Antminer S17 – Fan Speed Error

Hard power-cycle at the breaker. Kill the breaker, wait 30 seconds, restore. Clears any wedged firmware state where the PIC lost sync with the SoC and the thermal supervisor fails safe with a phantom fan error. If the error clears and does not return within an hour of hashing, you are done. This is the cheapest and fastest fix and clears a meaningful slice of tickets.

Read the kernel log and identify the failing fan. Pull up the web UI Kernel Log, or SSH in and run cat /var/log/kern.log | grep -i fan. Note which fan index is flagged (fan1, fan2, fan3, fan4). On the S17 chassis, numbering runs front-to-back with intake first. This tells you exactly which physical fan to inspect before you open the lid.

Power off at the breaker and visually inspect the named fan. Remove the top panel (S17, S17 Pro) or the side intake panel (S17+, S17e). Look for a broken blade, a zip-tie tail or cable tangled in the blade path, a cable pulled out of its header, or a dust-felted intake grille the fan cannot pull air through. Most of these are 30-second mechanical fixes — reseat, clear, or de-tangle.

Clean dust with compressed air and an anti-static brush. Blow out the intake grille, blow out the exhaust, and spin each fan by hand to verify it spins freely. Dust-loaded bearings are a common cause of stall-on-startup errors on S17-class hardware that has been in a dusty garage or basement shop for a year or more. Clean thoroughly before you order a replacement fan.

Factory-reset the miner. Hold the IP-report button for 5-10 seconds during the first 2 minutes after power-on to trigger factory reset. Wait for the miner to come back up on DHCP. Firmware regressions and misconfigured fan curves — especially on after-market or custom firmware — can cause phantom fan errors that clear with a reset. This is the last Tier 1 step before you pick up a multimeter.

Reseat the 4-pin fan cable at BOTH ends. Power off, breaker off. Disconnect at the fan and at the control board. Inspect both headers for green oxidation on the pins, bent pins, or burnt plastic. Reseat firmly — listen for the click. Route the cable clear of the fan blade path before you close up. Reboot and watch for the error. This one step alone clears about 30 percent of S17 fan-speed tickets in D-Central's repair queue.

Measure +12V at the control-board fan header under load. With the miner running, carefully probe the header pins with a multimeter on DC. You should read 12V (pin 1) to GND (pin 2). No voltage present means a blown fan-drive FET on the control board — jump to Tier 3 or Tier 4. Voltage present but the fan does not spin means the fan itself is dead — continue to the fan swap step.

Swap the fan cable with a known-good 4-pin PWM extension or a harvested cable from a donor miner. The FG (tach) wire fails open long before the fan itself dies; swapping the cable alone resurrects a surprising number of 'dead' fans. Route the replacement cable through the same clip path as stock so it does not end up in a blade. Reboot and verify.

Replace the failed fan with a 120 x 38 mm, 4-pin PWM, 12V fan. Stock candidates include Sanyo Denki 9GV1224M101, Delta PFB1212GHE, or Nidec V12E12BS4A7-57 on various S17 sub-revisions. Aftermarket replacements work as long as they hit ~5500-6500 RPM at 12V and use a 4-pin PWM connector with an FG tach line. Do NOT substitute a 3-pin fan — the firmware reads the 4th-pin FG signal to detect the fault you are trying to fix.

Swap fans between slots to isolate whether the fault lives with the fan or with the slot. Move the suspect fan to a known-good slot and a known-good fan to the suspect slot. If the error follows the fan, the fan is confirmed dead. If the error stays in the slot, the header or control-board drive circuit is dead and you are in Tier 3 or Tier 4 territory. Five-minute test that saves a $35 fan order.

Cross-flash DCENT_OS, D-Central's own open-source Antminer firmware, for granular fan control, per-fan RPM logging, and per-chip diagnostics. DCENT_OS exposes the per-fan polling loop and gives you a live tach chart — invaluable for confirming whether a borderline fan is marginal or genuinely dead. Alternatives: Braiins OS+, LuxOS, or Vnish. Verify your exact S17 sub-model is on the current support list before flashing any third-party firmware — S17 hardware is at the edge of some support windows.

Upgrade the fan wall to 4-wire PWM high-static-pressure fans. Stock S17 fans are end-of-life by design. Replace both (S17, S17 Pro) or all four (S17+, S17e) fans with longer-life ball-bearing or fluid-dynamic-bearing units rated 60,000+ hours — Noctua industrialPPC, Delta AFC1212DE, or Sanyo Denki 9G-series. Longer life, quieter, cleaner FG signal. This is the canonical Mining Hacker S17 rebuild.

Inspect and replace the control-board fan-drive MOSFET. If Step 7 showed no +12V at the header, the gate-driver FET has failed. On most S17 control-board revisions this is a small SOT-23 or SOT-223 N-channel MOSFET near the fan connectors. Desolder, replace with the same part number or an equivalent — check VDS, ID, and RDS(on) against the original — and resolder with hot air. This is fine-pitch SMD work; ship it if you are not comfortable with a hot-air station.

Reflash the EEPROM using Bitmain's hash-board code editor (documented for S17 series). Some S17 boards accumulate EEPROM corruption that presents as cascading errors with fan-speed as the first flag. BEFORE flashing, dump the existing EEPROM so you have a rollback path. Wrong EEPROM image on the wrong sub-revision bricks the board — verify your exact sub-model before writing.

Run a 30-minute nameplate burn-in after any fan-related repair and monitor per-chain temperatures and HW%. Power on, record temps at 5, 15, and 30 minutes. A healthy S17 at 56 TH/s nameplate runs chain temps in the 65-85 C range. Higher and you have a secondary thermal problem a fan swap did not fix — dried paste, degraded pads, or bent heatsink fins from the overheat event that preceded the error.

Stop DIY when the control-board fan-drive MOSFET is confirmed blown and you lack an SMD rework station, the hashboards show HW% elevation or physical damage from the thermal event, or you have replaced the fan AND the cable AND the error persists on clean factory firmware. You are now in test-fixture territory. Book a D-Central ASIC Repair slot — we have S17-specific bench fixtures, BM1397 chip inventory, and upgraded fan-wall stock.

D-Central bench process for S17 fan-speed tickets: full intake inspection, control-board removal and component-level diagnosis with fan-drive circuit probed under load, fan replacement with upgraded 60,000-hour units, hashboard HW% sweep to confirm no cascading chip damage from the thermal event, reassembly with fresh thermal paste on hashboards if they were cooked, and 24-hour nameplate burn-in with temperature and RPM logging. Standard turnaround 5-10 business days.

Ship the complete miner, not just hashboards, for a fan-error ticket so the bench can see airflow path, control-board fan drive, and hashboards together. Pack in original carton if available, double-box with at least 5 cm of foam on every side. Include a note with observed symptom (ERR_FAN_SPEED on fan N), firmware version, whether the miner ran post-error without airflow, and your contact. Saves diagnostic time and your repair bill.