Innosilicon A10 Pro – Fan Error

Hard power-cycle for 5 minutes at the breaker. Not a soft reboot. Full power-off long enough that capacitors discharge and the control board re-initializes clean. Record the exact fault string on the UI before and after. Sometimes a wedged fan-controller process on the control board clears with a cold boot, and you save yourself an hour of diagnostic for zero dollars. If the fault clears, monitor 24 hours before closing the ticket — it may recur and indicate a fan on the edge.

Shop-vac the intake mesh, fan blades, and visible heatsink fins through the exhaust. Use a soft-bristle brush for stubborn build-up. Never use canned air on an ASIC — it drives dust into chip sockets, capacitor channels, and fan bearings and makes every problem worse. A thorough clean on a 2-year-old miner recovers `3–8 °C` of thermal headroom before you do anything else, and often the fan fault disappears the moment the fans can pull air again. Schedule this as a 30-day recurring maintenance task.

Verify intake air temperature with an IR thermometer at the front grille. Aim at the grille itself, not room ambient. Target `≤ 30 °C` sustained for normal operation. Above `30 °C`, the firmware ramps fan PWM harder, which accelerates bearing wear and can push RPM demand beyond what a dust-loaded fan can deliver. In a Canadian garage in July, the intake can hit `35 °C` even when the room feels fine — add a box fan, duct cooler air from the shady side of the building, or relocate the miner before summer bites.

Inspect every fan blade edge with a bright flashlight for chips, cracks, bends, or embedded debris. Blade damage is often under `2 mm` and easy to miss at a casual glance. Rotate the fan slowly by hand and watch each blade edge as it passes the light. A single cracked or bent blade throws the fan out of balance, raises bearing load, and eventually trips the RPM threshold. If you find damage, move to Tier 2 fan replacement — there is no field repair for a broken blade.

Spin-down coast test. Power off. Spin each fan by hand and time how long it coasts to a stop. Healthy A10 Pro fan: `6–10 seconds`, smooth and silent. Worn bearing: `1–3 seconds`, gritty or clicky feel. If one fan clearly coasts shorter than the other, that bearing is dying — plan to replace it before it fails under load and torches a hashboard via thermal cascade.

Re-seat both fan connectors at the control board. Power off at the breaker. Unplug the 4-pin fan connector from the header, inspect the pins for corrosion, blackening, or bent geometry, and reconnect firmly until the latch clicks home. Vibration and thermal cycling walk these connectors loose over time — this alone resolves a material fraction of "fan reports zero RPM" tickets on the bench. While you're in there, inspect the wire at the strain-relief for fraying or cracked insulation.

Measure `12 V` rail at the PSU-to-control-board connector under load. Multimeter on DC, probe the `12 V` / GND pair while the miner is hashing at full power. Expect `11.6–12.6 V` sustained. Droop below `11.4 V` means PSU is tired or cable is undersized — swap PSU with a known-good unit. A sagging `12 V` rail starves the fans of the current they need to maintain commanded RPM, which trips the fault even though the fans themselves are fine.

Replace the failing fan with a matched 120 mm × 38 mm PWM industrial fan. Stock part numbers vary by Innosilicon revision — check the sticker on your specific fan. Match form factor (`120 × 38 mm`), voltage (`12 V DC`), connector (4-pin PWM), and static-pressure rating (`≥ 7 mmH₂O` at nameplate). Delta `AFC1212DE-SP00`, Nidec `TA450DC`, and Sanyo Denki 9G1212G101 are direct or near-direct replacements depending on revision. Stick with named industrial brands — generic case fans lack the static pressure to move air across dense ASIC heatsinks.

Replace fans as a matched pair, not one at a time. When one fan bearing has worn out on a 4-year-old A10 Pro, the other is typically within months of the same failure. Replacing one and running a new fan in series with a tired one creates an airflow imbalance that stresses the new fan's bearing and throws thermal imbalance across the three hashboards. Budget for two fans, swap both, reset the clock.

Verify line voltage at the panel. `240 V` split-phase should measure `235–245 V` at the outlet under load; `208 V` commercial should measure `202–212 V`. Low line voltage forces the PSU to draw more current, which in turn droops the `12 V` rail to the fans, which eventually trips RPM faults even on healthy fans. This is the "same hour every evening" fault pattern — when the neighbour's A/C kicks on at 6 PM, your fan faults like clockwork.

Inspect the 4-pin fan header on the control board under magnification. Power off. Pull the control board. Use a loupe or USB microscope to inspect every solder joint on the two fan headers for cold joints, lifted pads, or cracked-looking solder. Flux the joints and re-flow with a fine-tip iron at `340 °C`. This is a 15-minute bench job and resolves the fault pattern where one fan header reads zero RPM intermittently on a known-good fan. After reflow, test continuity from header pin to expected control-board net.

Re-apply thermal paste on all three hashboards. Arctic MX-6 or Thermal Grizzly Kryonaut, uniform thin layer. Clean old paste off ASIC tops and heatsink surfaces with `99%` isopropyl alcohol and lint-free wipes. Inspect any thermal pads for crumbling or dryness and replace with `1.0–1.5 mm` pads of matching conductivity. Dried paste forces the fan to ramp harder to maintain the same Tj — which wears bearings faster and is a stealth cause of fan-faults-that-are-really-thermal-faults. This is an 18–24 month maintenance item on 24/7 miners.

Replace the fan header if reflow doesn't hold. If reflowing the 4-pin fan header brings RPM reading back but it fails again within days, the header itself is mechanically fatigued. Desolder the old header with a hot-air station at `310 °C`, clean the pads with solder wick and flux, and install a new 4-pin Molex KK-254 or compatible header. Test continuity on every pin before reassembling.

Diagnose the fan-driver transistor / tach buffer if the header is mechanically perfect. With a known-good fan plugged in and the miner powered on (carefully), probe the PWM pin at the fan header with a scope — expect a `25 kHz` PWM waveform with duty proportional to commanded speed. Probe the tach pin with the fan spinning — expect two low-pulses per revolution. Missing PWM output = fan-driver transistor failed. Missing tach response = tach buffer failed. Both are small surface-mount components that can be identified from Innosilicon's informal schematic floating on repair forums — Tier 4 territory unless you're comfortable with board-level SMT work.

Roll firmware to the last-known-good version for your specific A10 Pro revision. Download from the official Innosilicon portal only. Verify the hardware revision sticker on the control board matches the firmware image — a mismatched image can brick the board. Do not trust third-party firmware rumors — DCENT_OS and other custom Antminer firmwares do not apply to Innosilicon, and anyone selling you a "custom A10 Pro firmware" is selling you a doorstop.

Stop DIY when: reflow of the fan header holds for less than a week, both fan positions intermittently fault with known-good fans, you've measured missing PWM or missing tach signal at the control-board silicon (not the connector), or you smell burnt PCB / see discoloration anywhere on the control board. You are now in full-board-level repair territory. [Book a D-Central ASIC Repair slot.](https://d-central.tech/services/asic-repair/)

What D-Central does at the bench. Test fixture with known-good fans and programmable PSU, control-board-level probing with scope and logic analyzer, fan-driver transistor / tach-buffer replacement, header replacement where fatigue is confirmed, thermal paste and pad refresh across all three hashboards, and post-repair 24-hour burn-in at nameplate with both fans logged for stability. If the controller is genuinely beyond economical repair on a 2020-era A10 Pro — and in 2026 that's a real question given the miner's post-Merge earning window — we'll tell you straight and help you plan salvage or re-coin.

Ship safely. Pack the miner in its original inner foam if you still have it, otherwise double-box with at least `5 cm` of foam on every side. If shipping hashboards separately, anti-static bag each one. Include a written note with: observed fault string, which fan indices faulted, current firmware version, last maintenance date, and a clear description of any symptoms that don't trigger a logged fault. That note saves diagnostic bench time, which saves you repair dollars.