Constant hard crashes when loading into zones

Hi fellow exiles and GGG,

I've been experiencing constant (~ 10/h when switching zones a lot) hard crashes (PC just shuts down without BSOD) when loading into new zones (campaign, maps, ...).

I've already tried everything that I could think of to fix the issue:
- Verify game files
- Tried different graphics settings
- Clear cached data (steam shader cache, %AppData%/Path of Exile)
- Updated everything - OS, GPU drivers, BIOS
- Load default BIOS settings
- Undervolting CPU

I also unsuccessfully tried to reproduce the crashes using different stress test tools:
- Full load OCCT stability test (CPU + RAM + Linpack + 3D Adaptive + VRAM) to check for PSU/power draw issues
- Prime95 stress test
- 3DMark stress test

My system info:
- Very fresh Windows 11 installation (10.0.26200 Build 26200) - sadly had to switch from Linux because of Vulkan issues
- AMD Ryzen 9 7950X3D
- NVIDIA GeForce RTX 3080
- ASUS ROG STRIX X670E-E GAMING
- 64 GB RAM
- Samsung SSD 980 Pro NVMe

I have multiple sensor logs I've gathered using HWiNFO when crashes happened that I am willing to forward to GGG developers if helpful.

Anyone experiencing similar issues? If so, any ideas on how to make my experience less painful? :)

Does the Event Viewer have any relevant incidents recorded?



On the left, click on Windows Logs, and then System

On the right, click on Filter Current Log, and tick event levels Critical and Error

---

Please revert any undervolting you have performed - if anything, I would recommend overvolting.

Hey, thanks for taking time to answer my post!

I've already skimmed through the Event Viewer and couldn't find anything related to the crash reason.

The only logs that are critical or errors are:
- [Kernel-Power] telling me the last shutdown was not clean (The system has rebooted without cleanly shutting down first)
- [EventLog] The previous system shutdown at 11:49:05 PM on ‎12/‎3/‎2025 was unexpected.
- [Service Control Manager] The Windows Search service terminated unexpectedly (probably unrelated)

As suggested I've removed the undervolting. How would overvolting (overclocking) my CPU improve my stability?

The tl;dr of it is that you want the lowest possible voltage as that will be best for your temperatures (less power = less heat = less strain on your cooler). But for a CPU and a GPU both, the higher the workload, the more power they need to be able to complete it. Due to the silicon lottery - essentially, some CPUs and GPUs of a given SKU are inherently better than others - the power required for a given task will vary slightly from one CPU to the next. If a CPU's voltage is insufficient for the task at hand, the system can just crash silently in the manner you have described.

Unless the problem was your system overheating and shutting itself down to avoid permanent damage - and in such a case, I would expect some throttling rather than it going from normal to 'dead' - and the undervolting was just enough to bring the temperatures back from the brink, I wouldn't expect undervolting to ever improve system stability. Essentially, you're using the exact same hardware and giving it the exact same workload, but affording the former less power to perform the latter.

Back in the day when manual overclocking was more prevalent than it is now, you would gradually need to increase the voltage to ensure stability at higher clock speeds (and ditto for VRAM), whereas other folks would try to undervolt the GPU while maintaining stability at factory spec clock speeds.

Essentially every crash ever has been caused by an application, a driver, or the Operating System entering into a state where it doesn't know how to keep running. Dodgy RAM causes programs to receive unexpected inputs - maybe Path of Exile, when trying to calculate a hit's damage, is instructed that your Fire Resistance is "daffodil"% instead of "77"%. Unstable CPUs trip over themselves and start performing tasks out of sequence. Badly written applications try to access a section of RAM after instructing the OS to empty it because they're finished using it.

The CPU you mention has 17,840 million transistors, and the CPU uses a portion of its allocated power to rapidly switch those on and off. Essentially, the system attempts to retrieve the answer to a calculation before it has actually managed to execute the calculation. This causes the affected software to receive "bad" inputs - and once the OS becomes aware of this, it perceives it as a critical failure and immediately triggers a shutdown. This is done because an unstable system can cause physical damage to its components, and can also corrupt files necessitating a reinstall - which the OS itself is not immune to, and which is invariably a massive pain in the... Wraeclast.

After a little more tinkering I seem to have found a setup that atleast doesn't constantly crash my PC.

I've used Process Lasso to pin the process to specific cores (and without SMT) via CPU affinity as seen in the image below:



This change of course heavily reduced my performance...

EDIT: Although much less likely (played for around 6 hours without crash) sadly still crashes my PC :(