Antminer D9 – Temperature Too High

Controlled cooldown and re-observation Power off the D9 at the PSU switch and leave it untouched for 15 minutes in open air. This lets silicon and PCBs return to a true baseline and gives you time to set up the rest of your diagnostics. After the wait, power on and let the miner run for `10 minutes` while you record per-chain PCB and chip temps on the UI at each minute mark. If temps climb back to trip point inside this window, you have a real, reproducible cooling fault — not a transient. Those per-minute numbers are your baseline for every fix that follows; write them down.

Clean the intake grille, fans, and heatsink fins with compressed air Take the miner outdoors or into a garage — this is messy. Use canned compressed gas or a shop compressor capped at `≤90 psi`. Spray through the intake grille, across both fan blades, and down into the heatsink fins from the top and bottom. Hold each fan blade stationary with a plastic pick while you blow — letting axial fans free-spin off compressed air damages their bearings. Flip the chassis and blow backwards through the exhaust to dislodge the compacted dust that sits at the base of the fin stack. Expect a visible cloud of dust leaving the unit; that's the problem.

Verify ambient at the intake grille itself Place a trusted thermometer directly in front of the D9's intake, `2–5 cm` off the surface, and let it equilibrate for 5 minutes. The reading must be `≤30 °C` for healthy continuous operation, ideally `≤25 °C`. If the room thermostat says `22 °C` but the intake reads `34 °C`, you have hot-air recirculation — the miner's own exhaust (or another miner's exhaust) is being pulled back into the intake. Re-position the miner, add a duct to direct exhaust away from the intake path, or separate stacked units with baffles.

Verify and clear the exhaust path The D9 exhausts out the opposite face from the intake. If exhaust is blowing against a wall, into a corner, into a curtain, or into another miner's intake, back-pressure drops airflow through the heatsinks even with healthy fans at full RPM. Maintain at least `30 cm` of clearance on the exhaust side, and make sure nothing soft (curtains, cardboard, packaging foam, insulation) can be sucked against the intake. Re-observe per-chain temps for 15 minutes after correcting the geometry.

Factory reset from the reset button Per Bitmain's own procedure, the D9 reset button is active only `2–10 minutes` after boot. Power on, wait 2 minutes, then press and hold `Reset` for 5 seconds. The miner will auto-restart after roughly 4 minutes on factory settings. This clears any user-level tuning misconfigurations that could be pushing the chips harder than stock (aggressive autotune profiles, custom frequency tables). Reconfigure pool and worker settings after reset. If temperatures stabilize on stock-vs-tuned, the prior config was the cause.

Replace a failed chassis fan If Step 5 of the Diagnostic Steps identified a stalled fan or a dead `FG` (tachometer) wire, replace the fan. The D9 uses standard Bitmain-spec chassis fans — `12 V`, 4-pin (PWM + FG) connector, high static pressure (Delta, Sanyo, or equivalent OEM-spec unit). Do NOT substitute a quiet PC fan of the same size; the D9 needs static pressure to push air through dense fin stacks, not low-noise airflow. Power off at the wall, remove the top cover screws, unplug the fan from the control-board header, remove the fan's mounting screws, install the replacement in the same airflow orientation (arrow on fan frame points toward exhaust), reconnect, boot, verify RPM on the dashboard.

Reseat hashboard data and power connectors Oxidation on the hashboard ribbon or power connectors produces marginal electrical contact that forces the voltage regulators to push harder to hit target frequency, dumping more heat into the boards. Power off, unplug at the wall (not just the PSU switch), wait 5 minutes for caps to drain. Disconnect and reconnect each ribbon and each power connector on both hashboards. If any contact looks dull or discoloured, clean it with `99%` isopropyl alcohol on a lint-free swab, let dry fully, reassemble. Boot and observe per-chain temps. This is a zero-cost fix that resolves a non-trivial slice of D9 thermal complaints on re-racked or recently shipped units.

Measure PSU voltage at the hashboard connector under load With the D9 running and hashing, carefully probe the DC bus at the hashboard-side power connector with a digital multimeter (Fluke 117 or equivalent). You should see a stable mining-rail voltage (Bitmain APW-family PSUs for the D9 deliver the domain voltage configured by the control board — expect the nominal to be stable with minimal sag). If the rail sags `>5%` under full load, the PSU is tired and is forcing the hashboards to compensate thermally — that sag is the thermal-trip root cause, not the cooling path. Swap in a known-good D9-compatible APW-family PSU and retest. See also [220V-to-110V wrong connection](https://d-central.tech/asic-troubleshooting/antminer-220v-to-110v-wrong-connection/) for the voltage-compatibility trap that mimics a thermal fault.

Reapply the heatsink clamp torque if you have recently disassembled the unit If the D9 has been opened up for cleaning and the heatsink retention hardware was disturbed, uneven clamp force produces uneven thermal coupling and one chip or one edge of a chain will run hotter. Follow Bitmain's torque pattern (diagonal sequence, two passes) to re-seat. If the chassis was shipped between locations recently, this is a common root cause and worth 10 minutes before you escalate to a full repaste.

Swap in a known-good replacement hashboard to confirm board-level failure If Diagnostic Step 4 isolated a bad board, and you have a known-good D9 hashboard on the shelf, install it in the suspect slot and run for 20 minutes. If temps normalize, the original board is your bench project; if temps stay high, the problem is upstream (fans, ducting, or control board). D-Central stocks D9-family spare hashboards intermittently; contact [support](https://d-central.tech/contact/) to check availability before you budget around shipping one in.

Repaste one or both D9 hashboards Highest-value advanced fix on a D9 that's 2+ years into continuous service. Power off, disconnect at the wall, wait 30 minutes. Remove the hashboard from its slot (track connector orientation). Remove the heatsink retention hardware — keep screws, clips, and any insulating shims organized by position. Gently lift the heatsink off the chip side; if dried paste fights you, warm with a hair dryer on low for 30 seconds, never pry with metal. Clean every BM1791 chip top and every mating heatsink contact with `99%` isopropyl alcohol and lint-free wipes until every surface is shiny and residue-free. Apply a rice-grain-size dot of Arctic MX-6 or Thermal Grizzly Kryonaut to each chip. Replace any crumbled or pumped-out thermal pads on the VRM / regulator components with fresh equivalent-thickness pads (measure the originals — typically `1.0 mm` or `1.5 mm`). Reseat the heatsink using the original hardware in the original torque sequence. Reinstall the board, boot, observe. Expect a `5–15 °C` drop on that chain once the new paste has cured under load for 15–30 minutes.

Flash a known-good firmware build or cross-flash DCENT_OS If thermal behaviour got worse after a firmware update, revert. If you want genuine per-chip thermal visibility and tuning control on the D9, cross-flash [DCENT_OS](https://d-central.tech/dcent-os/) — D-Central's own open-source Antminer firmware ([GitHub source](https://github.com/DCentralTech/DCENT_OS)) — which exposes per-chip HW error rate, autotuning, per-domain voltage control, and Stratum V2, all of which make `ERR_TEMP_HIGH` forensics dramatically sharper than stock `bmminer`. Braiins OS+, LuxOS, and Vnish are viable third-party alternatives if you have existing licences, but DCENT_OS is the Mining Hacker default on Antminer hardware — built by plebs, for plebs, open-source. Back up your pool and worker config before flashing anything.

Replace a failed NTC temperature sensor on a hashboard If Diagnostic Step 6 isolated a dead sensor (cold-boot reading stuck at `0`, `-40`, or `255 °C`), this is through-the-board SMD rework. You'll need a hot-air rework station, flux, and a replacement NTC of matching beta value (typically `10 kΩ @ 25 °C`, `B = 3950` — but *measure the footprint and match the part exactly*). De-solder the failed bead with hot air (keep airflow narrow — easy to cook adjacent SMD caps), clean pads, drop the new bead in with tweezers, reflow. This is advanced work. If you are not already comfortable with SMD rework on a mining hashboard, stop and route to Tier 4.

Re-flash the PIC chip on a hashboard reporting `fail to read pic temp` If the log line `fail to read pic temp for chain X` shows before the over-temp trip, the PIC microcontroller on the hashboard has lost its program or communication with the control board has degraded. Bitmain provides a hash-board-code-editor tool that can re-flash the PIC; it's not widely distributed outside the repair community but the procedure is well-documented on Bitmain's community forum. If the PIC re-flash brings the sensor reads back, your "thermal fault" was actually a PIC-comm fault all along — not unusual on Antminer BHB-family hashboards.

Stop and ship when DIY is exhausted If Step 7 of the Diagnostic Steps isolated a single-chip hot spot that didn't move after a clean repaste, if Step 6 showed a sensor-fault pattern and you don't have SMD rework capability, if the hashboard shows discoloration or scorch marks near any specific BM1791 position, or if you've already replaced fans and PSU and the fault persists — you're in chip-level repair territory. D-Central's ASIC repair bench has the test fixtures for the Bitmain 19-series architecture (which the D9 shares), including BM1791-family reflow capability and per-chip diagnostic rigs. Ship the hashboard only (not the full miner) — cheaper freight, faster turnaround. Mark the ticket `Antminer D9 thermal` so it hits the right bench. Canadian inbound means no cross-border RMA runaround. [Book a D-Central repair slot →](https://d-central.tech/services/asic-repair/).