I recently built a PC with a Ryzen 7 7800X3D, MSI PRO B650M-A WiFi motherboard, XFX RX 9070 XT, MSI MAG A850GL 850W PSU, 32GB of DDR5-6000 memory, a Thermalright Phantom Spirit 120 SE cooler, and a 2TB PCIe 4.0 NVMe SSD in a Lian Li DAN A3 case.
The system intermittently powers off completely without a blue screen. After some shutdowns, the motherboard's red CPU EZ Debug LED remains lit during the next startup, while other times the computer boots normally. It has even shut down once during a normal boot, and repeated unexpected power losses caused Windows to enter Automatic Repair.
BIOS defaults are loaded, EXPO and PBO are disabled, and the computer has been tested directly from a wall outlet. CPU temperatures appear normal. OCCT's CPU-only and GPU-only tests both complete successfully, but the combined Power test causes an immediate hard shutdown. This makes me suspect the PSU's protection circuitry, an intermittent EPS or GPU power connection, or a motherboard power-delivery problem.
I plan to disconnect and firmly reseat the 24-pin motherboard cable, CPU EPS cable, and GPU power cables at both ends, then clear CMOS. Is this behavior most consistent with a failing PSU, and what other checks should I perform before requesting an RMA?
3 Answers
It is still worth running a longer CPU-focused test such as Prime95, but passing separate CPU and GPU tests does not rule out a PSU problem because the combined load stresses the PSU and motherboard power delivery differently. A sudden loss of power with no crash screen is consistent with protection tripping or an intermittent connection, rather than a typical driver failure.
Reseating the cables and reinstalling drivers may make the problem disappear temporarily, but the fact that it returned during a game and the red CPU light came back means the underlying fault probably remains. After reseating and clearing CMOS, test at stock settings. If it still shuts off under combined load, swap in another suitable PSU if you can; if the replacement behaves the same, the motherboard, CPU power connection, or less commonly the CPU itself becomes more suspect.
A combined CPU-and-GPU test putting the system instantly into a hard power-off points more toward the power-delivery path than a normal software crash. Carefully unplug and reseat every modular PSU cable at both the PSU and component ends, especially the CPU EPS and GPU connectors, then clear CMOS. Also inspect for partially backed-out pins or cables being sharply bent against the case. If possible, test with a known-good PSU; that is the quickest way to separate a PSU fault from a motherboard or wiring issue.
I do not currently have another PSU available, so I will start with a complete cable reseat, a CMOS reset, and stock settings before deciding which component to replace or RMA.