Unpredictable Vulkan crashes with vk::Device::waitForFences: ErrorDeviceLost

Hello.

I play for a week now, with a new computer, on 2 characters, without any problem.
This morning, i start an old character and i don't plug the power supply, boom, that crash happens after 5 seconds in an area where the skills are loaded.
I plug on the power supply... no crash anymore.

This is definitely linked to hardware and power needs of the characters you play.

The Path they take to fix this problem is wait for player with problem to Exile themself out of this game.

I am having this problem. It seems this has never really been fixed, or is it just when a new league start and there are shaders that are not handled properly.
There was one map where I crashed 5 times in the exact same spot but the rest of the map was running fine, but when I got near that one spot, the computer slowed right down and got choppy losing frames and then would crash.

Hi,

i've managed to fix this on my site, not sure if you are using Automatic Tuning for soft overclocking your hardware with GeForce Experience, but turning it off fixed the crashes for me.

Not only for PoE but also for other titles that were working on Unreal Engine for example.

What auto tuning does is it tests couple of boost/voltage profiles and chooses the one with highest score that passed the short test, next it dynamicly boosts depending on the temperature/power limits.

It does not test workloads typical for PoE 2/UE5/Vulkan spikes.
Tests are short, linear and predictible.
PoE 2 is chaotic and does micro-spikes which avoids short GeForce tests.

So in short, OC can be " adaptive " but can also fail on a single specific shader.

Auto Tuning is good for benchmarks and AAA games with equal load, but not for games like PoE and shader-heavy types.

Even overclocking from a vendor can be unstable

"

Applications/Games crashing to desktop or causing PC to BSOD
1. Are you overclocking your GPU or using a factory overclocked graphics card? If so you can configure your graphics card to run at NVIDIA reference clock speeds by enabling Debug Mode. To enable Debug Mode, open the NVIDIA Control Panel (right-click over your Windows desktop to locate the NVIDIA Control Panel context menu option) -> click on Help from the top menu -> click on Debug Mode. Next see if you are still able to reproduce the application crash.

https://nvidia.custhelp.com/app/answers/detail/a_id/3141/~/how-to-provide-valuable-feedback-to-nvidia