Antminer S21 – Fan Speed Error

Hard-power-cycle the miner at the wall or breaker for 30 seconds. Do not rely on the web UI reboot — it does not clear a wedged fan-monitor state the way a cold power cycle does. Let the miner boot fully and watch the fan widget in the web UI. If all four fans report healthy RPMs within the first 30 seconds of boot and the error does not return, you are done; log the event and monitor.

Open the chassis (two screws per side on the S21 top cover) with the miner powered off. Physically inspect all four fans for visible damage, bonded dust, foreign objects (we have pulled zip-ties and shipping paper out of S21 fan bays), and for any cable pulled taut against a sharp chassis edge. Re-route any pinched cable. Close inspection with the chassis open saves multiple diagnostic hours.

Vacuum the intake grilles and fan blades with a shop-vac or compressed air. Dust-bonded blades drop RPM, trip ERR_FAN_SPEED, and cook the miner — this is the dumbest failure mode and the easiest to prevent. Also clean the exhaust side; even a partial blockage on exhaust raises static pressure and pushes the fans out of their healthy operating band.

Re-seat the 4-pin fan connectors at the control board. Firmly. You should feel and hear each latch click home. Loose headers from transit vibration or prior maintenance are the single most common cause of ERR_FAN_SPEED on an S21 — fixing this takes 60 seconds and resolves most tickets before escalation. Give each cable a gentle tug after seating to confirm it is locked.

Boot the miner with the chassis still open and verify all four fans spin up within the first 30 seconds. Visually confirm each fan is rotating and there is no contact with wires or foreign debris. If all four spin and the error clears in the web UI, confirm the per-fan RPMs are in the 5400-7200 RPM band under load before closing the chassis and bringing the miner back to production.

With the miner running and the chassis open, probe each fan's 4-pin header with a multimeter on DC. Red probe on +12 V (red wire), black probe on GND (black wire). Expect a steady 12 V within 30 seconds of boot. If any channel reads 0 V or unstable, the control board's PWM driver or +12 V rail to that channel is the problem — escalate to Tier 3. A healthy 12 V with a fan still reading 0 rpm points at the fan itself or its FG wire.

Replace the failed fan with a like-for-like unit. S21 uses a 12 V, high-static-pressure 12038 or 14038 axial fan (verify against your specific hardware revision). D-Central stocks replacement fans matched to the S21 family with tested PWM curves and correct FG pulse-per-revolution characteristics. Four screws, one 4-pin connector. Work with the miner fully powered off — 12 V at the connector can destroy an ESD-sensitive FG input if a probe slips.

Verify the replacement fan spins up on boot and the web UI reports a healthy RPM in the expected range (5400-7200 RPM at full load for S21). If the new fan also reports 0 rpm, the root cause is not the fan — return to Diagnostic Step 5 (fan-channel swap test) to isolate whether the problem is the cable, the channel, or the fan-monitor IC on the control board.

Re-crimp or replace a damaged fan cable if the FG signal is absent on an oscilloscope probe but the fan spins visibly. Use a JST SM or XH 4-pin crimp tool — do not twist-and-tape. A poor crimp on a high-RPM tachometer signal produces intermittent faults that chase you for months. D-Central stocks pre-crimped replacement cables matched to the S21 family's pinout and wire gauges.

If the S21 is new and arrived with ERR_FAN_SPEED out of the box, document everything — photographs of the chassis and fans, the exact log excerpt, the serial number, the shipping-box condition — BEFORE you open the chassis. Opening a sealed S21 voids Bitmain warranty coverage on the affected component. Warranty-claim first, repair second; the RMA cost is always zero, which beats any Tier 2 repair estimate.

Flash DCENT_OS (D-Central's own open-source Antminer firmware) for per-fan RPM history, per-chip HW%, tuning, and autotuning on the S21 class. Source: https://github.com/DCentralTech/DCENT_OS. Alternatives: Braiins OS+, LuxOS, Vnish. All four expose richer fan telemetry than stock Bitmain — specifically the per-fan RPM-over-time plot, which reveals whether the faulting fan was marginal for days or weeks before it tripped. This single diagnostic upgrade is how you catch the next ERR_FAN_SPEED before it shuts the miner down.

Run the firmware's fan-test / diagnostic mode (available on DCENT_OS and Braiins OS+). Command each fan to 100% duty for 60 seconds and observe reported RPM. A healthy S21 fan reports 7000+ RPM at 100% duty. A marginal fan reports 5500-6500 but does not fault. A failing fan reports < 5000 or cycles erratically. Document and replace marginal fans proactively — a fan replacement costs ~CAD 45 today; a cooked hashboard after an unattended fault later costs ~CAD 600.

Swap the faulting fan's cable to a known-good channel on the control board (move fan3 into fan0's header, power up, observe). If the fault follows the fan to the new channel, the fan or its cable is the problem. If the fault stays on the original channel with a known-good fan plugged in, the control board's fan-monitor circuit on that channel is damaged — candidate for component-level repair at D-Central's bench (blown SMD fuse, damaged FG input resistor, or PWM MOSFET replacement).

Emergency field-repair option only, for a broken FG wire with a healthy fan on an out-of-warranty miner: a 1 kΩ pull-up resistor from FG to +12 V combined with a 555-timer feeding ~200 Hz into the FG input will silence the fault and keep the miner hashing. WARNING: this disables genuine fan-failure protection. If the spoofed fan dies while the bypass is in place, the miner cooks its hashboards. Use only as a short-term bridge to a scheduled fan or cable replacement — days, not weeks.

Inspect the control board fan-monitor circuit under 10x magnification. Look for: blown SMD fuses on the fan +12 V rails (usually 1206-size, typically 5 A fast-blow near each fan header), damaged FG input pull-up resistors, scorched traces from a prior short or ESD event. A blown fuse is component-level replacement with the same rating. Scorched or lifted traces are bench territory — book a D-Central repair slot rather than risk further damage with a soldering iron at home.

Stop DIY when: the channel-swap test confirms control board damage, multiple fans fault across different channels, you see physical damage to the control board, or the miner is under active warranty. Your path is either a Bitmain RMA (warranty-intact miners) or a component-level repair at D-Central's bench. Book an S21 repair slot at https://d-central.tech/services/asic-repair/. Turnaround is 5-10 business days for fan-circuit-level repairs.

D-Central's bench process: the fan-monitor sub-circuit is isolated on a test fixture, each of the four fan channels is exercised with a known-load jig, blown SMD fuses are replaced, damaged FG input resistors or PWM MOSFETs are swapped, and the board is burned in for 24 hours at full fan load before it ships back. If the control board is beyond component repair (multi-channel damage, SoC-side fault, scorched PCB), we swap to a graded control board from S21-class inventory and re-pair it with the existing hashboards.

Ship safely: any loose fans sealed in anti-static bags, chassis and hashboards double-boxed with at least 5 cm of foam on every side. Include a note with the log excerpt, firmware version, what you already tried, and your contact info — it saves bench diagnostic time, which saves you money. D-Central ships Canada / US / international and gives a straight quote before any work starts. No diagnostic fee on accepted repairs.