Modern graphics cards do something that trips up otherwise sensible PSU choices: they draw brief, enormous bursts of power that last only a fraction of a second. These transient spikes can be double a card’s average draw, and a PSU sized only to the average can shut down when they hit. Headroom is the buffer that keeps your system stable.
As an Amazon Associate we earn from qualifying purchases at no extra cost to you.
Our Top Picks
What a transient spike is
A transient, or power excursion, is a very short surge in current, often lasting under a millisecond, as a GPU rapidly shifts load. A card averaging 300 W can momentarily demand 500–600 W during these microsecond bursts. The average is what your electricity meter sees; the transient is what your PSU’s protection circuits react to.
| GPU average draw | Possible transient peak | Suggested PSU |
|---|---|---|
| 200 W | ~350 W | 650-750 W |
| 300 W | ~500 W | 850 W |
| 400 W | ~650 W | 1000 W |
| 450 W | ~750 W | 1000-1200 W |
Why it can shut down a PC
Over-power protection reacts
A PSU includes over-power and over-current protection that trips if load exceeds a threshold, cutting output to protect the system. A tight PSU can interpret a legitimate transient spike as a fault and shut the whole PC off, producing the frustrating black-screen-mid-game reboot that no error log explains. More headroom stops the trip.
Headroom absorbs the burst
Sizing the PSU well above average draw gives the unit slack to ride out spikes without tripping. A 300 W-average GPU in a system pulling 450 W on average pairs best with an 850 W unit, leaving room for a 500–600 W GPU transient plus the rest of the system peaking together.
ATX 3.0 and 3.1 help
Newer ATX 3.0 and 3.1 power supplies are explicitly designed to tolerate transients, rated to handle spikes up to double their wattage for short windows. If you run a high-end GPU, an ATX 3.x unit can offer more stability at a given wattage than an older design, though generous headroom still matters.
Diagnosing a spike-related shutdown
When a PC shuts off instantly under load with no blue screen and no crash log, transient-driven PSU protection is a prime suspect, but it is worth ruling out neighbors. Overheating triggers a gradual thermal shutdown with warning signs, not an instant black screen, and a driver crash usually leaves a log or recovers the display. A clean, immediate power cut the moment a demanding scene loads points strongly at the PSU tripping on a current excursion.
To confirm, watch whether the shutdowns cluster around fast load transitions, such as a game menu snapping into a heavy scene, rather than during steady benchmark runs. Temporarily capping the frame rate or lowering the GPU power limit reduces the size of the spikes; if the shutdowns stop, you have your answer. The durable fix is a PSU with more headroom or a modern ATX 3.0/3.1 unit engineered to ride out spikes up to double its rating, which lets a high-end GPU behave without dragging the whole system down.
FAQ
Why does my PC shut off during gaming but not benchmarks?
Fast load changes in real games can trigger larger transient spikes than a steady benchmark, tripping a tight PSU’s over-power protection. If the shutdown is instant with no crash log, an undersized or older PSU reacting to GPU transients is a likely cause; more headroom or an ATX 3.x unit usually fixes it.
How much headroom covers transients?
Size the PSU so gaming load sits around 50–70% of its rating, which leaves ample room for GPU spikes. In practice that means roughly 850 W for a 300 W GPU build and 1000 W for a 400 W GPU, more if the unit is an older non-ATX 3.x design.
Bottom line
Transient spikes are sub-millisecond bursts where a GPU briefly draws far above its average, sometimes double. A PSU sized only to average power can trip its protection and shut the PC off. Leave headroom so gaming load is 50–70% of the rating, and prefer an ATX 3.0 or 3.1 unit with a high-end GPU for extra spike tolerance.
