The number printed on a power supply is its sustained output — what it can deliver continuously without shutting down. It is not what your system draws, and it is not a quality rating. Two 750 W units can behave very differently under the same load.

Your actual requirement is the sum of what each part pulls at full tilt. In practice the graphics card and the processor account for the overwhelming majority; storage, fans and memory are rounding errors by comparison. This is why the honest way to size a supply is to add up the real components rather than reach for a round number.

The headroom question

Running a supply at 100% of its rating is legal but unwise. Efficiency peaks somewhere near half load, fans stay quieter with margin, and capacitors age faster when constantly stressed. Leaving meaningful headroom above your measured draw is the usual advice — enough that the unit spends its life in its comfortable band rather than at the edge of it.

Headroom also buys you your next upgrade. A graphics card swap is the most common change people make, and it is the change most likely to push a marginal supply past its limit.

Add up your parts in the power supply calculator and you get a figure based on your actual build instead of a guess.

Modern graphics cards do not draw power smoothly. They pull in extremely short bursts that can briefly reach well above their rated draw — spikes measured in microseconds, far too fast to show up in any monitoring software you have running.

A supply that is comfortable with the average load can still trip its own protection circuitry on these spikes. The symptom is distinctive and badly misleading: the machine shuts off instantly under load, usually in a game or a render, with no blue screen, no error, and no problem at all during stress tests that ramp up gradually.

Why people misdiagnose it

Because the shutdown is clean and instantaneous, it looks like overheating or a failing card. People replace the wrong part. The tell is that it only happens under sudden load changes, never during a steady synthetic burn-in.

The ATX 3.x specification exists largely to address this. Units built to it are explicitly rated to ride out excursions well above their continuous output, which is exactly the behaviour a modern card demands. If you are buying new and pairing it with a current-generation card, this is worth more than a bigger wattage number on an older design.

Connectors decide compatibility

Wattage means nothing if the cables do not fit. Check what your graphics card actually requires and confirm the supply provides it natively, from its own modular ports. Daisy-chaining a single cable across multiple connectors, or stacking adapters to reach the required plug, puts more current through one harness than it was designed for.

Use the cables that shipped with your supply. Modular cables are not interchangeable between brands or even between models from the same brand — the connector shells match while the pinouts do not, and the result of mixing them is expensive.

What the efficiency badge does and does not buy

The 80 Plus tiers describe how much wall power reaches your components rather than being lost as heat. A higher tier means less waste, less heat dumped into the room and a quieter fan — but the electricity saving alone rarely justifies a large price jump on a machine that is not running constantly.

The tier is better read as a proxy: manufacturers rarely put their best internals behind a low badge. It correlates with build quality more reliably than it pays for itself.

If the machine does run long hours, the electricity cost calculator turns that into an actual figure for your rate rather than a rule of thumb.

Safe to ignore

Cosmetic sleeving, RGB on a component sealed inside a shroud, and total rail counts on modern single-rail designs. None of these affect whether the supply does its job.