Repair Guide · Antminer S21 · Whatsminer · Hashboard cleaning · Thermal throttling
Dust, hotspots and hashrate: why cleaning is the cheapest ASIC repair
Every ASIC miner is a forced-air machine, and every cubic foot of dirty air it pulls through the chassis is a slow withdrawal from its hashrate. Dust rarely kills a machine outright; it throttles it, unbalances it, and eventually shorts it. Cleaning is the lowest-cost intervention on the bench — and the one most farms skip.
- 10-30%
- Hashrate loss from dust fouling before operators notice
- 20-40%
- Thermal resistance increase from a 1mm dust layer on fins
- ~300 CFM
- Airflow through a current-gen Antminer chassis
A current-generation Antminer moves roughly 300 CFM of ambient air through its chassis. That air crosses hash boards, voltage regulators, and the fin stacks sitting on top of every ASIC. Whatever particulate rides along with it — desert sand, construction dust, pollen, diesel soot from a nearby generator — ends up deposited on the surfaces that are supposed to shed heat. Nothing about that process is dramatic. It is a slow, compounding loss that shows up first as a slightly lower hashrate, then as a board that drops out when the room gets warm.
The three failure modes dust creates
Dust does not damage a miner in one way. It works through three mechanisms at once, and they reinforce each other.
1. Heatsink fouling and thermal resistance
The fins on a hashboard heatsink exist to maximize surface area in contact with moving air. A dust layer between those fins insulates the metal from the airflow. Published field guidance puts the effect at a 20-40% increase in thermal resistance from a 1mm dust layer on aluminum fins. That is not a rounding error — it means the ASIC die runs hotter for the same fan speed, the firmware sees the junction temperature climb, and it throttles clock frequency to stay inside thermal limits. Hashrate falls, and the operator usually blames the pool or the PSU before pulling a board.
2. Hot spots and uneven board stress
Dust does not deposit evenly. It accumulates where airflow is slowest — typically the downstream end of a hashboard, behind connectors, and in the shadow of the control board. That creates localized hot spots: individual chips or small clusters running well above the board average. Over months, repeated thermal cycling at those hot spots is what turns a cleaning problem into a solder-joint or chip-level failure. Boards that fail "only when hot" and recover after a cooldown are almost always telling you about a hot spot, not a dead chip.
3. Fan load and bearing wear
A fouled heatsink forces the fan to work harder to move the same heat. Higher duty cycle means higher RPM, more bearing load, and earlier fan failure — and a failed fan on a running miner is a much more expensive event than a dirty one. Dust also collects on fan blades and hubs, unbalancing them and adding vibration that stresses board mounts and connectors.
The economics in one line
A cleaning cycle costs labor hours and downtime. A hashboard replacement or chip-level repair costs parts, shipping, bench time, and the same downtime — plus the hashrate you already lost while the board was quietly throttling.Why cleaning is the cheapest repair you can do
Compare the interventions available to a farm operator on a per-board basis.
Cleaning and re-seating
Dry regulated air, fin inspection, connector check, fan service
First response to any throttling, fan fault, or board dropout
Thermal re-paste / pad replacement
Board removed, heatsinks lifted, interface material renewed
After cleaning, when temperatures remain high
Fan replacement
Swap failed or noisy fan, verify RPM and tach signal
Bearing noise, low RPM, or fan alarms
Component-level board repair
Bench diagnosis, reflow or replacement of failed components
Persistent errors after cleaning and re-seating
Board replacement
New or refurbished hashboard
When repair cost exceeds residual mining value
Cleaning sits at the top of that list because it is the only step that costs almost nothing in parts and can be done before any diagnosis is committed. It is also the step that determines whether the steps below it are even necessary. A board that reads zero hashrate because a fan is choked with dust does not need a bench — it needs a blower and a re-test.
What a proper cleaning actually looks like
"Blow it out with a compressor" is not a cleaning procedure. Shop air carries moisture and oil, and blowing dust through an assembled chassis drives particulate deeper into the fin stacks and onto the boards. Do it as a bench task.
Step 1
Power down and isolate
Shut the miner down, disconnect power, and let the unit cool. Never clean a running chassis — the fans will pull debris straight into the boards.
Step 2
Pull the hashboards
Remove boards from the chassis so debris exits the assembly rather than recirculating through it. Handle by the board edges and work on an ESD-safe surface.
Step 3
Dry-air the fin stacks
Use dry, regulated air or a dedicated blower at a distance and low pressure. Work along the fin direction, not across it. Excessive pressure bends fins and stresses solder joints.
Step 4
Inspect the hot spots
Look specifically at the downstream end of each board, behind connectors, and around the control board. Discoloration, matte or chalky residue, and corrosion are the markers to log.
Step 5
Service fans and filters
Clean blades and hubs, check for bearing noise and free spin, verify tach output after reassembly. Replace inlet filters rather than shaking them out.
Step 6
Reassemble and re-test
Re-seat boards, verify all connectors, then run and compare per-board hashrate and reported temperatures against the pre-clean baseline.
Log before and after
Record per-board hashrate, chip temperatures, and fan RPM before cleaning and again after. Without a baseline you cannot prove the clean paid for itself.
Fix the source
Cleaning a miner in an unfiltered room is a subscription, not a repair. Inlet filtration and positive pressure reduce how often you repeat the cycle.
Escalate on evidence
Only send a board to a repair bench when cleaning, re-seating, and thermal service have failed to restore it. That keeps bench capacity for real failures.
When cleaning is not enough
Cleaning restores a miner that is dirty. It does not restore a miner that has already been damaged by being dirty. The signals that you have crossed from maintenance into repair are consistent:
- A board that still reports zero or XXXX ASIC readings after cleaning and re-seating.
- A board that hashes correctly cold and drops out once the room warms up — a hot-spot signature that often survives cleaning because the damage is already done.
- Visible corrosion, arcing, or burnt residue on the board or at connectors.
- Fan alarms that persist after a known-good fan swap.
At that point the board goes to a bench for component-level diagnosis. The important operational point is sequencing: clean first, then diagnose. Sending a dust-choked board straight to a repair queue wastes turnaround time on a problem that a blower would have solved, and it inflates the apparent failure rate of your fleet.
The maintenance math
Dust-driven hashrate loss is documented in the 10-30% range before operators typically notice it. On a machine running at 15 J/TH, a 15% hashrate loss is a permanent, silent tax on every hour the miner runs — and it compounds across a farm. Against that, a scheduled cleaning cycle costs labor and a few hours of downtime per unit.
The cheapest ASIC repair is the one you never have to book: pull the boards, clean the fin stacks, service the fans, log the temperatures, and fix the filtration so you do it less often. Everything downstream — re-pasting, fan swaps, board repair, replacement — gets cheaper and less frequent when cleaning is the first thing you try.
Frequently asked questions
How often should hashboards be pulled and cleaned?
It depends on the environment, not the calendar. A farm pulling desert air or sitting near construction should inspect heatsinks monthly and deep-clean on a 60-90 day cycle. Filtered, positive-pressure rooms can stretch to 6 months or more. Use inlet filter loading and fin inspection, not a fixed schedule, to set the interval.
Can I just blow the miner out with compressed air in place?
No. Shop compressed air carries moisture and oil, and blowing dust through a running chassis pushes particulate deeper into the fin stacks and onto the boards. Pull the unit, power it down, and clean with dry, regulated air or a dedicated blower at a distance, with the board out of the chassis so debris exits rather than recirculates.
Does cleaning void the warranty or risk damage?
Cleaning itself does not, but careless cleaning does. ESD damage from improper handling, bent heatsink fins, cracked solder joints from excessive air pressure, and damaged fan connectors are the usual casualties. If a board is already throwing errors, clean it first, then re-test before assuming a chip-level repair is needed.
When does a dirty miner need actual board repair?
When cleaning and re-seating do not restore the board. Persistent XXXX or zero-hash ASIC readings, boards that drop out only when hot, and visible corrosion or arcing on the board point to component-level failure — often the downstream result of months of thermal stress. That is when it goes to a bench, not a blower.
Repair & maintenance
Get Your Miners Back to Full Hashrate
Chip-level diagnostics, component-level repair, and burn-in validation — on-site across the USA, no shipping required.