Signs of a Failing Power Supply and How I Test One at My Somerville, NJ Shop

With the supply unplugged from the wall and every component, bridge the green PS_ON wire on the 24-pin connector to a black ground pin beside it, plug in, and flip the rear switch. If the fan spins, the supply can turn on. Corsair's support guide describes this procedure and warns that jumping the wrong pins "may result in injury and damages to the PSU," so count twice. Corsair also notes some models have a zero-RPM fan mode where the fan spins briefly and stops, which still counts as a pass.

A pass means the standby circuit and PS_ON logic work and the rails come up with zero load. It doesn't tell you the 12V rail holds when a graphics card asks for 400 watts in a millisecond, and it doesn't tell you the output is clean.

The plug-in PSU tester is one step better. It shows each rail under a small fixed load and reports the Power Good delay; Corsair's guide gives the acceptable ranges and says a Power Good reading of 0 ms means a failing unit. It catches a rail that's grossly out of range or missing. It won't catch the failure I see most: a supply that reads fine at a few watts and falls apart at several hundred.

A well-built supply is designed to fail safe. Seasonic's support documentation lists the protections in its units: over-current and under-voltage protection on each rail, over-voltage protection that shuts the supply down if a rail climbs past its trigger point, over-power protection, over-temperature sensors, and short-circuit protection. That's why a quality unit that fails usually just refuses to turn on, and the rest of the PC survives.

A budget unit, or an old one whose components have drifted, may lack some of those protections or have them set so loosely they never trip. That's the supply that pushes 13V down the 12V rail, or passes a surge from the wall straight through. I've replaced motherboards and graphics cards where the original fault was a supply that had been buzzing for a year. My page on power surge damage covers why the supply is often the first casualty.

Age and warranty realities

Warranty length is the honest way to judge the class of unit you own. Corsair covers its AXi, HXi, RMi and RMx lines for ten years and its other lines for seven, per its own announcement, and Tom's Hardware's current buying guide lists most of its recommended units at ten years, with the longest at twelve. A supply with a two-year warranty shows up on my bench at about the age its warranty implied.

A quality unit still inside its warranty is worth testing, because a confirmed failure gets you a free replacement from the maker. Past warranty, from a budget line, or already responsible for a dead part: skip the test and replace it. The cost is quoted after the diagnostic and is almost always less than the parts it protects.

ATX 3.x and the 12V-2x6 connector, in plain language

Intel's ATX 3.x design guide is the current desktop standard. The big change: the supply must ride out short power spikes from modern graphics cards, up to twice its rating for a fraction of a millisecond, without shutting down, which is why a GPU upgrade so often exposes an older supply. It pairs with the 12V-2x6 connector, the 12-pin plus 4 sense-pin plug that feeds high-end cards up to 600 watts on one cable. Tom's Hardware notes it uses shorter sense pins than the earlier 12VHPWR plug, so the card won't draw full power until the connector is fully seated, a direct answer to the melted connector problems of the first generation.

If your card uses that plug, or you plan to upgrade to one, the replacement should be a native ATX 3.x unit. With 8-pin cards it's the better buy but not mandatory. When the swap is part of a larger upgrade, I fold it into my power supply upgrade service so wattage and connectors match what's going in. For a supply that has simply failed, including the brand-specific units in many Dell, HP and Lenovo towers, see my power supply replacement service.

A Bridgewater customer's gaming tower that restarted mid-match. Fine on the desktop, rebooted within minutes in anything demanding. The graphics card had been upgraded months earlier and the original 600W supply from 2018 was still in the case. Under load the 12V rail dropped to 11.2V and the unit shut itself down. Replaced with an ATX 3.1 unit sized for the card, re-tested under load, no more restarts.

A Hillsborough customer's office desktop that wouldn't turn on. No fans, no lights. The paperclip test passed. With just the board and CPU attached it started fine; the fault was a shorted front-panel USB header. The customer had been ready to replace a power supply that didn't need it.

A Franklin Township customer's tower with a buzz and USB dropouts. The supply passed a plug-in tester with every rail in range, but the 5V rail showed heavy ripple under load and the fan bearing was gone. New supply, every symptom gone.