← All guides

Mining rig maintenance schedule: ASIC and GPU, on one calendar

Hardware · 9 min read

Educational content only, not financial advice. Whether a miner earns money depends on hardware, difficulty and electricity prices that change constantly; this guide is about keeping hardware alive, not about whether running it pays.

Every maintenance guide for miners agrees on the tasks, blow out the dust, check the fans, renew the thermal paste eventually, and disagrees on the calendar. One published schedule puts thermal paste at 12 months, another at 12 to 18, a third at 12 to 24. Fan replacement is “every 6 to 12 months” in one table and “every 12 to 24” in another. The disagreement is not sloppiness: the right interval depends on how dusty the room is, and most guides quietly assume different rooms. This guide puts the published schedules side by side, reconciles them into one calendar keyed to environment, adds the GPU-rig tasks the ASIC-focused schedules leave out, and states the stop rules, the findings that mean a cleaning session has become a repair, or should not be attempted at all.

Why the calendar matters more than the cleaning

Dust is a thermal problem before it is a dirt problem. Dust buildup on heatsinks and fans reduces cooling efficiency; when chip temperatures climb, most ASIC firmware throttles hashrate to protect the hardware, and a dirty machine can lose a substantial share of its rated hashrate without raising a single error alert. Hotter chips also draw more power at the same frequency, so the joules-per-terahash figure worsens at the same time as output falls, the machine earns less and costs more simultaneously. Fans fail in cascades: one dead fan leaves the remaining fans working harder, running hotter, and dying sooner. And sustained high temperatures degrade solder joints, driver components and the chips themselves, which is how a skipped cleaning becomes a hashboard replacement. The maintenance question is therefore never “is it dirty?” but “which interval keeps this room’s dust from becoming this machine’s temperatures?”

The published schedules, compared

TaskRepair-lab schedule (home miners)Exchange-guide scheduleDeterioration-data schedule
Dashboard / error checkDaily (automated alerts preferred)—Weekly
Intake filtersInspect/replace monthlyEvery 1 to 3 months (filter replacement)Monthly visual inspection
Compressed-air cleaningExternal monthly; full internal quarterly (6 to 8 weeks if dusty)Every 1 to 3 monthsQuarterly full blowout
Fan inspection / replacementProactive replacement every 12 to 24 monthsInspection/replacement every 6 to 12 monthsFan RPM verification quarterly
Thermal pasteEvery 12 to 18 months (1 to 2 hours per hashboard)Every 12 to 24 monthsAssessment every 6 months; reapply annually
Full board / connector inspectionConnector and cable inspection every 6 monthsFull board inspection every 6 monthsBiannual firmware and PSU-cable check

Read the disagreements carefully and they mostly dissolve. Paste intervals differ because one source assumes warm rooms and 24/7 running (12 to 18 months) while another states a general range (12 to 24). Fan figures differ because one is an inspection interval and the other a replacement interval. The reconciled calendar below keeps both ideas: inspect often, replace on condition or at the outer interval, whichever comes first.

The reconciled calendar

Weekly (minutes). Read the dashboard: hashrate against its own recent average, chip temperatures per board, fan RPM, rejected-share or error counts. Early detection is what prevents cascading failures, and it costs nothing but attention.

Monthly. Inspect or replace intake filters; external dust blowout (about 5 minutes per miner on the repair-lab timings); listen to every fan for bearing noise, vibration change or speed fluctuation; confirm exhaust paths are unobstructed. Manufacturer manuals say the same thing in fewer words: clean fans and heatsinks with compressed air, always with the device powered off and unplugged, and keep the environment clean, cool and dry.

Quarterly. Full internal cleaning (30 to 45 minutes per miner on the repair-lab timings; every 6 to 8 weeks instead if the room is dusty or pets share the space, pet hair is a leading cause of clogged heatsinks in home repairs). Verify fan RPM against specification, inspect cables and ductwork, and check for firmware updates, from the manufacturer’s official site only.

Every 6 months. Full connector and cable inspection: power connectors checked for darkened or melted plastic, bent pins, looseness; PSU cleaning and inspection; thermal paste assessment, temperatures trended against the baseline recorded at the last deep clean.

Every 12 to 18 months (warm rooms, 24/7 running) to 24 months (cool, filtered rooms). Thermal paste renewal and proactive fan replacement if condition warrants. Quality fans degrade after roughly 30,000+ hours of continuous operation, about three and a half years of non-stop running, so a fan that sounds rough at the annual check is replaced on evidence, not on hope.

GPU rigs, same calendar, different tasks. The ASIC schedules barely mention GPUs, but the rhythm transfers: monthly dust blowout of cards and heatsink fins with fans held still; quarterly fan-bearing checks by spinning each fan by hand and feeling for roughness; thermal paste and pad assessment when a card’s temperature at fixed settings trends upward across quarterly baselines. GPU fans are smaller and cheaper than ASIC fans and fail at least as often under 24/7 load, so the listen-monthly rule matters more, not less.

The environment sets the interval

Three environment facts move a rig between the short and long ends of every range above. Ambient temperature: repair guidance designs for roughly 15 to 35°C ambient; a rig in a warm room earns the short intervals. Humidity: a target of 40 to 60% relative humidity balances static risk (too dry) against condensation and corrosion (too humid). Airflow geometry: at least 30 cm of clearance on both intake and exhaust sides, exhausts never facing each other or a wall. Dust control compounds all three: intake filters replaced monthly and filtered positive-pressure air in the room are what let a rig honestly use the quarterly internal-cleaning interval instead of the 6 to 8-week one.

Quarterly deep clean: a numbered procedure

1. Record the baseline first: 24-hour average hashrate, chip temperatures (min, max, average per board), all fan RPMs, wall power draw, and any error-log entries. Without a baseline there is no way to prove the cleaning did anything.

2. Shut down through the web interface, do not just pull the plug, then disconnect power and Ethernet, and wait at least 15 minutes for chips to cool. Ground against static before handling boards.

3. Remove and clean or replace intake filters. Blow the exterior, fan grilles and vents from intake toward exhaust, following the natural airflow direction.

4. Open the case (photograph the layout before disconnecting anything). Hold each fan blade stationary while blowing it out, letting a fan spin freely in compressed air can damage its bearings, then spin each by hand: smooth and free is healthy; roughness, grinding or wobble means replacement now, not at the next interval.

5. Blow heatsink fins at about a 45-degree angle, working along each hashboard; brush compacted dust loose first. Inspect boards for corrosion, burn marks, bulging capacitors and damaged traces, and every power connector for darkening, melted plastic or bent pins.

6. Blow out the control board, reseat ribbon and data connectors, reassemble, reconnect, and power on. Allow 10 to 15 minutes to reach steady state.

What success looks like: chip temperatures measurably lower than the step-1 baseline, hashrate restored to (or toward) the machine’s known-good figure, fan RPMs consistent across identical fans, and no new error-log entries over the following hour. If temperatures dropped but hashrate did not recover, the problem is firmware or board-level, not thermal, further cleaning will not fix it.

Stop rules: when cleaning becomes repair

Thermal paste is a decision, not a reflex. One careful source warns explicitly against a universal repaste interval: unnecessary disassembly can damage hardware or affect warranty coverage. Reapply when temperatures climb despite clean heatsinks and healthy fans, at the 12 to 24-month outer interval, or whenever a heatsink has been removed, once the seal is broken, old paste is never reused. When the job is done: heatsink screws loosened and tightened evenly in a star pattern, old paste fully removed with 99% isopropyl alcohol, a thin even layer per chip, and one hashboard at a time so a problem can be traced to the board that caused it.

Electrical findings end the session. Burned or loose 6-pin connectors, darkened plastic, or a PSU whose measured output deviates beyond about ±5% are repair items, not cleaning items. Likewise the troubleshooting triage for a board that disappears from the dashboard, power off, reseat ribbon cables and power connectors, check the PSU is delivering correct voltage before condemning the board, belongs to a repair session with the machine off the calendar, not squeezed into a quarterly clean.

The economic stop rule. When a repair quote exceeds roughly 50% of the machine’s current used-market value, or its efficiency makes it unprofitable at current difficulty regardless of repair, replacement beats repair. Maintenance extends the life of hardware that still has a case; it does not create one.

FAQ

How often should an ASIC be cleaned? Externally monthly, internally quarterly in a filtered room, every 6 to 8 weeks in a dusty one. Published schedules converge on that range once environment is accounted for.

How often does thermal paste need replacing? Every 12 to 18 months in warm 24/7 service, up to 24 months in cool filtered rooms, or immediately when temperatures rise despite clean heatsinks and good fans. Do not open a machine on a fixed date if its temperature trend is flat and its warranty is live.

What temperature is a warning sign? Sustained chip temperatures above roughly 80 to 85°C are treated as a warning sign for many models: improve cooling and airflow first, then investigate paste and fans.

Can cleaning be done while the miner runs? No. Manufacturer guidance is explicit: powered off and unplugged before cleaning, every time.

Next: Underclocking for efficiency →