I'm preparing to build a new PC and would like to pass my current system to my brother-in-law, but it unexpectedly shuts down during demanding games and benchmarks. The system uses a B550M Aorus Elite motherboard, Ryzen 5 5600 processor, 32GB of Corsair Vengeance DDR4-3200 CL16 memory, a Gigabyte RTX 3070, and a Corsair CX650 power supply. Reducing the GPU power limit in Afterburner makes the shutdowns much less frequent, although they can still happen occasionally at reduced power. I'm planning to disassemble, clean, and rebuild the computer in a different case with fresh thermal paste and new fans. What is the best step-by-step way to identify whether the problem is the power supply, graphics card, temperatures, or something else?
3 Answers
I’d start with the power supply, especially since lowering the GPU power limit reduces the problem. A sudden loss of power under a combined CPU and GPU load often points to PSU protection triggering, a failing PSU, a loose power connection, or a problem with the GPU’s power delivery. Make sure the graphics card uses separate PCIe power cables rather than one daisy-chained cable, and reseat the 24-pin motherboard, CPU EPS, and GPU connectors. The CX650 may be enough on paper, but an aging or faulty unit can still fail during short power spikes. The most reliable PSU test is swapping in a known-good, suitably rated unit; software cannot accurately verify PSU health. You can also test CPU and GPU separately, then together, to see which load causes the shutdown.
A GPU temperature of 83–84°C is warm but generally within the normal operating range for an RTX 3070 and would usually lead to throttling rather than an immediate complete power loss. Check the hotspot temperature as well, since the core temperature alone may not show the whole picture. Clean the heatsink, confirm the fans are working, improve case airflow, and monitor temperatures with logging enabled. If the card is reaching its thermal limit, reducing the power limit or applying a mild undervolt can help, but that should be treated as a workaround until the underlying shutdown is diagnosed.
Troubleshoot methodically instead of changing several parts at once. First restore stock settings, disable overclocks and memory profiles temporarily, and check Windows Event Viewer for unexpected power-loss events. Run a CPU-only test, a GPU-only test, and then a combined test while watching temperatures, clocks, and power. If only the combined test shuts the system off, the PSU or power cabling becomes much more likely. If the GPU-only test causes it, inspect the card, its connectors, and its temperatures. If the system merely freezes or restarts instead of losing power, also test the RAM and update the motherboard BIOS and chipset drivers. Rebuilding the machine and replacing thermal paste may improve cooling, but it won’t fix a defective PSU or graphics card.
The shutdowns happen during demanding games and benchmarks. I saw roughly 400W of combined CPU and GPU draw, and the GPU reached about 83–84°C. How should I test the PSU, and are those temperatures high enough to cause the shutdown?