How Loud Is an ASIC Miner, Really?
The scale, explained once
Decibels are logarithmic, and three rules cover everything an operator needs. +10 dB is 10× the sound intensity and reads roughly twice as loud to human ears. Two identical machines add +3 dB, not double the number — two 75 dB miners are 78 dB, ten are 85, a hundred are 95, which is why a farm is a different animal from a unit. Each doubling of distance subtracts about 6 dB — in open air with nothing to reflect the sound, the condition acousticians call free field; real rooms reflect, and do worse. Worked once: 75 dB at 1 m → 69 at 2 m → 63 at 4 m → 57 at 8 m → 51 at 16 m. Meters usually report dBA — the same scale weighted to human hearing; the glossary has the fine print.
That is the whole toolkit. Point it at your own floor plan:
You add a second, identical 75 dB miner next to the first. What does the meter read?
The ladder, in lived terms
Classes, not a catalog — per-model noise rows live on /rigs, each labeled with whose number it is. Figures as of 2026-09-27:
| Noise class | dB at 1 m | Whose number | In lived terms |
|---|---|---|---|
| Open-source single-chip (Bitaxe class, ~18 W) | 35–40 | community-measured · no manufacturer spec exists | quiet-room hum, under a refrigerator |
| Hydro-cooled flagship | ~50 | manufacturer spec · measured-confirmed | steady pump hum ≈ quiet conversation |
| Air-cooled flagship (S19/S21/M60S class) | 75 | manufacturer spec at ~1 m, 25°C · shops read 73–76 [measured] | vacuum cleaner at arm’s length |
| Old-generation air-cooled (S9 class) | 76–80 | community/shop-measured | a louder vacuum with a harsher pitch |
| Degraded: worn fans, dust, hot ambient | up to ~90 | community-reported, anecdotal | gas-lawnmower territory |
Lived-terms anchors from CDC/NIOSH-class reference charts: refrigerator ~45 dB, normal conversation ~60, vacuum cleaner ~70–75, gas lawnmower ~85–90. The ~90 dB row is an anecdote class, not a spec — community reports of S19-class units with worn fans, dust-loaded boards or hot rooms pushing fans to maximum; no controlled measurement of it has been published.
Now the honest twist in this lesson’s title: “a vacuum cleaner” is fair on level and misleading on everything else. A vacuum runs twenty minutes; this runs 24/7, straight through the night. And the sound is not a broadband whoosh — a high-pitched fan whine carries annoyance well past what its dBA number suggests. Level, duty cycle, tone: three axes, and the spec sheet reports one.
Two thresholds, 45 dB apart — most pages quote only one. NIOSH puts the hearing-risk line at 85 dBA averaged over 8 hours, with every +3 dB halving the safe time; a 75 dB miner at 1 m sits below that line. The WHO Europe night guideline for sleep is 40 dB outside bedrooms — a full 35 dB under the spec figure. “Won’t damage your hearing” and “livable” are different questions, and at home the sleep number is the one that decides.
Do they get louder with age? Commonly, yes. Fans are the wear item, and worn bearings, dust-loaded boards and hotter ambient all push RPM — and pitch — up; old S9-class units read 76–80 dB [measured] with a harsher tone than their spec suggests. A miner that has grown louder is telling you something about its remaining lifespan — the fan hour ratings, and what they actually mean, live in the lifespan lesson.
Reading any noise row, anywhere: three questions. Whose meter — manufacturer spec or independent measurement? At what distance — the −6 dB rule needs a reference point? In which mode — rated fans at 25°C, an eco profile, or a worn unit in a hot attic? A dB figure missing any of the three is marketing, not data — and “whisper-quiet home miner” labels routinely decorate 45–65 dB devices, which is a running refrigerator-plus in a bedroom at 3 a.m.
A listing calls a 62 dB machine “whisper-quiet — perfect for home”. What is missing before that number means anything?
Why it screams: noise is the exhaust of heat
A 3.5 kW air-cooled flagship converts essentially every watt it draws into heat — about 11,900 BTU/hr (unit arithmetic, 1 W = 3.412 BTU/hr, on the manufacturer’s wattage), a space heater that never cycles off. Moving that heat through a shoebox-sized steel case takes serious airflow: typically four industrial 120×38 mm fans at roughly 6,000–7,000 RPM [manufacturer figures via parts listings] — high-static-pressure units built to force air through hot, dense hashboards, nothing like the lazy fans in your PC.
Here is the physics that locks noise and heat together: fan sound power rises with roughly the 5th power of fan speed — a standard engineering approximation. Double the RPM and you add about 15 dB; halve it and you drop about 15, and lose most of your cooling with it. Your desktop whispers because it sheds a few hundred watts with slow fans; a miner screams because it sheds 3,500 with fast ones. Pulling 200 trillion drawers a second is silent — cooling the arm that does it is not.
That coupling is the master key to this whole subject: every quieting trick is secretly a heat-routing trick. Nothing below — distance, boxes, water, firmware — escapes it. You never solve noise alone; you decide where the heat goes, and the noise follows the decision.
The room test: apartment, bedroom, garage
The overall should-you-mine-at-home verdict lives in Module 1’s feasibility lesson — what follows is the placement framework for a machine you’re already considering. Four gates, in the order they usually kill the plan:
Gate 1 — the circuit. 3.3–3.6 kW at 120 V is roughly 28–30 A — past any standard US wall outlet. A full-size air-cooled miner is a 240 V dedicated-circuit appliance, dryer-class. That is arithmetic on the spec sheet, not electrical advice — have an electrician confirm anything that touches your panel. For most apartments this gate closes the case before noise gets a hearing.
Gate 2 — shared walls. The WHO night guideline is 40 dB outside bedrooms. Starting from 75 dB [spec], the free-field math needs five to six doublings of open air — 32 to 64 meters — to get there, and no apartment has them. Worse, the −6 dB rule assumes air: structure-borne vibration through studs, joists and slabs ignores the math entirely, which is why the classic neighbor complaint is a hum in the wall, not a noise in the air.
Gate 3 — lease, neighbors, ordinances. A receipt instead of a lecture: in October 2024, residents of Granbury, Texas — represented by Earthjustice — filed a private-nuisance lawsuit against Marathon Digital (MARA) over 24/7 noise from its Wolf Hollow bitcoin facility, with more than 40 residents alleging health effects including tinnitus, hypertension and vertigo; MARA said its measurements were below legal limits and built a sound wall, and the litigation remained active into 2026. That is industrial scale — recall the +3 dB rule: a hundred 75 dB machines land near 95 — but it is the same physics your neighbor hears through one wall, and the same complaint, one lawsuit smaller.
Gate 4 — the heat, again. The room the miner is in becomes the room the heat is in: ~11,900 BTU/hr with no thermostat and no off-season. A bedroom wall fails gates 2 and 4 before anything is plugged in. A detached garage with free air and its own circuit: maybe — run the widget above with your real distances, and remember an enclosed garage reflects. That is the honest spread.
Quieting, as categories only
No products here — the no-affiliate rule is absolute, and “silencer” listings routinely skip the thermodynamics. Five honest categories, each a decision about where the heat goes:
Distance. The cheapest −6 dB you will ever get, and it stacks: every doubling of open-air meters buys another 6. It works until walls, reflections and structure-borne vibration break the math — distance is air, and walls are not distance.
Enclosure. Every dB a box blocks is heat it traps. The fans answer by climbing the 5th-power curve — spending the quiet you just bought — or the unit throttles; an unventilated DIY box is an overheat and fire hazard, stated plainly. A serious enclosure is a ducted-airflow project, which is to say: a heat-routing project.
Firmware derating. Run the chips and fans slower and the machine gets genuinely quieter — roughly −15 dB per halving of fan speed — at the price of hashrate. An honest trade with no hardware at all; the efficiency math it moves along lives in the J/TH lesson.
Hydro or immersion. Move the heat into water and the scream disappears; what remains is ~50 dB [spec] of pump and radiator hum. Quieter by ~25 dB — roughly five times less perceived loudness, never silent — and bought with plumbing, pumps and new failure modes. Hydro variants appear on /rigs as presets land, noise rows labeled like every other.
Or change the problem. An 18 W machine has almost no heat to shed, hence almost no noise to make: 35–40 dB [measured] from one small fan. That is the Bitaxe path — and the category difference between a desk object and a 3.5 kW appliance is its own lesson.
You build a sealed, sound-insulated box around a miner. A week later, what is the most likely state?
FAQ
- How loud is an ASIC miner really?
- Air-cooled flagship miners (S19/S21/Whatsminer class) are spec’d at 75 dB at about 1 meter [manufacturer spec]; independent shop measurements read 73–76 dB when fans and ambient are nominal [measured]. That’s a vacuum cleaner at arm’s length — except it never switches off, and worn fans or hot rooms are reported to push units toward ~90 dB [community-reported].
- Why are Bitcoin miners so loud?
- Because noise is the exhaust of heat. A 3.5 kW miner turns essentially every watt into heat — about 11,900 BTU/hr, a space heater that never cycles — and pushes it through a shoebox with four industrial fans at roughly 6,000–7,000 RPM. Fan noise rises with roughly the 5th power of fan speed, so the airflow that keeps the chips alive is precisely what screams.
- Can I run an ASIC miner in an apartment?
- An air-cooled full-size unit fails the placement gates in every direction: a bedroom shares a wall (the WHO night guideline is 40 dB — about 35 dB below the 75 dB spec figure, and an apartment has no distance to spend); 3.5 kW is roughly 29 A at 120 V, past any standard wall outlet; and few leases survive either. The honest apartment path is an 18 W Bitaxe-class miner — and the overall home-feasibility verdict lives in Module 1.
- How can I make an ASIC miner quieter?
- Four honest categories, each a heat-routing decision: distance (−6 dB per doubling, open air only); an enclosure (every dB blocked is heat trapped — fans spin up or the unit throttles, and an unventilated box is a hazard); firmware derating (trading hashrate for silence); or moving the heat into water — hydro or immersion, with ~50 dB of pump hum remaining [spec]. Or shrink the problem: an 18 W open-source miner has almost no heat to shed. We recommend no silencer products — only the physics.
- Are water-cooled (hydro) miners silent?
- No — quieter, not silent. Hydro variants are spec’d around 50 dB [manufacturer spec, measured-confirmed]: a steady pump hum near quiet-conversation level, roughly five times less perceived loudness than the 75 dB air-cooled figure. You buy those 25 dB with plumbing, pumps and a radiator.