System freezing yet remains active—possible GPU issue or another cause.
System freezing yet remains active—possible GPU issue or another cause.
Recently, while watching a video, my screen abruptly froze for a brief moment before going completely dark with "No display detected." My keyboard and PC remained powered on. Even after pressing the power button, nothing changed for a short time. Eventually, I had to perform a hard shutdown. Initially, I dismissed it as a grid spike after an hour of troubleshooting, especially since my brother experienced the same issue nearby. However, two days ago, I faced 5-6 crashes in a single day. It began when I was out on a trip and locked my PC. When I returned, the monitor was off and the keyboard was disconnected—something unusual. After logging back in, it restarted but displayed an "unsafe shutdown" entry in Event Viewer. Hours later, after locking it again, it rebooted with just that error. It didn’t crash completely until right after a gaming session with a friend, when it repeatedly froze for about two minutes in Windows without doing anything. Before deciding to shut it down permanently, I checked Event Viewer and the reliability page, which showed new errors. Some notable ones included: LiveKernelEvent 141 nvlddmkm (possibly the Nvidia driver) and "Desktop Window Manager stopped working." There were also several "unsafe shutdown" related errors. This situation makes me suspect the GPU (GTX 1080Ti) might be failing, but it’s possible another component is at fault. I’m curious why the whole system would freeze completely, not just respond to a safe shutdown, especially since other signs like graphics output were absent. On my rig, the first crash happened two days after installing Nvidia's 531.68 driver, but today I performed a clean install of the latest Windows 10 22H2 and updated BIOS for my motherboard.
It has been quite warm lately. It hit 92°C for a while before I noticed how severe it became and changed the fan settings, though it still can spike to 90°C with improved clocks. None of the crashes happened during heavy gaming sessions. I played Hitman 3 briefly before the first crash, and the one before that was light enough to finish at around 60-65°C. I purchased and assembled it in June 2017. I didn’t remember trying this during a crash where the video kept playing while the screen went dark, but the audio stopped after a few seconds…
It’s clear it’s time to upgrade the graphics card. Most manufacturers don’t apply the best thermal paste, which usually lasts only a few years. Options like MX-4 or Thermal Grizzly Kryonaut are ideal choices.
Usually the thermal paste wears out quickly, so it doesn’t last very long.
It seems even top-tier cards vary a lot. Your friend’s GTX 1060 stays under 40°C at idle while you use more power, which is interesting. Regarding crashes, it might not just be the GPU—switching to iGPU helped, but keeping the 1080Ti plugged in still feels risky.
Companies will opt for the most affordable thermal paste just to meet warranty periods. The GTX 1060 generates similar heat to a 1080 Ti because fewer transistors are used and it operates more efficiently. Given the current GPU temperatures, I still think the graphics card might be the issue. If you test it on another system and the problem continues, then you’ll know it’s the cause.
I experienced a similar black screen issue with my late RX580 card. As you noticed, the only clue in Event Viewer was Kernel Event 41, indicating the PC didn’t shut down correctly—something like a forced restart. I’m aware my system uses AMD, not NVIDIA, but I was certain the fault lay with the card itself. My backup RX550 performed fine, and even my RX580 crashed under similar conditions in a test environment. In short, I considered the card dead and sold it as defective for parts. Additionally, GPU-Z shows my card operated at x8 PCIe lanes instead of the expected x16. This isn’t related to PCIe version changes (1.1 to 3.0 or 4.0) during rendering tests. If you run GPU-Z, check how many lanes it reports for your 1080ti and see if it also fails during rendering.