3 but I haven't been able to see a pattern yet. It might be something performance related, although that behaviour is inconsistent: I have games that struggle to run but that never crash, and I have games (like Warlock now) that have system wide stability issues. If I can find another example, I'll make a generic issue.
It should be noted that the symptoms you've described, a system restart without a dying message usually hints at a general hardware (thermal? power? borderline ram timings?) or kernel issue, but modern computers are complicated and I'm not going to be able with troubleshooting to that degree. When there's a software-induced system-wide failure, the kernel usually has a chance to write a dying message to the system log or to netconsole if you set that up and that can be pondered, if not, then the odds of bad hardware increases.
Honestly, I think that there might be a bug in the NVidia drivers that is only triggered through certain conditions using certain VK extensions.
Thermals, power and similar I rule out, since I don't have such crashes running other applications, even if my system is struggling. The power draw should be similar for different applications if the GPU is at 100%. I also kept an eye on the thermals and % usage using Mangohud, and nothing was out of the ordinary.
Memory is also unlikely. I've had a dead memory stick before (10 years ago now) and its crashes were far less predictable. If I can start one specific application with one set of graphics settings, so it causes a crash within 5 minutes... that's way to predictable. The Kernel should be randomly assigning memory for security reasons, so if they give exactly the same piece of memory to Warlock every time, then I have something to complain about on the Kernel mailing list.
Lust but not least, the Systemd logs are empty. There is the usual log-spam but nothing that gives me a clue. I'll might be able to set a netconsole logger up in the future.
Thus, for now I think it might be driver related issue. NVidia 4.65 is currently in testing so I'll make sure to give it another go when that's released for Fedora 33. Until then, let's just see if somebody else has similar issues.
proton 6.3x1 2021-04
Compatibility Report
System Information
I confirm:
Log
steam-907740.log
Symptoms
Full system crash. The screen literally goes black and the machine power cycles.
Reproduction
1b. Make sure that the system is heavily taxed: The crash occurs when I run the game at 'high quality' 1440p uncapped, but when I go down to 1080p capped, the game remains stable for the duration of the Demo.