Start with the same game scene and compare its usual frame rate with a deliberate FPS cap. Record fan speed and where the noise is loudest. A cap changing the sound suggests load-sensitive noise, but it cannot identify GPU versus PSU alone. Use the result to decide whether to keep, service or exchange the part.
Contents
By LK Wood IV · Published 2026-10-04 · ~8 min read · St. Louis County, MO
A new graphics card starts buzzing, and the power supply sits just below it. Which one do you replace?
Before ordering either part, repeat the scene that bothers you, change the frame cap and listen again while watching the fans, because changing the GPU’s workload also changes what the power supply has to deliver. A cap alone cannot identify the culprit.
My ASUS ROG Astral RTX 5080 OC produced a faint whine only in uncapped game menus, audible with my ear at the case panel. Any frame cap removed it. That is one report from my card, running with an EVGA SuperNOVA 1000 GT, and the reusable observation records that same unit. It gives you a test to repeat; it cannot promise the same result from another Astral or another PSU.
Decide whether this is a noise test at all
Do not continue a listening test if there is smoke, a burning smell, visible connector damage or repeated shutdowns. Power down and get the fault assessed. A phone recording cannot establish that a crackling electrical sound is harmless.
For a stable system, ordinary coil whine is a plausible explanation. Corsair describes vibrating inductors, and ASUS describes load-related resonance in graphics cards. Neither description means every buzz from a PC is coil whine.
Keep the PSU enclosure closed.
Do not insert a tube, screwdriver or microphone through its grille. Corsair warns that PSU capacitors can retain charge after unplugging. Have the manufacturer handle a faulty PSU fan, or replace the supply.
Use one scene and change one setting
Download the GPU and PSU noise test log (CSV). It contains suggested test rows with blank results. Fill it in for your machine.
Start with the game or application where you hear the noise. Keep the same save point, camera position, graphics settings and case-panel state, and record the exact card and PSU models, including the revision if known, so that a later comparison does not accidentally mix two board designs or two versions of a supply. Mute the speakers.
| Check | What to record | What the result tells you |
|---|---|---|
| Desktop before the game | Fan state, noise location and whether the sound is present | A baseline for this session |
| Repeatable noisy scene | Actual FPS, GPU fan RPM if available, and noise at your seat | The symptom under the workload you care about |
| Same scene capped at 60 FPS | Whether the cap takes effect and whether pitch or loudness changes | Whether this lower frame-rate condition changes the noise; 60 is a test setting, not a required gaming target |
| Same scene at your preferred cap | The same observations | Whether a usable everyday setting solves your complaint |
| Restore the original setting | Whether the original sound returns | Whether the change repeats rather than coinciding with a different scene |
Use the game’s own limiter when available. Verify the resulting FPS on screen. A setting called “60” is no help if the menu ignores it. AMD documents that frame-rate limits can reduce GPU power demand, including in menus. Changing GPU demand also changes the load supplied by the PSU. If the noise changes, either remains a candidate. You have found a useful condition to repeat.
Do not combine the first comparison with other changes. Save the undervolt or new fan curve for a separate run; otherwise, even if the sound disappears, you will have no way to tell which change helped or which setting you should restore if a game becomes unstable later. If a cap solves the problem, stop there. For further tuning, use the GPU undervolt settings database and validate your own card’s stability separately.
Separate the fans from the electrical noise
Watch the fans as well as the FPS counter. A buzz that changes immediately with a scene or cap while reported fan RPM stays steady is a reason to investigate electrical noise. A rattle that begins with a fan starting is a reason to investigate that fan. Record the timing.
If a fan is visibly motionless while the noise continues, its rotation cannot be producing that sound. Do not extend that observation to every fan in the case. Some PSUs deliberately stop their fans: Corsair’s zero-RPM reference lists specific models and part numbers. A silent fan does not mean the PSU is electrically inactive.
For a GPU or case fan with a manufacturer-supported control, you can compare two permitted fan speeds at idle and then restore automatic control. Leave the pump alone. Do not force a fan to stop, put a finger into moving blades or disable cooling under load. If a card’s fans fail to operate when its own documentation says they should, follow the fan troubleshooting branch before continuing a gaming test.
Listen from outside the case near the graphics-card area, then near the PSU exhaust. Keep your hands and recording device clear of vents and fans. Note the apparent location, including “unclear.” If panels obscure the source, let a shop isolate it instead of reaching into a powered PC.
Leave the case closed for these comparisons.
Read reports for your exact product
The coil-whine census has a different job from this test log: it stores one accepted owner report per physical unit. The September 26 dataset still contained 47 reports across 44 models when checked on October 4. No individual model had reached the five-report publication floor.
These examples help choose the next check:
| Product in the census | Available context | Useful next step |
|---|---|---|
| ASUS ROG Astral RTX 5080 OC | One report, my menu-only observation with an EVGA SuperNOVA 1000 GT | Try the menu-cap comparison before buying a different PSU |
| XFX RX 6950 XT Speedster MERC 319 | Three reports; exact SKU is present in two | Match the model and available SKU, then read each workload; three units cannot establish a model-wide rate |
| Sapphire NITRO+ RX 9070 XT | Two reports | Do not substitute the wider RX 9070 XT group for this particular board |
| Corsair SF450 and Seasonic FOCUS PX-650 | One PSU report each | One owner’s result cannot choose a quiet replacement PSU for your GPU |
The source is the public submissions CSV, with issue numbers linking back to the owners. The GPU-family table mixes board designs. It cannot establish that an ASUS PRIME, Sapphire NITRO+ and PowerColor Red Devil will sound alike simply because all three use an RX 9070 XT.
The missing detail often matters more than the model name. “Capping FPS changes it” leaves out whether the owner tested a menu at 60 FPS with the case on the floor or gameplay at 144 FPS with the case beside their head, which makes it a poor substitute for repeating the test where you actually sit. Keep that context in your log.
Keep it, exchange it or test another supply
My first choice is to keep a stable system when a reasonable cap removes the annoyance.
If the card remains distracting at your normal seat during normal gameplay, contact the seller while its return or exchange window is open. Explain the exact symptom and ask which remedy applies. You do not need to call the card electrically defective to explain that its noise is unacceptable to you.
If the sound clearly comes from the PSU, start with that supplier’s support. If the source remains unclear and you have no suitable spare, pay for diagnosis before buying replacements by guesswork. For a loaned or shop-supplied comparison PSU, check the build’s power and connector requirements using the PSU wattage calculator and the component manufacturers’ instructions.
Change only the supply and its cables, then repeat the same scene and cap. Use the comparison PSU’s own supplied or explicitly approved cables throughout. Seasonic warns that modular cables can differ even between models from the same brand. Matching plugs are insufficient. A quieter second setup gives you a practical choice. It does not establish that the original PSU was defective, that the GPU alone caused the noise, or that a higher wattage rating guarantees silence. Record which configuration you kept.
What the manufacturer guidance says about service
The US ASUS graphics-card FAQ describes slight high-frequency noise as a component characteristic. Corsair’s US PSU support article describes the mechanism and directs concerned owners to support. Those pages do not decide an individual warranty claim, and they are not a blanket statement about every region or seller.
Send the exact model, purchase date, country, workload, FPS cap and a recording from your usual seat. Include any crashes or other faults separately. Ask for a written decision under the terms that apply to your purchase. Do not consume the return window running a promised “burn-in cure.”
Once you have identified the component, submit your unit to the census. A quiet result is useful too. If the source is still unclear, keep that uncertainty in the log rather than assigning the noise to a GPU or PSU in the public data.
Sources
Checked October 4, 2026. Manufacturer guidance below is separate from the subjective owner reports.
- Corsair US support: PSU coil whine.
- ASUS US support: graphics-card high-frequency noise.
- AMD: Frame Rate Target Control and GPU power demand.
- Corsair: why users should not replace a PSU’s internal fan.
- Corsair US support: zero-RPM PSU model reference.
- Seasonic: modular cable compatibility.
- TechFuelHQ census: submissions CSV and my Astral report, issue #1.
Frequently asked questions
How do I tell whether coil whine comes from my GPU or PSU?
How can I separate coil whine from fan noise?
Will a different PSU fix GPU coil whine?
Should I return a graphics card for coil whine?
Does coil whine automatically qualify for a warranty replacement?
Sources and corrections
- Last updated
- Methodology
- See our methodology for research and review standards.
- Corrections
- Spotted an error or a stale number? Email contact@techfuelhq.com. Confirmed corrections are recorded here with the date and change.