Hardware Monitors in use at the same time as the built in monitoring also would trigger display driver resets, the Health Monitoring Chipsets on most of these GPU’s are designed to only respond to so many requests per/second.
App would send a Request for Info from the Chip, and if it didn’t respond within a threshold it would try to send a reset code to the chip or resend the same Request causing a Request Polling pile up in the chipset, which if the GPU is in use, and the chip is already reporting data to the nVidia/AMD Driver, it will trigger the driver to crash and windows to recover the display driver state.
I was able to trigger it 100% by running any Monitoring Software and opening the AMD Driver Window that shows health status of GPU, then telling the monitoring software to reduce polling time, and which case, after a few minutes the Chip that responds to those requests, will trigger something in the GPU, and all displays will cut off, windows will come back with a GPU Driver recovery bubble.
I can also force this on the Windows 10 Desktop just by increasing polling rate by 10x.
Likely because the AMD App was requesting info at a certain polling rate, and then a separate program was requesting info at certain polling rate, and the chip likely see’s them as double polling and triggers a GPU reset, or the chipset simply can’t handle that many requests, and even less under load temps, and has a meltdown and sends a reset code to GPU.
plus those Hardware Spec chips that control those readouts aren’t exactly under any cooling, so increased polling along with increased heat from the GPU and Ram, would likely overheat that chip.
I basically shut off all the extra crap while gaming, even the 3rd Party MSi Tool, and just use AMD’s Built In Health Monitoring, and their phone app to see it outside the game window.
I would hope nVidia has something similar.