Is That Whine Coming from the GPU, the PSU, or the Motherboard? Coil-Whine Source Isolation & Load-Test Diagnosis Guide

2026-11-08 | GPU Repair | 18 minutes | HK GPU Fix Editorial

Is That Whine Coming from the GPU, the PSU, or the Motherboard? Coil-Whine Source Isolation & Load-Test Diagnosis Guide

Key Takeaways

A high-pitched whine inside a PC can come from the graphics card, the power supply, or the motherboard, because all three contain inductors that can vibrate under load. Pinpointing the source means changing one component's load at a time and watching which load the sound follows, not trusting where it seems loudest.

The Question Nobody Has Answered Yet

This site already explains what coil whine is and how to make it quieter. It also classifies fan noise and separates audio-path interference from airborne sound. But every one of those guides starts from the same unexamined assumption: that the whine is coming from the graphics card.

That assumption is the problem this guide exists to correct. The graphics card, the power supply unit (PSU), and the motherboard all contain inductors, and any inductor can whine. The three components sit centimetres apart inside the same case, and the sound reflects off the side panels, the glass, and the airflow, so the human ear cannot reliably tell which one is producing it. A whine that sounds as if it comes from the graphics card may be coming from the PSU or from the motherboard's CPU power delivery instead.

An earlier article on this site states the gap plainly: a frame-rate test changes the load on the graphics card and the PSU at the same time, so it cannot separate the two. That warning is the starting point here. This guide does not repeat the physics of coil whine or the list of ways to quieten it. It answers one question only: which component is making the sound.

One older English article on this site asserted that the noise source is the graphics card's inductor and not the PSU, but it offered no isolation test to support that claim — it simply assumed the source. That assertion is treated here as the mistake it is, and nothing from that article is reused: no figures, no prices, no branding, and no claims from it appear in this guide.

The Three Suspects: Wherever There Is an Inductor, There Can Be Whine

Coil whine is caused by magnetostriction — the tiny vibration of an inductor's core as current passes through it. That mechanism is already covered in the site's guide to coil whine, and it is not repeated here. What matters here is where inductors sit:

All three suspects share one property: when their load changes, the sound changes. That shared property is exactly what makes source isolation hard.

Why Locating the Source Is So Difficult: Three Confusing Factors

Load coupling. Gaming raises the graphics card load, the PSU total load, and often the CPU load all at once. All three suspects get louder together. Any observation of the form "it only whines when I play games" is therefore useless as source evidence — it is consistent with all three components.

Distance and reflection. The graphics card, the PSU, and the motherboard's CPU power area are a few centimetres apart inside the case. Sound reflects off the side panel, the glass, and the fan airflow, and the human ear's localisation error can be larger than the distance between the components themselves.

Similar pitch. All three produce a high-pitched whine whose tone follows the load. Listening to the character of the sound cannot separate them. A spectrum analyser can show that a sound exists at a particular frequency, but a narrow frequency peak cannot say which coil produced it — a spectrum is a clue, not a component diagnosis.

Why the Frame-Rate Cap Test Cannot Identify the Source

The frame-rate cap test taught in the site's earlier coil-whine guide — cap the frame rate at 60 and see whether the sound gets quieter — is a valid test for one question and an invalid test for another. It is valid for "is this coil whine?" If the sound falls with the frame rate, it is load-related whine rather than a fan or a bearing. It is invalid for "which component is whining?" because capping the frame rate lowers the graphics card load and the PSU load at the same time, so both suspects go quiet together.

Proving that the sound is whine is not the same as proving which component is whining. Confusing the two is the error this guide is built to prevent.

Step Zero: Confirm It Is Airborne Whine (Do Not Skip This)

Before any isolation test, confirm that the sound is airborne whine at all. Two quick checks separate it from the two lookalikes:

Only when both checks leave the whine still present does the isolation process below apply.

Step One: The Load Isolation Matrix (the Decisive Test)

The core method is not "listen more carefully" but "change only one suspect's load at a time". Use the same repeatable scene for every test, and change one variable per test.

TestGraphics card loadCPU loadSound appears?Points to
Idle desktopLowLow—Baseline — record whether there is any sound at all
3D scene or game menu, frame rate uncappedHighLowYesSound follows the graphics card load
CPU stress test, graphics card idleLowHighYesSound follows the CPU load — motherboard power delivery is the leading suspect
3D and CPU load high togetherHighHighYes, louder than 3D alonePossibly two sources, or the PSU total load
Frame rate capped at the monitor refresh rateMediumLowQuieter or goneProves load-related whine, but does not identify the source

Run the tests in this order and record the result of each one before moving on. The decisive comparisons are the second and third rows: a sound that appears only under 3D load follows the graphics card, and a sound that appears under a CPU-only stress test follows the motherboard's CPU power delivery. If the sound appears under both, do not force a single answer — two sources can exist at the same time.

Step Two: Positional Listening (Supporting Evidence, Not Decisive Evidence)

Under the same load, use a rolled-up paper tube or a funnel — one end to the ear, the other pointed at different positions: the graphics card area, the PSU exhaust vent, and the CPU socket area. Compare which position sounds loudest.

State the limitation plainly: case reflections mislead. Positional listening is supporting evidence only and can never settle the question on its own. It must be combined with the load isolation matrix from step one. A position that sounds loudest is a hint, not a verdict.

Safety note: positional listening is done from outside the case. Before removing a side panel, shut the machine down and disconnect the power. Never reach into a running machine, and never hold a tube, a microphone, or any object close to a spinning fan or a live component.

Step Three: Swap Comparison (Final Confirmation)

Swap one component at a time. Swapping two at once destroys the comparison, because the result can no longer be attributed to either one. The same one-variable rule from step one applies here.

Never open a PSU. A power supply contains high-voltage capacitors that can remain charged after the power is disconnected. Opening one is dangerous and is not part of any diagnosis in this guide. A PSU swap means replacing the whole sealed unit, never opening it.

Step Four: Reading the Results

Result combinationConclusion
Sound only under 3D load, and gone after the card is moved to another machineGraphics card inductor whine
Sound under a CPU stress test, and positional listening points at the CPU power areaMotherboard power-delivery inductor whine
Sound only at high total load, and gone after the PSU is swappedPSU whine
Sound under each load separately, at different positionsTwo sources at the same time — do not force a single answer
No test separates the sourcesRecord the test results and hand them to the inspection technician; do not replace parts on a guess

The last row is a valid outcome. An inconclusive isolation is not a failed diagnosis — it is a recorded result that keeps the next person from starting over. Replacing a PSU or a graphics card on the basis of a guess, when the tests could not separate the sources, is how money is wasted.

Common Misunderstandings, Taken Apart One by One

"A whine is always the graphics card." Wrong. Three components contain inductors, and the PSU and the motherboard can whine just as the graphics card can. The older article that asserted the graphics card as the source offered no isolation test for the claim.

"I capped the frame rate, the sound got quieter, so it must be the graphics card." Wrong. Capping the frame rate lowers the graphics card load and the PSU load together, so both suspects go quiet together. The test proves the sound is load-related whine; it cannot identify the source.

"Whichever position sounds loudest is the source." Wrong. Case reflections, fan airflow, and the short distances between components all mislead the ear. Positional listening is supporting evidence only and must be combined with load isolation.

"Replacing the PSU will definitely fix the whine." Wrong. If the source is the graphics card or the motherboard, a new PSU changes nothing. Locate the source before replacing anything.

"The louder the whine, the more dangerous it is." Wrong. Whine volume has no relationship to danger. The signals that actually matter are cracking sounds, a burning smell, visual artefacts, or a black screen.

"Whine means the card is about to die." Wrong. Whine on its own is a normal physical phenomenon, not evidence of a fault. It only becomes a concern when it appears together with other symptoms.

"A spectrum app showed a peak, so I know which component it is." Wrong. A spectrum shows that a sound exists at a particular frequency. It cannot say which component produced it.

What Whine Does Not Prove

Whine is not evidence of a VRM fault, and it is not evidence that a card is failing. A VRM fault shows itself through short circuits or loss of power, not through sound. Ceramic capacitor whine, caused by the piezoelectric effect, is a separate branch worth one sentence only: it exists, it is not the subject of this guide, and it is not evidence of failure either. Do not let a whine, however loud, become a reason to declare a component dead.

Changing the mounting orientation of the card can change how the sound propagates — but orientation is a boundary note here, not a fix.

When the Source Cannot Be Isolated

If the load matrix, positional listening, and swap comparison still cannot separate the sources, or if the whine appears together with other symptoms — cracking sounds, a burning smell, artefacts, or black screens — the next step is a physical inspection. Bring the recorded test results: which loads produced the sound, which positions sounded loudest, and which swaps were tried. Those notes are the most useful thing a technician can receive, because they prevent the whole diagnosis from being repeated from scratch.

HK GPU Fix offers a diagnostic inspection at its Shek Mun shop: Room 1209, Dali Plaza, Shek Mun. The published inspection fee range is HK$400–2,800 in total, and the actual price is subject to a post-inspection quotation. Phone 2186 8010, WhatsApp 6558 6806. There is no on-site service; inspection is carried out at the shop.

Frequently Asked Questions

How can I tell whether the whine is from the GPU, the PSU, or the motherboard?

Change one component's load at a time. Run a 3D scene with the CPU idle, then run a CPU stress test with the graphics card idle, and record which load the sound follows. That load isolation matrix is the decisive test — listening to where the sound seems loudest is not.

I capped the frame rate and the whine got quieter. Does that prove it is the graphics card?

No. Capping the frame rate lowers the graphics card load and the PSU load at the same time, so both suspects go quiet together. The test proves the sound is load-related whine, but it cannot identify which component is whining.

Can the motherboard really whine? I thought only graphics cards did that.

Yes. The power phases near the CPU socket contain inductors, and their load follows the CPU load. A whine that appears during a CPU-only stress test, with the graphics card idle, points at the motherboard's power delivery.

What is the first thing to check before trying to locate the whine?

Confirm it is airborne whine at all. Unplug the speaker and headphone cables — if the sound disappears, it is audio-path interference. Then check whether the sound disappears when the fans stop — if it does, it is fan noise. Only then does source isolation apply.

Is positional listening with a paper tube reliable?

It is supporting evidence, not proof. A rolled-up paper tube pointed at the graphics card area, the PSU exhaust, and the CPU socket area can show which position sounds loudest, but case reflections mislead the ear. Always combine it with the load isolation matrix, and do it from outside the case.

If I swap the PSU and the whine stays, what have I learned?

The PSU is excluded as the source. That is a real result, not a wasted test. Move on to the graphics card: test it in a different machine and see whether the sound follows the card or stays in the original machine.

Can two components whine at the same time?

Yes. If the sound appears under a 3D load and under a CPU-only load, at different positions, there can be two sources at once. Do not force a single answer — record both results.

Is coil whine dangerous, or does it mean my card is failing?

On its own, whine is a normal physical phenomenon — the vibration of an inductor under load — and it is not evidence of a fault or of impending failure. The signals that matter are cracking sounds, a burning smell, visual artefacts, or a black screen.

Can I open the PSU to find the whining part?

No. Never open a power supply. It contains high-voltage capacitors that can remain charged after the power is disconnected. A PSU swap means replacing the whole sealed unit, never opening it.

A spectrum app shows a clear peak at one frequency. Does that identify the component?

No. A spectrum shows that a sound exists at a particular frequency, but a narrow frequency peak cannot say which coil produced it. Use the spectrum as a clue that whine is present, and the load isolation matrix to find the source.

Related articles

Related services

If you need Power connector repair, see these service pages: