PSU Bottleneck Calculator — Is Your Power Supply Limiting Performance?

A power supply does not bottleneck a system the way a processor or graphics card does. It does not gradually slow things down. It delivers what the parts request, or it fails to, and failure looks like the machine turning off rather than fewer frames. The PSU bottleneck calculator above works from that distinction, which is why it asks for your symptom before your wattage.

A PSU Does Not Bottleneck — It Truncates

A processor or graphics card that runs short of capacity degrades smoothly. Ask more of it than it can give and frame rate falls in proportion. That is a slope, and every other page on this site is about slopes.

Power delivery does not work that way. A supply has a rating, and below that rating it delivers whatever is asked of it at full quality. There is no partial mode where it provides eighty percent of the requested current and the system runs eighty percent as fast. It either delivers or it protects itself.

There are two real mechanisms, and both are step changes.

Sustained load exceeding capacity. When the parts request more than the unit can supply for a sustained period, protection circuitry engages. What you see is the system shutting down or restarting under load, typically minutes into a demanding game rather than at the desktop.

Inadequate delivery causing lower power states. Where a supply cannot hold voltage stably, components detect the condition and fall back to conservative clocks to stay within safe limits. This one does look like a performance loss, which is why it gets misdiagnosed — but it is a defensive step down, not a gradual limit, and it typically arrives all at once rather than creeping in.

The practical consequence: if your symptom is a frame rate that is merely lower than you expected, your power supply is very unlikely to be the cause. If your symptom is the machine turning itself off, it is a leading suspect.

Why Does a PC Shut Down Under Load?

A shutdown under load is a protection circuit reacting to a brief current spike, not a shortage of average capacity.

Modern high-power graphics cards do not draw a steady current. They draw brief excursions far above their rated figure, lasting a fraction of a millisecond, as rendering load shifts abruptly between frames. This is normal, designed behaviour. The problem is that protection circuitry inside a power supply responds to current, not to averages, and a unit rated comfortably for your system’s average draw can trip its over-current protection on one of these spikes.

Here is why it confuses people so reliably. Stress-testing software ramps load smoothly and holds it steady, which is exactly the condition a supply handles best. So the system passes every test you throw at it. Then it shuts down at one specific loading screen, or during a particular scene transition, or when a shader compilation kicks in — and because it is repeatable and tied to a moment in software, it reads as a driver fault or a game bug. People reinstall drivers for weeks over this.

Newer supplies are specified for this behaviour explicitly. The ATX 3.x generation of the standard defines how far above its rated figure a unit must tolerate an excursion and for how long, and that tolerance — not the number of watts printed on the box — is what decides whether yours trips. Almost all of the load in question lands on the 12V rail, since both the processor and the graphics card are fed from it, so a unit’s 12V capability tells you more than its headline total does.

Two things follow. Wattage alone is a poor guide, because two units with the same rating can behave very differently on transient loads depending on the quality of their internal components and how conservatively their protection is set. And the cabling matters at high current. Use the cable supplied with the unit rather than a chain of adapters, seat the connector fully, and avoid tight bends right at the plug. Poor contact under high current produces faults that look exactly like a failing power supply.

How Much PSU Headroom Do You Actually Need?

Headroom serves three purposes, and only one of them is capacity.

Transient absorption is the first, and it is the reason the recommendation in the tool above is not simply your draw plus ten percent. A card prone to excursions needs the unit to have room to ride them out.

Efficiency is the second. Conversion efficiency is not flat across a unit’s range — it peaks somewhere around the middle and falls away at both extremes. A unit running at a small fraction of its rating is working in its least efficient region, which means more waste heat for the same delivered power. This is why buying enormously more wattage than you need is not automatically better: it moves you down into a worse part of the curve rather than up. The 80 Plus tier on the box describes that curve rather than a single point on it, which is worth knowing for what it does not cover: a higher tier tells you about conversion loss and heat, and says nothing about whether the unit will ride out a transient spike.

Upgrade room is the third, and it is the honest reason to buy above what you need today rather than any performance claim.

One thing the tool deliberately does not do: prescribe a purchase. It estimates draw and reports headroom. What you buy depends on how long you intend to keep the unit and what you expect to put in the machine next.

Is Your Power Supply the Problem?

The signature is failure rather than slowness.

Look for shutdowns or resets under load rather than at idle — a system that is stable browsing and dies in a game points at delivery. Resets during scene transitions and loading screens rather than during steady play are the transient signature described above. Clocks sitting below expected sustained values, on parts and cooling that should support more, can indicate a component defensively stepping down. And coil noise that rises and falls with load is a sign the unit is working hard, though not on its own a fault.

Which power symptoms indicate the supply and which do not, as described above.

What is not a power supply symptom: a frame rate that is simply lower than you hoped. That is what the PC bottleneck checker is for, or the graphics card compute and memory check if you want to know which of the two your card is short on.

If your clocks are dropping under sustained load but the system never actually shuts down, power delivery on the board is a more likely culprit than the supply itself — the motherboard bottleneck calculator covers that case. The power supply estimation model sets out the bands this page uses and where they come from.

Fixing a Power Delivery Problem

In order, and none of it raises frame rate:

1. Sort the cabling first. Use the cables that shipped with your own unit rather than ones from another supply, because pinouts differ between makes. Give a graphics card its own connectors instead of daisy-chaining several loads onto one cable, and seat every connector fully — a partly seated one carries current through fewer contacts than it was built for.
2. Reduce the load before replacing the unit. A modest power limit or undervolt on the graphics card lowers both sustained draw and the height of the spikes, which is often enough to settle a system that was shutting down.
3. Then replace the supply. Size it for transient headroom rather than for the average figure, and treat the result as a stability fix — a larger unit does not produce frames.

Which component to spend on first, when several are candidates, is covered by the general upgrade guidance on the homepage. A power supply is the clearest case of a part that fails without ever slowing anything down, which what a bottleneck is and is not sets out alongside the other three common misreadings. Where the symptom is lost frames rather than a shutdown, start with the two-minute check instead.