It was a Friday night, I had just downloaded the latest patch to the game that I was grinding away at the time, queued up for a ranked game, and then, four minutes into the game… bam. Black screen. Then the Windows logo.
My PC had just rebooted like nothing had happened, no warnings, no errors, no blue screen, nada. I just sat there for a second and looked at the boot screen and wondered “did that just really happen?”
Chances are if you are reading this you know the feeling. It’s not a crash. Crashes give you something a freeze, a blue screen of death with a stop code, a game that just closes itself, your desktop returning you to the desktop.
That is not what is happening. What is happening is that your whole PC has reached around, grabbed itself by the neck, and then pulled the cord out.
It took me about three weeks to go down that rabbit hole and actually fix it. And I want to take you through what I did specifically to diagnose the problem (not just give you a random “go check your drivers” list… because that’s not what actually fixed it for me).
How I First Noticed the Pattern
The first time it happened, I blamed it on the game. New patch, buggy build, whatever the default fallback. But then it happened again two days later, and I was just looking at Chrome with maybe fifteen tabs, no play, no load, nothing to really make it work. That’s when I noticed it.
In the following week I began to mentally keep a running tally (then, after the recording stopped, an actual tally in Notepad because my brain was tracking to slow) of when it occurred:
- Gamified intervals On two occasions within the playing session, always after 15–30 min continuous play.
- Just sat on your desktop with YouTube in the background idle when fully idle.
- Once while I was exporting a video in DaVinci Resolve.
And that idle happening is what really, really surprised me. If it was just a power or heat draw thing based on gaming load, then why was it happening while I was doing basically nothing? That mismatch is actually a good sign, and I nearly didn’t notice it because I was so fixated on “probably my GPU.”
Initially, What I Thought (And Why I Was Off)
I initially suspected a GPU driver issue… I’d recently updated my Nvidia drivers (20.10.17.3283) and they’d started to crash at roughly the same time that the reboots did. So I used DDU in safe mode to roll back and reinstall some older driver versions and was content.
It “fixed” it for a four days. Then it happened again. Classic case of correlation is not causation, the timing was right so I thought this is it, but it was not actually the root cause. My second suspicion was heat. I have a mid tower case and it’s got average airflow, so I fired up HWMonitor and let it do its thing for a couple of days. Temps are ok, though.
GPU never over 72C while gaming, CPU 65-70C. Nothing special, but definitely not close to heat shutdown, so again, I checked that off the list.
The Actual Diagnostic Process That Cracked It
This is the part that most articles gloss over and I want to be clear about what I really did, in order.
Step 1: Checking Event Viewer
Honestly, this is the first step for everyone that I can think of, and I wish I had done it first instead of trial-and-error. And so:
- Hold down the Windows key, type “Event Viewer”, run it.
- In left panel, expand Windows Logs > System.
- Keep an eye out for the red “Error” entries at the time a few seconds before your last reboot occurred.
- Keep an eye out for Event ID 41 (Kernel-Power) and Event ID 6008.
In my case, I just happened to have a bunch of Event ID 41 entries, “The system has rebooted without cleanly shutting down first.” That confirmed that something was killing power to the system suddenly. This isn’t a software crash, this is the power going away suddenly. That one entry is what saved my bacon.
A software crash (a bad driver, corrupt game file, Windows bug) has a high chance of leaving a BSOD or crash dump in the logs. An unclean power-down with no crash dump is almost certainly a hardware problem, in fact in many cases it points to a bad power supply unit.
Step 2: Testing RAM with MemTest86
This is a definite issue that bad RAM can lead to with random reboot – It is certainly possible if you’ve ever turned on XMP/DOCP profiles by hand (which I did to my B series of motherboard months ago, and honestly had forgotten about it).
Install MemTest86 (and use the free one from their official website, not the ancient version that was included with some Windows disk tools), created a bootable usb stick with their utility, and let it run overnight, 8 passes.
No errors. My RAM was clean. This turned out to be good info even if it wasn’t the fix, because it eliminated one entire class of potential causes. I could stop blaming RAM after that. You can also read:
- What Happens to a Game You Bought When the Servers Shut Down
- Steam vs Epic Games Store: Which Platform Is Better?
Step 3: Comparing PSU wattage and GPU Power Draw
This was the step that finally cracked the case for me, and it’s also the step that everyone seems to ignore because “checking your PSU” feels like “more advanced” troubleshooting.
Previously, I had a 550W PSU. My GPU (an RTX 3070 at the time) has a rated TDP of 220W, however transient power surges on Nvidia 30-series and 40-series cards can spike well above the rated TDP for short periods of time, sometimes 1.5 to 2x the rated draw for a brief instant. Throw in a mid-range CPU drawing 100-150W under load, plus drives, fans, RGB, and a motherboard, and that short power spike probably pushed me right up against, or just above, what my PSU was delivering reliably, especially given it was a budget-tier non-Gold-rated unit that didn’t handle transient loads all that well.
This is how I really managed to verify it for you:
- The gauge on GPU-Z shows you the real rated TDP of my card, and if you look at the sensor tab, you can see the power draw during a stress test.
- I performed a combined stress testing using OCCT (it’s a dedicated power supply testing mode that is made to induce transient load spikes on purpose).
- You got it – roughly 90 seconds into the OCCT power test, my PC crashes and reboots – no BSOD, the same as in the real world.
That reproduced was my proof. I was no longer just guessing. I had duplicated the failure on demand.
The Fix
I replaced my 550W budget non-modular power supply with a 750W 80+ Gold PSU by a reputable brand (I’ve used Corsair and they were good, but I hear Seasonic and EVGA are good as well). Since I’ve done this I haven’t had a single random reboot in 4 months, during long gaming sessions and my render work.
What I Would Do Differently If I Did This Again Mistakes I’d Avoid
Some advice I’d tell myself at 18 to save some pain:
Don’t immediately go to reinstall Windows. I was about ready to go down that route out of frustration. It wouldn’t have fixed the hardware power problem, but would have cost me time and preferences for nothing.
Don’t ignore Event Viewer. I wrote it off as a last resort step, when really that should’ve been my first stop. Event ID 41 is basically the innermost circle of the crash … it tells you up-front if you’ve got a power/hardware problem versus a crash.
Assuming a “quiet” PSU fan indicates the PSU is healthy. Mine was never odd, never smelt strangely, nothing. Silent failure when suddenly loaded is common with budget units and newer GPUs.
Not just stress testing with something like a Cinebench or a game. Transient power spikes aren’t always flagged like they are by a proper PSU stress testing utility.
Final Thoughts
In cases where your PC is restarting with no Blue Screen of Death, that fact that you’re not getting one actually has a message behind it: The system lost power or experienced a hard failure, rather than something on the software side caused an error that could be logged. But check the Event Log, test your RAM with MemTest86, and test if your PSU can really handle the impact of your GPU’s actual power spikes in the real world, not its ‘rated’ TDP in the datasheet.
It’s a little slow.. Somewhat like figuring out what isn’t broken and swapping it with the new, but it’s the only way I could actually find the root cause without throwing money at the wrong thing.
