Skip to content

We're upgrading our operations to serve you better. Orders ship as usual from Laval, QC. Questions? Contact us

Bitcoin accepted at checkout  |  Ships from Laval, QC, Canada  |  Expert support since 2016

ERR_FAN_SPEED Critical

Antminer S9 – Fan Speed Error

ERROR_FAN_LOST / fan speed below threshold — one or both of the S9's two 120 mm axial fans read 0 RPM or below the firmware's ~1500 RPM floor; the thermal watchdog kills the mining engine before the BM1387 array cooks.

Critical — Immediate action required

Affected Models: Antminer S9, S9i, S9j, S9k, S9 SE

Symptoms

  • `kern.log` shows `ERROR_FAN_LOST: fan X speed 0 rpm` or `fan X speed below threshold` (floor typically `<1500 rpm`)
  • Web dashboard shows `fan1` or `fan2` at `0 RPM` or well below ~4300–6000 RPM nameplate
  • Miner boots, reaches chain-init, then shuts down hashing within 30–120 seconds
  • One fan visibly not spinning while the other runs, or both stopped
  • Fan audibly ticking, grinding, buzzing, or chirping before failure (classic S9-era sleeve-bearing death)
  • Fan spins when finger-flicked but won't start on its own (bearing drag or dust-loaded rotor)
  • Web UI hashrate reads `0 GH/s` while the firmware is still reachable over HTTP/SSH
  • `check_asic_number_with_power_on` lines appearing in `kern.log` after a fan error — boards cooked
  • Exhaust temperature climbs 5–15 °C in the final minute before shutdown
  • Same slot errors again within months of a fan replacement (header, cable, or FET — not the fan)
  • Miner runs fine with lid open and desk fan blowing, errors the moment you reseal the chassis
  • Bearing noise worse on cold start, quiets after warmup (grease migration — terminal)

Step-by-Step Fix

1

Hard power-cycle at the breaker for 30 seconds — not a soft UI reboot. Soft reboots leave the PIC18 supervisory state intact; a hard cycle resets the entire thermal/fan loop. Boot back up and watch for the error in the first 2 minutes. If it doesn't return, you had a transient PIC state issue and no hardware is actually broken. This single step clears roughly 10–15 percent of S9 fan-error tickets with zero parts cost.

2

Open the top lid with a Phillips #2 screwdriver — six screws, thirty seconds. Inspect for: a disconnected 4-pin fan cable, a cable pinched under the lid, a zip-tie tail or foreign object jammed in a fan blade, or obvious blade damage. Clear any obstruction, reseat any loose connector firmly until you hear the click, close the lid, and power back on. Roughly 70 percent of S9 `ERR_FAN_SPEED` tickets are this visible on first inspection.

3

Shop-vac both fans and grilles. Remove fans if possible (four screws each), remove intake/exhaust grilles, wipe fan blades with a dry microfiber cloth, blow out heatsinks and hashboards with compressed air or shop-vac on reverse. Five years of dust is the number-one silent killer on S9-era hardware — it stalls fans, insulates heatsinks, and accelerates bearing wear. Reassemble and test. Plan on this as a 30-minute job done every 90 days as preventive maintenance going forward.

4

Verify ambient intake air temperature with an IR thermometer or a 10-dollar digital thermometer held at the front grille — not room-middle, not hallway. Target is `≤ 30 °C` ambient at the intake. The S9 was designed for ~25 °C data-centre conditions; at 35 °C-plus intake, fan RPM must scale higher to compensate, bearings age measurably faster, and marginal fans fail first in summer. Improve room airflow or move the miner before you blame the fan.

5

SD-card reflash to known-good firmware. Download an official S9 recovery image from `support.bitmain.com/downloads` (or DCENT_OS or Braiins OS for S9), flash to a 1–8 GB SD card with balenaEtcher, insert into the control board SD slot, hold the IP-reset button while powering on until the green LED flashes, then release. Wait ~10 minutes for the reflash. This clears phantom PIC state and firmware corruption — highest-ROI Tier-1 step on any S9 that's been through multiple owners.

6

Measure the fan header `+12V` under load with a multimeter on DC. Probes between pin 2 (`+12V`, red-wire position) and pin 1 (`GND`, black) at the control-board fan header with the fan disconnected and the miner trying to drive fans on. Expect `+11.8–12.2 V` steady. Below that — or a pulsing rail that never settles — means the fan-drive FET or the `+12V` feed is compromised. This 90-second test tells you whether the fault is fan-side or board-side before you spend a dollar.

7

Re-seat and clean every fan connector. Power off at the breaker, wait 60 seconds for caps to bleed, unplug both fan connectors at the control board. Inspect every pin for green oxidation, bent sockets, or stretched contacts. Clean pins with 99% IPA on a cotton swab, let fully dry, reseat firmly. Add a small dab of dielectric grease on reassembly to slow future corrosion — this single preventive step buys years of reliability in humid or dusty environments.

8

Swap the 4-pin fan cable with a known-good harness. Harvest from a dead donor S9 (plentiful on eBay, Kijiji, and local scrap), or buy a generic 4-pin PWM extension cable. Run the new cable, leave the original disconnected in place, reboot. If the error clears, the original cable had either a broken `FG` tach wire or a cracked solder joint at the connector. Crimp failures on S9-era cables are epidemic — don't assume an electrically-continuous cable is mechanically sound.

9

Replace the fan with a compatible unit: `120 × 38 mm`, `12 V DC`, 4-pin PWM, `~4300–6000 RPM`, high static pressure. Proven drop-ins include Delta AFB1212SHE, Sanyo Denki 109R1212H1021, Nidec V12E12BS1M5, or the generic replacements D-Central stocks for S9/L3+/T17 chassis. Orientation matters — S9 fans push front-to-back; the airflow arrow on the fan hub must align with the chassis airflow direction or you've bolted in a very expensive blanking plate.

10

Consider pivoting to a D-Central Bitcoin Space Heater conversion instead of chasing another fan. An S9 with a chronically failing fan wall is a perfect conversion candidate — rehouse in a silent low-RPM chassis, underclock to ~8–10 TH/s, duct exhaust into a room you were going to heat anyway, and turn a dead-end miner into a sats-earning heating appliance. Canadian winters don't care about difficulty adjustments — heat is heat, and this rig's next 5–10 years make economic sense.

11

Flash DCENT_OS or Braiins OS for better fan diagnostics. DCENT_OS — D-Central's own open-source Antminer firmware — gives you per-fan RPM telemetry, adjustable RPM thresholds, API endpoints for Prometheus/Grafana, and the ability to safely override the fan-speed floor during bench diagnostics. Braiins OS for S9 (community-maintained fork) is the alternative. Flash via SD card per step 5. On an S9, third-party firmware matters more for *telemetry* than for overclocking — the BM1387 is past its silicon-lottery window anyway.

12

Replace the fan-drive MOSFET on the control board if step 6 showed `0 V` at the header under load. Identify the part (commonly an SOIC-8 or SOT-23 N-channel MOSFET near the fan header on the S9 control board), read part number from silkscreen or datasheet cross-reference, desolder with hot air at ~330 °C, clean pads with flux and solder braid, reflow replacement. Expect a bench-experienced tech 20–40 minutes. A donor control board from a scrap S9 is often the cheaper path if you're not already set up for SMD rework.

13

Inspect and replace damaged passives in the fan-drive circuit. Look near the fan-drive FET and `PIC18` for cracked MLCC capacitors, discoloured resistors, or a damaged inline fuse. Control boards that have lived in hot environments for years routinely develop cracked caps near power-switching nodes. Replace with equivalents (0603/0805 MLCCs, rated voltage matching silkscreen). Any board with burn marks or brown flux residue around the fan circuit is a Tier-4 send — the damage is wider than visible.

14

Swap the control board with a known-good donor. The cheapest path when fan-drive circuitry is damaged and SMD rework is off the table. Pull a control board from a scrap S9 with confirmed-working fans, swap into the target chassis, re-flash firmware to match the hashboard revision. S9 control boards cross-compatible across S9/S9i/S9j/S9k/S9 SE with minor firmware image differences — verify your firmware matches the hashboard stamp before you power up or you may brick the new board.

15

Refresh thermal paste on hashboards while the chassis is open. Not a direct fan-error fix, but a fan-error shutdown leaves hashboards that were running hot, and 5+ years of dried paste contributes to the thermal margin problem that makes marginal airflow loss trip `ERR_FAN_SPEED` faster than it should. Use Arctic MX-6 or Thermal Grizzly Kryonaut between each BM1387 and its heatsink. Clean old paste with 99% IPA before reapplying. Uniform thin layer, not a glop.

16

Stop DIY on an S9 when: the control board has visible burn damage, you've replaced fans/cables and reflashed firmware and the error persists, or the economic math says scrap-or-convert rather than repair. An S9 is worth 50–200 CAD on the secondary market — bench labour above 200 CAD rarely makes sense unless you're specifically preserving the unit for heating, research, or sentimental purposes. Book a D-Central repair slot or consider the Space Heater conversion path instead.

17

D-Central bench process on S9-class fan faults: intake and exhaust fan replacement with graded aftermarket 120 × 38 mm fans, control-board MOSFET reflow, cable-harness refurbishment, SD-card firmware image verification with DCENT_OS or Braiins OS, 24-hour burn-in at stock voltage and frequency, exhaust-temperature mapping before ship. Typical turnaround 5–10 business days Canada-wide; US and international shipments welcomed. Ship hashboards separately in anti-static bags, double-boxed with 5+ cm of foam.

18

Space Heater pivot path: an S9 with a dodgy fan wall, a tired PSU, or boards that hash but not at full frequency is often better rebuilt as a D-Central Bitcoin Space Heater than repaired as a miner. We rehouse the guts in a silent low-RPM chassis with an insulated exhaust duct, detune voltage and frequency for longevity, and ship you a heating appliance that earns sats for every kWh you were going to spend heating the room anyway. Quote on request; build times vary with inventory.

When to Seek Professional Repair

If the steps above do not resolve the issue, or if you are not comfortable performing these repairs yourself, professional service is recommended. Attempting advanced repairs without proper equipment can cause further damage.

Related Error Codes

Still Having Issues?

Our team of Bitcoin Mining Hackers has been repairing ASIC miners since 2016. We have seen it all and fixed it all. Get a professional diagnosis.