Antminer S9 – Temperature Too High

Hard power-cycle the miner. Pull the PSU cord for 30 seconds - not a soft reboot - then power back up. Firmware sensor state can wedge after a thermal trip, and a real power-off clears it. Watch the first 10 minutes of kernel log for the temp fault to recur; if it doesn't, you were done before you started.

Move the miner to cooler or better-ventilated space. If the S9 lives in a closet, stacked under another miner, or in an unvented corner, pull it into open air, aim a box fan at the intake, or duct the exhaust out. Retest under full hash. If chip temps now stay under the fault threshold, ambient was your cause - solve the install.

Blow dust out of every heatsink fin and both fans. Compressed-air can or shop air at 40-60 PSI - not a hair dryer, not a weak blower. Power off, take the miner outside or into a garage, spend a full 5 minutes on it. Blow rear-to-front so you push dust out the intake. This single step fixes 60% of S9 temp faults. Retest under full hash for 30 minutes.

Check any aftermarket intake filters for clogging. Stock S9s don't ship with filters, but home-mining installs often add window-AC-style foam filters to cut shop dust. Clogged filter equals no airflow equals instant overheat. Vacuum or replace the filter, retest. If you don't have filters: skip this step.

Eyeball both fans at full load and read per-fan RPM from the UI. Both fans should be a solid blur and report within 10% of each other - roughly 5500-6000 RPM on a stock S9. A fan visibly slower, wobbling, or ticking is dying. Spin each fan by hand with power off: healthy fans coast 2-3 seconds smoothly; dying ones seize or grind. Replace any bad fan.

Reseat both fan connectors on the control board. Power off, pull the top cover, trace each fan cable to its 4-pin header. Unplug, inspect for pin corrosion or bend, reseat firmly. A loose fan header causes intermittent fan-lost faults that present as temperature faults because one fan isn't actually running at speed. Listen for the click when you reseat.

Swap in a known-good fan if Step 5 flagged one bad. Match voltage (12 V), connector type (4-pin PWM on most S9 revisions), and RPM class (6000 RPM). Generic equivalents work - QFR1212GHE, PFB1212UHE, or any 120 mm 12 V 6000 RPM axial with matching pinout. Mind fan orientation: intake pulls air in, exhaust pushes air out. Retest under full hash.

Measure ambient with an IR thermometer at the intake grille while the miner is running - not room-middle, not the hallway, at the grille. Must read <= 35 C. If higher, solve the room before the miner: duct exhaust out a window, add a basement-to-outside vent, move the miner out of a closed cabinet. Above 35 C intake, no amount of cleaning will save you.

Swap hashboards between the three slots to isolate a bad chain. Power off, label slots 0/1/2 with tape, move the suspect board to a known-good slot. Power on and watch which chain now reports the high temp. If the fault follows the board, the board is your problem - Tier 3 re-paste or swap. If it stays in the slot, suspect control board, ribbon cable, or PSU rail to that slot.

Re-seat every hashboard ribbon cable and power connector. Old S9s accumulate oxidation on the 2x7-pin ribbon connectors. Unplug, inspect for corrosion or blackening, clean contacts with 99% isopropyl alcohol on a lint-free swab, reseat firmly and listen for the click. Corroded ribbons can cause sensor read failures that present as temperature faults.

Flash DCENT_OS - D-Central's own open-source Antminer firmware - via SD card. DCENT_OS gives you per-chip temperature visibility, per-chain tuning, safer undervolting, and modern stratum support, none of which Bitmain stock firmware exposes on an S9. Alternatives if you prefer: Braiins OS+ (S9 support) or any community open-source S9 fork. Let the miner stabilize 20 minutes post-flash, then read per-chip temps. This converts the S9 from a black box into an instrumented machine.

Re-apply thermal paste on every chip of the suspect hashboard. Remove the board, unscrew any heatsink brackets or gently lever heatsinks off the chips. Clean old paste with 99% IPA and a lint-free wipe. Apply a rice-grain dot of Arctic MX-6 or Thermal Grizzly Kryonaut per chip - excess paste insulates, do not glop it on. Reseat heatsinks with even downward pressure, reassemble. Full-board re-paste on an S9 runs 45-60 minutes; expect a 5-15 C chip-temp drop depending on how dried the old paste was.

Re-adhere any loose heatsinks with thermal adhesive - not regular paste. Stock S9 heatsinks are bonded to BM1387 BGA packages with thermal epoxy (Arctic Silver Thermal Adhesive class). Clean old adhesive from chip and heatsink with IPA, apply a thin even coat of fresh thermal adhesive, press the heatsink on with firm even pressure. Cure per manufacturer spec - typically 4-24 hours before powering up. A dropped heatsink on one chip alone can drag the entire chain's reported temp up 10-20 C.

Inspect and, if skilled, replace the chain temperature sensor. If Diagnostic Step 6 isolated a failed sensor path and you have QFP/SMD rework experience, the sensor chip is user-replaceable - typically a small SOT-23 or T451-class thermistor package on the hashboard. Match the part number from the silkscreen before ordering. If you don't have a hot-air rework station and the experience to use it, this is Tier 4 - ship to D-Central instead.

Roll firmware back to a known-good Bitmain build for your exact S9 hardware revision. Some late-era S9 factory firmwares shipped with sensor-polling edge cases that trigger spurious temp faults. Check Bitmain's S9 download archive for an earlier build matching your hardware rev (sticker on the control board). Flash via SD card. Rarely the fix on its own, but a cheap test to rule out a firmware regression.

Stop DIY and book a D-Central repair slot when you see bulging capacitors, cracked MLCCs, burnt silicone/paste smell, PCB discoloration, or per-chain isolation points at the control board or PMIC. You're now in test-fixture territory. D-Central's repair bench has the S9 jigs, BM1387 chip stock, PIC16F1704 reflash tooling, and QFP/BGA rework stations. Book at d-central.tech/services/asic-repair/.

What D-Central does at the bench: full board inspection under microscope, chain isolation with programmable load, sensor and PIC16F1704 replacement, BM1387 chip reflow or replacement with graded-salvage or NOS stock, full repaste with Thermal Grizzly Kryonaut, and 24-hour burn-in at nameplate hashrate in our test chamber before the board ships back. Most S9 thermal repairs complete inside 5-10 business days.

Ship hashboards safely. Remove each board from the chassis, bag individually in anti-static, double-box with 5 cm+ of foam on every side. Include a note with observed symptoms, ambient conditions, firmware version, and any steps you already tried - it saves diagnostic time, which saves you money. D-Central is in Quebec, Canada; we accept ship-ins from across Canada, the US, and internationally.