My PC started shutting off completely when I played demanding games like Cyberpunk 2077, with a Kernel-Power 41 error. After some digging, I realized my old PSU (NZXT C650, 650W) wasn’t meeting the recommended requirements for my i7-11700K + 4070 Ti Super (which is usually 750W+).
For background:
-
This PC has a history of instability. For example, I once tried running 3 RAM sticks without realizing that would cause problems, and also dabbled with overclocking. Back then it would crash constantly until I downclocked my 2× sticks to 2999 MT/s (instead of the rated 3200).
-
It has also crashed while training LLMs: it could train for a couple of hours, but would die if I opened Edge or other apps at the same time.
I bought a Corsair RM850e (850W) to fix things. At first, the PC wouldn’t even power on. After swapping back and forth to the old PSU and eventually breadboarding the system on cardboard, the PC booted and ran FurMark and Cinebench with the new PSU. I tested it multiple times successfully before putting everything back in the case.
The next day, it crashed in Cinebench. I reseated the PSU cables, and it passed both benchmarks again.
But last night, after hours of light use (VSCode connected to a server and a bunch of Edge tabs), the PC crashed again. When it does, it enters an on/off loop for several minutes, and only boots properly after I unplug, discharge (hold power button), and reconnect.
Motherboard: MSI Z590-A PRO, which has debug LEDs. The CPU LED lights up on these failures. The system usually passes POST, but with the new PSU at the very beginning, I could only reach the Windows login screen before it powered off again.
Specs:
-
CPU: Intel i7-11700K
-
RAM: 2× HyperX Fury 3200 MT/s Dual Channel (bought separately, running together)
-
GPU: Gigabyte 4070 Ti Super 16GB OC Gaming
-
Motherboard: MSI Z590-A PRO
-
Cooling: NZXT Kraken AIO (don’t recall model)
-
PSU (old): NZXT C650 (650W, 80+ Gold)
-
PSU (new): Corsair RM850e (850W, 80+ Gold, ATX 3.1)
What I’ve tried so far:
-
Tested RAM sticks individually → no difference.
-
Windows Memory Diagnostic = no errors.
-
sfc /scannow= no issues. -
HWiNFO monitoring didn’t show anything obvious, but maybe I didn’t read it correctly.
-
Temps are fine: CPU/GPU both <70°C under load. I once saw 100°C in Dragon Age: Veilguard but that was a known game bug and I fixed it by limiting CPU usage.
Behavior now:
-
I can use the PC for hours, even play Cyberpunk 2077 in 2K with DLSS and Path Tracing enabled, with no problems.
-
But the crashes are inconsistent. Sometimes benchmarks pass, sometimes they don’t. Sometimes the crash happens only after long light use.
-
The failures always involve the system fully powering off, then cycling, with the CPU debug LED lit.
At this point, I’m worried it could be the motherboard’s VRMs or CPU power delivery. I want to believe the Corsair PSU is fine, since it’s brand new. But I’m running out of ideas other than replacing the motherboard (and maybe upgrading to a 14700K while I’m at it).
Any advice on how to confirm whether this is a motherboard/CPU power delivery issue, or if the PSU could still be at fault?
