Addressing ongoing BSOD with Event 41 and no available dump file
Addressing ongoing BSOD with Event 41 and no available dump file
Another detail I overlooked that might be connected and bothers me is how often my PC automatically turns off the screen to conserve energy. Sometimes when something like Spotify is playing, it keeps running, but I can’t wake it back up except by pressing the power button until it shuts down and restarts.
The results might not reflect the whole situation. Memtest86 could be showing a partial outcome, so it's worth checking other factors.
Memtest isn't overly aggressive on RAM. Consider using TM5 with Anta Extreme settings for stronger testing. For stability checks, y-cruncher works well—it increases CPU heat so you can observe RAM behavior under load. Running prime95 will stress the GPU as well, giving a realistic view. If both tools perform consistently, your system is stable.
Running a memory tester within an operating system wouldn't be advisable since it can't access memory that's currently being used. If a bad sector is active, the tester can't locate it. Operating systems already consume a significant amount of memory, and these utilities are designed for the CPU, not the GPU.
I heavily upgrade my RAM, so I’m familiar with how to check for issues. I use TM5 Anta Extreme, OCCT, y-crunch and memtest occasionally for practice. Memtest often misses problems. Yes, I usually test only 48 or 64GB of RAM, which is why I run several cycles to ensure full coverage. A few 3-5 cycles with TM5 should spot errors. Then I switch to y-crunch to verify its stability. Reading an OC guide will give you basic details. This process requires a lot of troubleshooting, especially when pushing RAM limits. TM5 is mainly a RAM tester, while y-crunch also checks the CPU. The only thing it might miss is if you set the tREFI to maximum. But a screen freeze caused by poor RAM retention is easy to spot.
Yesterday I managed to get it running smoothly on two sticks by omitting the side panel. It might have been a coincidence or could indicate thermal problems. Recently, after reinserting the side panel, it crashed after about 15 minutes, but this time the failure was much more gradual. The system reported a WHEA error upon restart and generated a dump file, confirming it hadn't done so before (as that directory didn’t exist and only one minidump was present). Here’s the link to all dump files in my OneDrive: https://1drv.ms/f/c/30eefcb1fb616e57/Ekd...g?e=qMXcUu This issue also appeared in hwinfo64.
The manually gathered dump file was devoid of data, yet perfmon detected it. It revealed a WHEA_Uncorrectable_Error, indicating a hardware fault with the CPU or a PCIe component. The CPU tracks its own and PCIe device errors and transmits them to Windows when issues arise. In your case, it highlighted a problem reading from the L1 cache. Your event logs displayed six such uncorrectable errors on the 24th, though all parameters were missing, suggesting a sudden shutdown. These parameters hold the error packet details, clarifying the cause of the crash. You also noticed numerous WHEA events. The WHEA logs resemble CPU activity, typically generated when the CPU detects a critical issue it can't handle, prompting an immediate reboot and sending the error packet at startup—information Windows uses to log the event. Your logs also recorded several shutdowns and restarts during that time. Decoding these manually is challenging due to the size of the error packet. Usually, we use specialized tools from a tech support group that parse events and decode packets for us. Compared to WHEA BSODs, error packets from crashes are much smaller and simpler. BSODs rely on AMD/Intel docs, while event logs follow UEFI guidelines. I’m open to the chance that my interpretation of the WHEA data might be off. A failing CPU is likely the main issue, but since removing the side panel stabilized it, overheating could be the real cause. If you’re adjusting overclocks or undervolts, take them out. Your CPU temperatures appeared normal in the screenshot, yet idle temps can be misleading. Even if they’re okay during load, they might not reflect stability when idle. If your CPU stays cool under stress, a VRM (Voltage Regulator Module) issue is possible—especially since there are no temperature sensors for the VRMs. The 5000 series models often face voltage problems; we can experiment with small adjustments. One option is to enable a voltage offset setting on your motherboard: set the CPU Core and SoC offsets to +0.050v (read this twice—0.05v, not 0.5v). Another approach is to apply a static voltage: 1.3v for the Core and 1.1v for the SoC. If your board uses incremental settings, try matching them closely. You can’t use both at once, so test one at a time. The first setting is more broadly applicable for 5000 series chips experiencing memory controller errors (which I’m less confident about diagnosing from event logs). The second adjustment has proven effective for many high-end models like the 5800x, 5900x, and 5950x across various crash scenarios.
Thanks a lot for your time so far! It seems the side panel isn't an issue now. I've tested it crashing with both sticks in single and regular dual channel without it. The last time it triggered a "hypervisor-error" during BSOD. I've shared the sysnativefilecollectionapp result from the previous link, marked as #2. I plan to try undervolting this evening when I return. Appreciate your support!
I already sense this will reveal an NMI before I even notice it. NMI stands for Non-Maskable Interrupt, a special interrupt that forces the CPU to pause and address it right away. It bypasses normal command queues that would otherwise wait. This kind of issue usually points to hardware faults. Most consumer devices trigger an NMI when the CPU is under stress, though we can't determine the exact cause or origin. Reviewing the dump file confirms an NMI was generated. To be precise, this isn't an undervolt; it's more like an overvolt affecting the SoC (system on chip). The voltage spike would be minor, possibly raising temperatures slightly, but it shouldn’t be significant.