System freezes when GPU usage is high
System freezes when GPU usage is high
I was really hoping things would stay stable after extended play sessions (about half an hour) without any freezes and after using generative AI. The GPU usage spiked, then dropped sharply to zero and back up quickly—very intense. After these spikes, I experienced crashes in the games I played. I plan to share a Windows dump as suggested in the earlier discussion. Note: the minidump folder is empty. Edit 2: I captured the BSOD and the error message reads WHEA_UNCORRECTABLE_ERROR, suggesting a hardware problem. Could this mean my 3080 GPU is damaged? Edit 3: After a restful night, I managed to play for about 20 minutes without any crashes before exiting and revisiting the issue. It might be related to AI usage, possibly a software conflict. Maybe it's not hardware-related, but something else. Edit 4: Playing for an hour straight didn’t cause a crash, so the problem likely lies between the Python libraries I have installed. It seems to happen when AI tools are active or when launching games after generating images. That’s strange but makes sense. I might still have an NVMe drive or controller issue, so I’ll wipe everything and start fresh.
Here’s a revised version of your message:
I wanted to share all the details I collected again. Sorry for posting this multiple times.
The absence of a dump file suggests the NVMe SSD is likely the culprit, though certain configurations can block such reports. I noticed you reviewed the SMART status, but it offers little value for NVMe drives. The overall condition rating remains unreliable—it reflects total write attempts rather than real-time health. It doesn’t indicate current stability, but rather how much of the manufacturer’s warranty has been consumed. Understanding SMART details can be helpful, but with NVMe storage, those metrics have been significantly reduced. There’s now just one meaningful indicator left: Critical Warning. Among the many faulty NVMe SSDs I’ve encountered, fewer than five have displayed any SMART faults.
Method 1: Investigate the BSOD parameters to identify the root cause. Navigate to Event Viewer → Windows Logs → System, locate the shutdown event marked with a red cross (Event ID 41), and open the Details section. Copy the relevant information.
Method 2: If the system freezes during a BSOD (which you can’t capture a dump for), skip this step. However, if it restarts after a short pause, follow the bot’s instructions to disable automatic reboot. For manual restart, press the power button. To display extra details on the BSOD screen, edit the registry at HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\CrashControl, add a DWORD value named "DisplayParameters" with the value 1, and save changes. Reboot to see the updated information. This adjustment applies only to WHEA BSODs.
Found event ID 41 here. The XML file details show various system information and event data.
I went there and completed the task. No longer need to restart automatically. I’ll be able to notice the error without using my phone and playing it at 25%.
Had a crash but didn’t see the BSOD. Reinstalled ComfyUI from scratch—still no Windows reinstall needed. Worked fine, then let the PC run alone for a few hours before playing. First 10 minutes showed a black screen, I had to restart. Planning to try again now after the crash. Updated: crashed again 18 minutes into the session.
I encountered several failed attempts before the crash, suggesting possible issues with corrupted files in Windows or drivers, likely related to an NVME drive problem. The typical solution involves using DISM and the System File Checker. It appears some files were damaged, so I plan to retry loading the entire AI system to determine if crashes persist. Edit: it did crash quickly, and I now have specific error codes to analyze. Here they are: 0x0000000[...]10 (too many zeros there!) 0xFFFFAF86E60EE028 0xFFFFAF86CA163AAC 0xFFFFAF86CA2DF1A0 In the event viewer, I see no event ID 41 but only 131. Also, after checking Reddit about the "container error," it seems a failed driver update might be the cause, so reinstalling Windows could resolve many uncertainties.
This error suggests a NVMe issue with the SSD. The problem might also lie in the slot or drivers, though storage is the only driver reported by WHEA. Often the faulty drive itself is the cause. You could try re-seating it. Incorrect installation is common with M.2 devices. If the slot is near the GPU, it could be overheating. Use the SSD diagnostic tool to check for firmware updates.
Another Samsung 980... I used the "samsung magician" tool recently and was running the latest firmware. Yes, the drive is next to the GPU, which might cause overheating, but it didn’t happen sooner. After backing up everything—so I won’t lose data and can reinstall Windows 10—I’ll open the case, clear dust, reinsert the drive, and try again. Thanks for your help! I didn’t say much but I’m really glad I can access the knowledge you and others have.
Nice! I recently moved the PC a few days ago and again recently to recalibrate everything, including installing arms for my two ultrawides. It seems I might have bumped the PC a bit, but I managed to reinstall the drive. So far, I can play about 45 minutes without any crashes and exit games normally. I'm now pushing the GPU to its limit by creating images with AI locally, and it seems to work. We'll see how things progress over the next few days. Edit: I generated around 50 images in a row—previously it was impossible before the drive was replaced, so maybe that finally helped.