The Off-Key Bard
Technology

Why Do Electronics Make Coil Whine, and Can You Really Hear Electricity?

Or: why your graphics card squeals at 300 FPS even though nothing appears to be moving. You aren't hearing electricity. You're hearing tiny parts that were never meant to be loudspeakers.

Or: Why Your Graphics Card Squeals at 300 FPS Even Though Nothing Appears to Be Moving

You're in a game when an unfamiliar sound creeps in. It isn't the steady rush of fans or the clatter of a hard drive. It's a thin, high-pitched whine from somewhere inside the case, like a tiny insect trapped behind the graphics card. You quit the game and it vanishes. Open a simple menu, where the card is suddenly drawing hundreds of frames per second, and it comes back louder. The pitch shifts as you move around.

The instinct is to assume something is about to die. Electronics aren't supposed to make noise, right? Some of them can. The effect is called coil whine, and it's one of the stranger demonstrations that a computer is a physical machine. No speaker, motor, or fan is involved. A tiny component in the power circuitry is being pushed around by rapidly changing electrical forces, and it moves just enough to make a sound.

Why Chips Need Their Own Power Plants

To see where the sound comes from, start with the power. A desktop power supply delivers a few fixed DC rails, such as 12 volts. But a processor core runs at roughly one volt, and memory chips and controllers want their own carefully regulated supplies. Those voltages must stay steady while the load changes in an instant, as a processor jumps from idle to full tilt.

One way to drop 12 volts to one is a linear regulator, which simply burns the difference as heat. For a part drawing large currents, that would waste a staggering amount of power. So modern boards use switching regulators. The cheap mental model is a very fast faucet. A transistor switch connects and disconnects the input hundreds of thousands or millions of times per second, and energy-storage parts, an inductor and capacitors, smooth the pulses into a steady output.

The inductor is the star of this story. It's a coil of wire, often wound around or embedded in a magnetic core. Current through it builds a magnetic field and stores energy, and it resists sudden changes in current. When the switch is on, energy flows into the field; when it's off, the field gives the energy back. Pair it with capacitors and a rapidly pulsed input becomes a calm, low-voltage, high-current output. The name "voltage regulator module" covers all of this. PowerElectronicTips' explainer describes the usual design: a multiphase synchronous buck converter that steps a 5 or 12 volt bus down to what the CPU or GPU needs.

Multiphase matters for what comes next. Processors and graphics cards with higher current requirements often use more power phases, dividing the load among multiple buck-converter stages, typically with an inductor in each phase. Some controllers can also activate or deactivate phases as demand changes, improving efficiency at lighter loads. The same source says the phases are interleaved, switched at intervals of 360 degrees divided by the number of phases, and they share input and output capacitors. A big graphics card might therefore carry many inductors, each experiencing its own rapidly changing magnetic forces. Remember that scale: a 250 watt GPU at about one volt means on the order of 250 amps on the core side, ignoring conversion losses. Large currents plus fast changes give mechanical vibration plenty to work with.

How a Magnetic Field Becomes Sound

Changing current means changing magnetic fields, and changing fields exert forces. Three separate mechanisms can turn that into motion.

Magnetostriction. Some magnetic materials change shape slightly when magnetized. Repeat the magnetizing, and the core flexes in time with the current. The change is tiny; Power Integrations' application note on audible noise says that for common power ferrites it is under one part per million. It's still real, and it's one of several mechanisms the note lists for noisy magnetics.

Forces on the windings. Current-carrying wire in a magnetic field feels a force. The same note puts it plainly: "Current in the coil produces attractive and repulsive forces that tend to move the wires." If a winding isn't perfectly immobilized, it can shift microscopically with each change in current, and that motion can pass into the core, the housing, the solder joints, and the board.

Piezoelectricity in capacitors. Despite the name, "coil whine" isn't always a coil. Many small ceramic capacitors in power circuits are made of barium titanate, whose piezoelectric and electrostrictive properties make it flex when voltage changes. A review of "singing" multilayer ceramic capacitors in the journal literature puts it this way: the capacitor's own vibration is in the megahertz range, far above hearing, but it is transferred through the solder joint into the circuit board, which deforms and vibrates, and the board's vibration can land in the audible range of 20 hertz to 20 kilohertz. The authors identify three parts: the capacitor as the source, the solder joint as the path, and the board as the acoustic resonator. The Power Integrations note adds that high-dielectric-constant ceramics such as Z5U and Y5V are the most prone, while other dielectrics are quieter.

That last point is the central trick. A component moving a few nanometers can't push much air on its own. But once the motion couples into a board, an area of fiberglass and copper acts like a soundboard. A microscopic vibration, acoustically amplified by a large surface, can be surprisingly loud.

But the Switching Is Too Fast to Hear

Here's the puzzle. Many regulators switch at frequencies above 20 kilohertz, beyond what we hear, and hearing is commonly given as 20 hertz to 20 kilohertz, with the upper end falling with age. So why do we hear anything?

Because the switching frequency isn't the only frequency in the system. The load itself varies, and that variation modulates the electrical activity. A GPU doing work in bursts creates current changes at much lower rates, along with harmonics and sidebands. Regulators also change strategy when the load changes. A Power Electronics News article on acoustic noise in switched-mode supplies notes that burst mode or skipped cycles can create an audible switching pattern even when the underlying frequency is above 20 kilohertz. The component doesn't have to switch at 8,000 hertz to make an 8,000 hertz sound.

Add resonance. Everything physical has natural frequencies where it responds strongly to a repeated push, like a swing pushed at the right moment. Inductors, housings, and boards all have them. When some audible-frequency component of the electrical pattern lines up with a resonance, a small vibration becomes a big one. This is why two supposedly identical graphics cards can sound different. The circuit design is the same, but small differences in winding tension, core properties, mounting, solder, and board stiffness change how strongly each one vibrates.

Why a Menu Can Be Louder Than a Game

Back to the screech in the menu. In a demanding game at 90 frames per second, the load is heavy but fairly steady. A simple menu may have no frame limit, so the card renders 400, 500, or thousands of frames per second. Each frame is trivial, but the GPU alternates between bursts of activity and brief pauses at a furious rate, and the power demand moves with it. The regulator's switching pattern, ripple, or load modulation may shift into something that excites a resonance, and out comes a whine, often right when the frame rate peaks.

Don't over-read the frame counter, though. A card at 500 frames per second does not make its coils vibrate at 500 hertz. Each frame involves shaders, memory transfers, cache activity, and scheduling, and the regulator sees the sum of all of it. Frame rate correlates with the pattern of demand but doesn't set the pitch. The chain runs: the game shapes the workload, the workload shapes the regulator's behavior, the regulator shapes the mechanical vibration. That's why the whine can change pitch and character with no simple formula, and why it sometimes seems unrelated to how demanding the game looks.

Not Just Graphics Cards

GPUs are especially prone because they combine huge power draw with low voltage and heavy bursty loads. But the same physics is everywhere. The power supply itself contains high-frequency transformers, inductors, and capacitors, and sometimes the sound only appears when the GPU loads it heavily, so a noise that seems to come from the card may be the PSU. Motherboards carry the CPU's regulators near the socket and can whine as the CPU boosts or idles. Laptops pack regulators and charging circuits into a small shell, where an electrical tone is easy to hear at close range, especially when the fans are stopped and nothing masks it. Scrolling a web page can cause quick bursts of GPU and CPU activity that produce a faint chirp. Phone chargers and laptop adapters contain switching transformers that can hum or whine, and the pitch may change as the battery charges and its demand tapers.

Some devices are noisier at idle than under load. Regulators use light-load modes to save energy, working in bursts rather than switching continuously, and those bursts can create low-frequency modulation. Under heavier load the switching smooths out and the sound goes away. The circuit isn't struggling; its efficiency strategy has an audible side effect.

Your own setup also decides how obvious it is. A computer next to your ear reveals faint tones that would vanish under a desk. A solid closed case can attenuate some high-frequency noise, while a vibrating panel can add to it. And here's the ironic part: swap in quieter fans and a bigger cooler, and a whine that was always there suddenly stands out. Hearing is very sensitive to narrow tones against a quiet background.

Is It Dangerous?

Usually not. A noisy inductor can be working within specification, delivering stable power while irritating you. Coil whine is typically an acoustic characteristic, not evidence of impending failure. A squealing GPU can run for years. It may be present from day one, and it doesn't reliably get worse with age; some people report it settling, others see no change. Temperature changes materials' stiffness and magnetic properties, so a device may sound different cold versus after an hour, but there's no universal rule.

Different symptoms deserve a different response. Crackling, popping, arcing, smoke, burning smells, repeated power interruptions, or unexplained overheating are not coil whine. Stop using the equipment and disconnect power if it's safe. Don't open a power supply to investigate; its capacitors can hold a dangerous charge even when unplugged. A faint, steady whine that follows workload and a sign of electrical failure are very different things.

What Actually Helps

Since the sound depends on the electrical workload, the safest fixes change the workload.

  • Cap the frame rate. If the whine appears at very high frame rates, a limiter or V-Sync is the first experiment. If the monitor refreshes at 165 hertz and a menu runs at 600 frames per second, a cap stops the card rendering hundreds of unseen frames, which changes the load pattern. It isn't guaranteed to work, but it needs no hardware changes.
  • Adjust power limits or undervolt. Lower current demands can change the regulator's behavior and sometimes the noise, though they may only shift the pitch, and results differ from card to card.
  • Check the PSU and the source. The power supply interacts with the card's regulators, so swapping it can alter the result, but a pricier PSU won't automatically fix anything, and swapping hardware without finding the source is an expensive experiment. Confirm it isn't a failing fan bearing instead; fans change with airflow and speed, while coil whine can change almost instantly with electrical demand.
  • Use the warranty. If a new part is exceptionally loud, whether it qualifies for service depends on the manufacturer.

Manufacturers have their own tools. Encapsulated or molded inductors restrict movement, and potting or varnish can damp vibration. Power Integrations notes that adhesive bonding of core pieces is the most effective transformer fix, and that varnish choice has electrical side effects: a high-dielectric varnish can raise capacitance and cut efficiency. That's exactly why the popular home remedy is a bad idea. Dripping glue, epoxy, nail polish, or silicone onto a noisy component can trap heat, degrade thermal pads, damage nearby parts, compromise insulation, and create fire risk. Engineers choose damping materials with heat and electrical behavior in mind. A noisy component is not an invitation to improvise repairs on energized hardware.

The Bard's Take

We like to think of computers as pure information: windows, icons, ones and zeros, logic gates opening and closing in the abstract. Coil whine punctures that. Every bit lives in hardware, and hardware is made of materials that respond to forces.

Current makes a magnetic field. Change the current and the field changes. Changing fields push on windings, flex magnetic cores, and squeeze ceramic capacitors. The movements are microscopic, but if their rhythm lines up with a resonance, and the board amplifies it, you hear it. That's why the sound follows your workload, why it differs between identical cards, and why it can show up in power supplies, motherboards, laptops, and chargers alike.

It usually isn't the sound of electricity escaping, a component wearing out, or hardware about to fail. It's the sound of electrical energy interacting with physical materials never meant to be loudspeakers. You aren't hearing the bits being calculated, and you aren't hearing electrons racing down a trace. You're hearing a power-delivery system working hard enough to make itself heard, and a circuit board that, for a moment, sings.

Sources

← Back to the Desk — Technology