Air-cooled ASIC miners use high-speed fans to move heat into the surrounding air. Hydro-cooled miners transfer heat into a coolant loop, while immersion systems transfer it into dielectric fluid surrounding the submerged hardware. Hydro and immersion cooling reduce miner-level fan noise, but pumps and external heat-rejection equipment can still generate substantial facility noise. The practical choice depends on the site's cooling infrastructure, maintenance requirements, and acoustic conditions—not just the miner's published noise rating.

Bitcoin mining hardware selection is often reduced to hashrate versus power efficiency (J/TH). Cooling method also changes where heat and noise are generated, how a facility must be engineered, and what an operator must monitor after deployment. The comparison below uses manufacturer specifications and system-level considerations to explain those differences.

Air, Hydro, and Immersion Cooling at a Glance

This framework is qualitative. It does not establish that one method is universally quieter or more efficient at a given property boundary; that requires a site-specific acoustic and thermal assessment.

What Makes ASIC Miners Loud?

An air-cooled ASIC generates most of its audible noise from high-speed fans that pull ambient air across the hashboards and exhaust heated air out of the chassis. Noise output rises with fan speed, and fan speed is normally controlled by the miner's internal temperature response to ambient (inlet) temperature and workload. For example, BITMAIN specifies the S21 XP's 76 dBA noise figure at maximum fan RPM. This is a stated maximum-fan-speed test condition, not an average operating noise level; other models need to be checked against their own measurement notes.

A manufacturer-published dBA figure is also incomplete without knowing how it was measured. A-weighted decibels (dBA) account for the ear's differing sensitivity across frequencies. A sound-pressure reading depends on microphone distance and room acoustics, while sound power describes acoustic output at the source; the two are not interchangeable. Without a stated measurement method, a single dBA rating should be treated as a limited, device-level reference point rather than a prediction of how loud a facility will be to a nearby worker.

Air Cooling: Simple Deployment, Fan-Driven Acoustics

Air cooling remains the simplest ASIC deployment method: no miner coolant plumbing, no coolant loop, and no dielectric fluid to manage. BITMAIN's Antminer S21 XP illustrates typical air-cooled specifications: a manufacturer-stated typical hashrate of 270 TH/s, wall power of 3,645 W, and wall efficiency of 13.5 J/TH, with power and efficiency specified at 25°C inlet-air temperature (BITMAIN S21 XP specifications). These figures are internally consistent: 270 TH/s multiplied by 13.5 J/TH equals 3,645 W.

The same specification lists 76 dBA at maximum fan RPM, with an operating range of -20°C to 45°C ambient. BITMAIN states that actual hashrate may vary by roughly ±3%, and actual wall power and efficiency may vary by roughly ±5%. Within the miner's operating limits, elevated intake temperatures or poor airflow management can increase cooling demand and cause its fans to run harder and louder than in a cooler, well-ventilated environment.

Hydro Cooling: Quieter ASICs, Not Necessarily Quiet Facilities

Hydro-cooled ASICs circulate liquid coolant through the miner's cooling system, which removes most of the miner-level fan noise. BITMAIN's own guidance describes its hydro miners as theoretically noise-free at the unit level and states that audible noise generally originates from the associated heat-dissipation equipment (BITMAIN hydro-miner guidance). In other words, hydro cooling relocates the noise source; it does not eliminate it from the facility.

The Antminer S21 XP Hyd. specification illustrates the scale of that supporting infrastructure. BITMAIN lists a typical hashrate of 473 TH/s, wall power of 5,676 W, and wall efficiency of 12.0 J/TH, with power and efficiency specified at 35°C inlet-coolant temperature. Its permitted coolant-inlet range is 20°C to 50°C, required coolant flow is 8.0–10.0 L/min per miner, and maximum coolant pressure is 3.5 bar (BITMAIN S21 XP Hyd. specifications). As with the air-cooled model, the power and efficiency figures are internally consistent: 473 TH/s multiplied by 12.0 J/TH equals 5,676 W.

BITMAIN also states that hydro miners require dedicated external heat-dissipation equipment and cannot operate as standalone units. Pumps and external heat-rejection equipment are therefore part of the mining system, including its noise and energy profile.

The 13.5 J/TH versus 12.0 J/TH comparison is not direct proof that liquid cooling alone produced a more efficient miner. The figures describe different products tested under different inlet conditions: 25°C air versus 35°C coolant. They are useful manufacturer specifications, but they do not constitute a controlled cooling-method experiment.

Immersion Cooling: Low Fan Noise, Different Maintenance Demands

Immersion cooling submerges ASIC hardware in a dielectric fluid, reducing miner-level fan noise while transferring heat into the fluid. The heat must then be removed through the system's heat exchangers and external heat-rejection equipment. Depending on the design, that equipment can include pumps and fan-equipped dry coolers or cooling towers. For example, a published ASIC immersion-cooling installation uses a four-fan dry cooler and a circulation pump (Dry Coolers: Cryptominer Immersion Fluid Bath Cooling). Removing the miner's fans therefore does not make the full installation silent.

Immersion also introduces engineering considerations distinct from air and hydro cooling. A 2026 IEEE study on single-phase immersion-cooled data-center servers identifies potential failure mechanisms including absorption, environmental stress cracking, corrosion, and oxidation. Its abstract concludes that directly applying existing reliability standards developed primarily for air cooling can miss immersion-specific failure mechanisms and lead to inaccurate assessments (IEEE: Reliability Risk Assessment and Standardization Strategy for Single-Phase Immersion-Cooled Data Center).

These findings concern potential risks in single-phase immersion-cooled servers; they do not establish failure rates for ASIC miners or show that every immersion installation will experience those problems. For mining operators, they support evaluating fluid and component compatibility alongside maintenance procedures, fluid condition, and pump or heat-exchanger upkeep.

How to Compare ASIC Specifications Fairly

Manufacturer specifications can be compared as reference points when their test conditions are clearly stated. Differences in inlet temperature, product design, or operating mode limit what can be concluded from that comparison. In particular, different inlet-air and inlet-coolant temperatures should not be treated as equivalent conditions, and comparing two products does not isolate the effect of cooling method alone.

Before comparing ASIC listings, check:

• The stated inlet-air temperature or coolant condition.
• Whether the noise figure specifies fan speed, measurement distance, and measurement type.
• Whether hashrate and power are labeled as typical or measured values.
• Whether efficiency (J/TH) uses wall power and hashrate from the same operating condition.
• Which equipment is included in the power measurement.

Miner-level wall efficiency does not by itself describe facility-level energy use. When comparing complete cooling setups, include the electricity used by external pumps, fans, and other cooling equipment, using a consistent measurement boundary and period. Otherwise, a comparison may include onboard fan power for an air-cooled miner while omitting the external cooling load of a liquid-cooled installation.

Identical stated J/TH also does not imply identical deployment requirements. In its September 16, 2025 launch announcement, Bitdeer listed the SEALMINER A3 Pro Air at 290 TH/s and the A3 Pro Hyd at 660 TH/s, both at 12.5 J/TH (±5%) (Bitdeer A3 launch specifications). These historical launch specifications illustrate why matching efficiency figures do not make two products interchangeable: the hydro variant has substantially more hashrate and power draw per unit, changing electrical and cooling-infrastructure requirements.

Evaluating a Site, Not Just a Miner

Fleet-level noise does not scale simply by counting miners. For independent, uncorrelated sound sources, sound levels combine logarithmically. If ten sources each contribute 76 dBA at the same receiver location, their combined level is approximately 86 dBA:

76 dBA + 10 × log₁₀(10) = 86 dBA.

This calculation does not apply simply because ten miners each carry a 76 dBA specification measured elsewhere. It is a mathematical illustration, not a prediction for a specific facility. Actual site noise depends on distance, spacing between units, enclosure design, sound barriers, indoor reflections, and the operating state of fans or cooling equipment.

U.S. occupational noise guidance provides a separate reference for worker exposure. NIOSH recommends an exposure limit of 85 dBA averaged over an eight-hour workday. OSHA's general-industry rules set a hearing-conservation action level at 85 dBA and a permissible exposure limit of 90 dBA over eight hours (NIOSH: Noise-Induced Hearing Loss; OSHA: Occupational Noise Exposure). These are U.S. occupational references, not universal site-boundary noise limits.

A single miner's 76 dBA maximum-fan-speed rating cannot be directly compared with worker-exposure thresholds. Exposure depends on the sound reaching the worker over time, including noise from other machines and cooling equipment. Facilities intended for regular staff presence should assess sound at relevant work locations and exposure over the workday rather than relying on component-level specifications alone.

Monitoring After Deployment

Once a cooling system and ASIC fleet are commissioned, several measurements should be tracked separately:

• Device-level data: hashboard temperature, fan speed or coolant flow and pressure, and wall power.
• Facility-level data: cooling-equipment power consumption and performance, plus measured sound levels at relevant locations.
• Pool-side data: pool-estimated hashrate and share-submission data, such as accepted and rejected shares where reported.

Pool-estimated hashrate is useful for checking mining activity after a cooling-system change, but it reflects share-submission activity rather than a direct electrical or acoustic measurement. It should not be used to infer device efficiency.

ViaBTC's mining guide states that, after a miner has been stable for around 10–15 minutes, its operation status and earnings can be checked through the Workers and Earnings pages (ViaBTC Help Center: BTC Mining). This is a reference for reviewing pool-side information, not a prescribed waiting period for thermal stabilization or acceptance of a new cooling setup.

FAQ

Is hydro cooling quieter than air cooling?

Hydro-cooled ASICs typically produce little or no fan noise at the miner itself, but the facility still requires pumps and external heat-rejection equipment, which can be significant noise sources. Overall site noise depends on the design and placement of that equipment, not on the miner alone.

Can I compare noise specifications across different ASIC models directly?

Only when the measurement conditions and quantities are comparable. Check fan speed, microphone distance, and whether a rating describes sound pressure or sound power. Without that information, published ratings are limited reference points rather than a reliable ranking of site noise.

Does switching to liquid cooling automatically improve J/TH?

Not automatically. Miner-level J/TH is the ratio of wall power to hashrate under specified operating conditions. ASIC design, operating mode, and temperature also affect it. Published specifications can be compared with their conditions stated, but they do not isolate the effect of cooling method. A facility-level comparison must also include external cooling power.

What should be monitored after commissioning a new cooling setup?

Track device readings such as temperature and fan speed or coolant flow, facility-level cooling performance and power consumption, and pool-side mining activity separately. Pool status alone does not establish that cooling performance or thermal stability is satisfactory.

References

• BITMAIN, S21 XP Specifications
• BITMAIN, S21 XP Hyd. Specification
• BITMAIN, Find More About Hydro Miners
• Dry Coolers, Cryptominer Immersion Fluid Bath Cooling
• IEEE Access, Reliability Risk Assessment and Standardization Strategy for Single-Phase Immersion-Cooled Data Center (2026)
• Bitdeer, SEALMINER A3 Series Is Here — Next-Gen Mining with Higher Hashrate and Efficiency (September 16, 2025)
• NIOSH, Noise-Induced Hearing Loss
• OSHA, Occupational Noise Exposure
• ViaBTC Help Center, BTC Mining