Antminer S9 – Hashboard Not Detected

Hard power-cycle at the breaker for 60 seconds — not a soft restart, not the power button. Full breaker-off. This discharges PSU bulk caps and resets the PIC16F1704 on every hashboard, clearing roughly 15% of `find 0 asic` tickets outright. Re-boot and watch the kern.log. If the error returns within 24 hours, continue to the next steps. If it stays clear, monitor for 48 hours — treat as a one-off brown-out and log the event.

Inspect the chassis before opening it: dust blocking intake, shipping damage, scorch marks on the front PSU-side panel. Then open the chassis (Phillips #2, approximately 8 case screws) and confirm every heatsink on the failing board is still bonded. S9s ship with loose heatsinks that drop during transport; a dropped heatsink on a hot chip kills the chain and presents as `find 0 asic`. Lift and tilt the miner — rattling = dropped heatsink.

Re-seat the two data ribbons on the failing chain, at both ends (board side and control-board side). Power off at the breaker first. Press straight down on the connector; do not rock. Check for bent pins and oxidation — clean with 99% isopropyl and a lint-free wipe if needed. This single step resolves approximately 45% of all S9 `find 0 asic` tickets in D-Central's repair queue. Re-boot and check the log before proceeding.

Re-seat the three 6-pin PCIe power connectors on the failing hashboard. Pull each, inspect contacts for blackening (heat damage — connector is done, replace pigtail), bent pins, or green/white oxide. Reconnect firmly, listen for the click. A connector that clicks but pulls out with gentle force is a failed latch — replace the cable; do not zip-tie. Repeat on all three. Any burnt connector is a direct `find 0 asic` cause and a documented S9 defect.

Confirm ambient intake temperature ≤ 30 °C with an IR thermometer at the front grille — not room-center, at the intake. A chain that enumerates cold and drops out warm is thermal-driven, not electrical. S9s were designed for data-centre airflow; stuffy garages or enclosed workshops starve them. Fix airflow (ducted inlet, ambient cooling, or seasonal operation only) before suspecting silicon.

Multimeter the 12 V rail at the failing board's 6-pin connector under load. DC volts, red probe on 12 V, black on ground, miner actively hashing at full power on the two good chains. Target ≥ 12.4 V sustained. Below 12.2 V = PSU sag, burnt connector, or undersized circuit. S9 pulls roughly 110 A at 12 V at stock; any connector or PSU weakness at that current presents immediately as a dead chain.

Swap PSU with a known-good unit — APW3, APW3++, APW7, or quality aftermarket (Delta DPS-2000AB is a community favourite). If the rail was low in Step 6 and reads ≥ 12.6 V with the new PSU, original PSU is tired. Do NOT hot-swap PSU cords on a live PDU — per Bitmain's Do's and Don'ts guide, this kills PSU output caps in seconds and is the most common cause of 'dead S9 PSU' mistaken for hashboard failure.

Swap the suspect data ribbon with a known-good ribbon from another chain. S9 ribbons are $3-5 cables; keep 3-4 spares in the bench drawer. Fault following the ribbon = dead ribbon, replace. Fault staying with the board = board-level issue, continue. This is often the cheapest fix in the tree and is worth trying before any slot swap.

Swap the suspect hashboard between chassis slots. Label the three slots (S9 chains 6, 7, 8) with masking tape. Move the suspect board to a known-good slot, re-boot, watch the kern.log. Fault following the board = bad board, proceed to Tier 3 per-chip visibility. Fault staying in the slot = control-board-side problem, proceed to Step 15 SD-reflash. This isolation step is the single most diagnostic action in Tier 2.

Inspect all 63 heatsink bonds on the failing board. Gently push each heatsink laterally — any that move were never properly bonded or have lost their thermal adhesive. Re-bond with Arctic Alumina Adhesive or equivalent thermal epoxy; do not trust thermal tape on the S9 at this point in its lifecycle (12-18 month failure mode in warm environments). A dropped heatsink on a hot chip is the most common late-life chain-death cause.

Flash DCENT_OS (D-Central's open-source Antminer firmware — the Mining Hackers' option, recommended) for per-chip ASIC enumeration on BM1387. Alternatives: stable S9 builds of Braiins OS+, LuxOS, or older Vnish releases. All expose per-chip visibility that Bitmain stock firmware hides. Let the miner stabilize 10 minutes after flash. Note the exact failing position — chain-head (0) or chain-tail (62) are the most common single-chip kills that present as `find 0 asic`.

Isolate the bad chip via dichotomy / 1V8 technique. Short the RO (reset output) line and the 1.8 V line between chip 31 and chip 32, then re-probe. The failing half contains the dead chip. Repeat on the failing half. Three to four iterations isolate a single chip position. This is the standard S9 bench technique and the key reason many 'dead' S9 hashboards are salvageable at the bench for a $0.50 chip + labour.

QFN hot-air reflow on the identified dead BM1387 chip. Preheat the board bottom to ~150 °C on a preheater. Apply Amtech NC-559-V2-TF flux heavily over the chip. Top-side hot air at 280-310 °C for 30-40 seconds directly on the suspect chip. Let the board cool naturally on the preheater — pulling it off early cracks joints. Reapply thermal interface, reassemble. Do not attempt this without a proper hot-air station; a Weller soldering iron will destroy the board. Practice on a Bitaxe Hex first if you have not reflowed a QFN before.

Reflash the PIC16F1704 if `fail to read pic temp` is the primary symptom. This is not publicly documented by Bitmain but is well-known to repair shops. The PIC firmware can be reflashed with the hash board code editor tool (community-circulated utility — email D-Central repair for guidance; bad flashes brick the PIC permanently). $25 labour at D-Central's bench versus $200 replacement board. This step alone saves hundreds of S9 hashboards per year from unnecessary retirement.

SD-card reflash the control board with a stable S9 firmware image per Bitmain's S9/T9/R4 recovery procedure. Use a fresh 4-16 GB SanDisk (cheap cards die fast on S9 — cumulative log writes). If running custom firmware (Braiins, Vnish, DCENT_OS), flash to that image not stock — the hashboard EEPROM may have been re-signed and stock will not enumerate it cleanly. Fault clears = done, monitor 48h. Fault persists = Step 16.

Stop DIY and ship the hashboard to D-Central. Triggers: per-chip view shows a fully dead chain with no identifiable bad position (boost-circuit fault), PIC reflash fails twice, visible component damage, reflow returns within 30 days, or parts-plus-labour exceeds CAD $150. At CAD $150+ the Bitcoin Space Heater S9 Edition conversion is a better ROI than continued mining repair. [Book ASIC Repair →](https://d-central.tech/services/asic-repair/).

Ship the hashboard safely to D-Central. Anti-static bag, double-boxed with ≥5 cm of foam on every side. Include a note with observed symptoms, firmware version at time of fault, kernel log excerpt (or a photo of the log screen), and your contact info. Diagnostic time is billed hourly — documenting the fault saves diagnostic time, which saves your repair bill. Typical S9 hashboard repair $85-200 CAD, 5-10 business day turnaround, Canada-wide flat-rate shipping, US/international welcomed.

Evaluate the Bitcoin Space Heater S9 Edition as an alternative to repair. An S9 outputs ~4,500 BTU/h at stock (80% of a portable space heater) and continues earning via solo-pool lottery-ticket mining while it heats your garage, workshop, or basement from October through April. D-Central's turnkey conversion is noise-dampened, ducted, and thermostat-integrated. For a Canadian home miner with power costs $0.08-0.15/kWh and heat bills to offset, the space heater math closes in a single winter — something no other S9 repair guide acknowledges.