How to test power supply fan noise under different loads
A power supply can be electrically excellent and still undermine a quiet PC if its fan ramps aggressively. Measuring that behavior requires more than placing a sound meter beside the case while running a single benchmark. Fan speed depends on load level, internal temperature, ambient conditions, the unit’s fan-control profile, and whether the system creates short or sustained power spikes.
A useful test therefore treats the PSU as part of a repeatable acoustic system. The goal is to capture idle noise, gradual load changes, sustained full-load behavior, and the transition between passive and active cooling modes. Results should reveal both the measured sound level and the character of the noise, such as bearing hum, clicking, tonal whine, or sudden ramping.
This method works for ATX, SFX, and SFX-L models, although compact units often require extra care because their smaller fans and denser internal layouts can produce a different acoustic signature. Consistent placement and temperature control matter more than expensive test equipment.
Prepare a controlled test environment
Begin in a quiet room with HVAC systems, desk fans, and unnecessary electronics switched off. Measure the room’s background sound before powering the test system. A sound level meter is preferable, but a calibrated measurement microphone connected to an audio interface can also provide useful results. Phone apps are suitable for rough comparisons, not precise published measurements.
Place the microphone at a fixed distance from the PSU exhaust, commonly 30 centimeters or 1 meter. Keep its height and angle unchanged for every unit. A tripod helps prevent small positioning errors, while a non-reflective surface beneath the system reduces sound bouncing back toward the microphone.
Test with the same chassis, fan configuration, cable arrangement, and room temperature. If the power supply is tested outside a case, use the same open-bench position each time. Record ambient temperature and humidity, because a warmer room can trigger higher PSU fan speeds even when electrical load remains identical.
Select hardware and monitoring tools
The system should draw enough power to exercise the PSU meaningfully without changing platforms between tests. A modern processor and graphics card are useful because they produce distinct CPU-heavy, GPU-heavy, and combined workloads. Disable motherboard fan profiles that could create distracting changes in system noise, or document their settings and keep them fixed.
Monitor AC input power with an inline power meter rather than relying only on software estimates. Software readings can miss motherboard and conversion losses, while the wall measurement reflects what the PSU actually receives. Log DC component estimates separately when available, but use one primary load reference for comparisons.
Record fan behavior with an infrared thermometer aimed at the PSU casing, a thermocouple attached to the housing, or a temperature sensor placed near the exhaust. Do not open the PSU to install a probe. Internal capacitors can retain dangerous voltage even after the unit is unplugged, and removing the cover can void its warranty or compromise safety.
For broader hardware context, compare the unit with models documented in Hardware Hounds reviews, especially when a published result includes thermal or acoustic measurements. A consistent external test remains essential because laboratory distances and equipment vary between outlets.
Build a repeatable load sequence
Allow the computer to sit at the desktop for at least 15 minutes before recording idle noise. Capture several readings rather than one instantaneous number. A 30- to 60-second average, plus the lowest and highest values, shows whether the fan is steady or periodically pulsing.
Increase power demand in fixed stages, such as 20, 40, 60, 80, and 100 percent of the PSU’s rated output. At each step, hold the load for several minutes and wait for the reading to stabilize. For longer thermal testing, hold each stage for 10 to 15 minutes. Small power supplies may need longer because their internal temperature rises slowly after the electrical load changes.
Use separate workloads to expose different behavior. A CPU stress test can produce a fairly stable draw, while a GPU benchmark may create rapid transients and more heat near the PSU intake. A combined CPU-and-GPU workload represents a demanding gaming or rendering scenario. Avoid calling a brief benchmark “full load” if the meter shows only a short power spike.
Capture sound pressure and fan behavior
Take the sound reading from the same microphone position at every stage, and note whether the PSU fan is stopped, spinning slowly, or running at high speed. A passive mode can make a unit appear silent at idle, but its eventual transition to active cooling may be abrupt. Record the load and temperature at which that transition occurs.
Decibel readings should be reported as A-weighted values when the aim is to describe perceived room noise. Use the same response setting and averaging mode for each result. Because decibels are logarithmic, a two- or three-decibel difference can be meaningful, while a one-decibel change may be difficult to hear in a normal room.
Numbers alone do not describe annoying sounds. Listen for motor ticking, bearing chatter, pulsing, resonance, and high-pitched electrical noise. A PSU that measures only slightly louder can be more distracting if it produces a narrow tonal sound. Make brief recordings from the test position when possible, and annotate each clip with load, temperature, and fan state.
| Test stage | Typical duration | What to record | Useful interpretation |
|---|---|---|---|
| Desktop idle | 15-minute warm-up, 60-second reading | dBA, ambient temperature, fan state | Shows passive-mode behavior and baseline system noise |
| Light load | 5–10 minutes | Wall power, dBA, fan speed if available | Reveals early fan activation or pulsing |
| Mid load | 10 minutes | dBA, PSU surface temperature, workload type | Shows normal gaming or productivity acoustics |
| Heavy GPU load | 10–15 minutes | Peak and average watts, dBA, tonal character | Exposes heat buildup and GPU-related transients |
| Combined load | 15–30 minutes | Stable watts, dBA, temperature trend | Tests sustained cooling performance |
| Recovery | 5–10 minutes | Time to slow or stop, dBA decay | Shows hysteresis and post-load fan control |
Separate PSU noise from the rest of the system
The graphics card, CPU cooler, case fans, pump, and coil whine can easily mask the PSU. Establish a baseline with the PSU installed but the system at idle, then repeat selected loads while manually fixing other fan speeds where safe. Do not stop essential cooling fans during a high-power test.
A practical isolation check involves changing the microphone angle slightly toward the PSU exhaust while keeping its distance constant. If the sound changes noticeably, the PSU is contributing to the measured level. Another option is to use a directional microphone, but avoid blocking airflow or placing equipment close enough to alter the exhaust pattern.
If the PSU fan remains stopped, the system may still be louder because of the GPU or case fans. Report the total system dBA and, separately, the observed PSU fan state. Claiming that a PSU is “silent” is misleading when the measurement describes a complete computer rather than the power supply alone.
Electrical noise deserves its own note. Coil whine can originate from the PSU, graphics card, or motherboard VRM, and it may change with frame rate rather than wattage. Test with a frame-rate cap and without one if the sound is relevant to the use case. A microphone recording and a written description help distinguish fan noise from electronic squeal.
Interpret thermal ramping and acoustic results
The most useful result is usually a curve showing load or wall power on the horizontal axis and sound pressure on the vertical axis. Add markers for fan activation and major speed changes. A slowly rising curve indicates gradual fan control, while a sudden jump suggests an aggressive thermal threshold or a narrow control range.
Compare sustained and short-duration results. A PSU may remain quiet for five minutes because its internal components are still cool, then become louder after 20 minutes. This is especially important for high-efficiency units with semi-passive modes. The recovery period also matters: some fans stop quickly, while others continue running to remove stored heat.
Efficiency affects noise indirectly. A lower-efficiency unit converts more input power into heat at the same output, potentially causing earlier fan ramping. However, efficiency alone does not predict acoustics. Fan diameter, bearing quality, heatsink design, firmware behavior, and the PSU’s thermal sensor placement all influence the final result.
When comparing models, state the exact load, test duration, room temperature, microphone distance, and whether figures represent the PSU in a case or on an open bench. A 35 dBA result at 30 centimeters cannot be directly treated as equivalent to 35 dBA at 1 meter. Clear test notes prevent readers from drawing false conclusions from superficially similar numbers.
Recommendations for reliable measurements
- Use the same microphone, distance, room, chassis, and cable layout for every PSU.
- Measure wall power and ambient temperature alongside dBA and fan state.
- Include idle, staged, GPU-heavy, CPU-heavy, combined, and recovery periods.
- Report average and peak readings, while describing tonal or intermittent noises separately.
- Never open a power supply or alter its internal fan control for a normal acoustic test.
A strong PSU fan-noise test produces evidence that readers can reproduce rather than a single impressive number. Keep the load stages long enough to expose thermal behavior, isolate competing noise sources, and publish the conditions beside every result. That approach makes differences between a quiet workstation unit, a gaming-oriented model, and a compact high-density PSU much easier to understand.
Use the same procedure in future hardware reviews and preserve the raw logs, audio clips, and temperature notes. Publishing the resulting charts and observations gives builders a practical basis for choosing a power supply that matches their performance needs without adding unnecessary noise to the room.