How much PSU headroom does your GPU really need?

The power supply is the least exciting part of a build and the one most likely to cause baffling crashes when you get it wrong. With GPU power draw climbing every generation, sizing it sensibly matters more than it used to. Here is the practical version.

Why headroom matters

Running a PSU near its maximum all the time is inefficient, runs hotter, and ages it faster. More importantly, modern GPUs draw brief power spikes (transients) well above their rated TDP — for a few milliseconds a card can demand far more than its sticker number. A PSU with no headroom can trip and shut down even though the "average" math looked fine.

The rule of thumb

Add up your components' draw and target a PSU where your full-system load sits around 60-70% of its rated wattage. That gives you efficiency in the sweet spot plus enough margin to absorb transient spikes.

The numbers are rising

This is the part that catches upgraders out. GPU TDPs have climbed sharply — we measured an ~92% jump in average board power over a decade in our power and efficiency report. A flagship like the RTX 5090 pulls 575W on its own. A PSU that was comfortable for a previous-gen build can be undersized for a new flagship, even if your other parts didn't change.

Practical guidance

• Mid-range GPU: a quality 650-750W unit is usually plenty.
• High-end GPU: 850W is a safe baseline.
• Flagship (500W+ cards): 1000W and up, and don't cheap out on the unit.
• Always check the GPU's TDP on its spec page and the maker's PSU recommendation, then add headroom.

Buy a good unit from a reputable line, size it with margin, and you never think about it again. Skimp, and you will spend a weekend chasing crashes that look like everything except the real cause.