Whatsminer M50S – Hashboard Not Detected

Full power-cycle at the breaker for 60 seconds. Not a web-UI reboot - a genuine AC-off cycle. This is the only way to clear a wedged I2C bus. Observe the WebUI as enumeration completes; note whether the missing chain returns at all or still reports 0 chips. If it clears, monitor for 24 hours - if it returns, you have an intermittent fault and move to Tier 2 anyway.

Apply the current stock firmware via WhatsminerTool V9.0.1. Download the correct image for your exact M50S hardware revision from MicroBT's support portal. Verify the sticker on your unit before flashing - wrong-revision firmware bricks the control board. A clean firmware flash rules out corrupted calibration blocks. Watch the miner.log export after reboot for asic_init_fail or eeprom read fail entries.

Factory-reset via the physical button. Hold the reset button for 5 seconds while powered. This clears the WebUI config and forces a fresh detection pass. If the chain was masked-out by a config artifact from a prior user, this reveals it. Does not reflash firmware, only resets config. If resetting makes the chain reappear, you inherited a partially-configured miner.

Verify ambient and airflow. Target ambient at the miner intake: 5-30 degrees C. Below 5 degrees in a Canadian winter garage is fine but may throw transient codes during cold-start - let the miner warm up for 20 minutes before declaring a fault. MicroBT codes 213, 236-238, 268, 329 often manifest as 'missing chain' when ambient is the actual trigger.

Power off at the breaker. Wrist-strap grounded. Pull the suspect hashboard. Slide the lock, pull straight out - no torque, no leverage. Inspect the adapter-board ribbon for oxidation, bent pins, or stretched contacts. Inspect the copper busbar contact face for discoloration. Clean both contact surfaces with 99% isopropyl and a lint-free wipe. Reseat firmly until you feel the mechanical detent.

Torque the copper busbar screws to MicroBT spec. This is the single highest-hit fix on M50S HB_NOT_FOUND tickets. The busbar carries the high-current DC rail; a loose screw looks fine but fails under thermal expansion as the board heats up. Use a torque screwdriver - over-torque cracks the busbar, under-torque fails under load.

Slot-swap the suspected-bad board. Label slots 0, 1, 2 with tape. Move the missing board to a known-good slot. Power on, observe for 2 minutes. If the fault follows the board, the board is bad (advance to Tier 3/4). If the fault stays in the slot, the control board or adapter is at fault (skip to step 13).

Measure voltage at the adapter-board power rail. With the miner powered and hashing on the two working chains (suspect board removed), measure DC voltage at the adapter-board output pads for the removed slot. Expect the rated hashboard voltage under load (verify against service manual for your M50S revision). If the rail is dead or sagging with no load, the adapter is faulty. If it collapses under the hashboard, the board is shorting.

Inspect for burnt components with magnification. 10x loupe or a phone macro lens. Look for: cracked MLCCs near the PMIC, bulging electrolytics, blackened resistors, lifted BGA corners on any ASIC, scorched traces on the adapter board. Photograph anything suspicious before proceeding - if you ship to D-Central, these photos save diagnostic time and repair cost.

Re-run WhatsminerTool diagnostics. Export miner.log, power.log, system.log, api.log, upfreq_test.log. MicroBT does not document the schema for any of these, but community reverse-engineering maps the common patterns. The asic_init_fail: chain X, chip Y line tells you which chip position dropped enumeration.

Thermal-image the hashboard under brief power. Apply power for 30 seconds max. Scan the full board with a thermal camera (FLIR ONE Pro or better) or IR thermometer on a grid pattern. Healthy chips all run within a few degrees. A cold chip (10+ degrees below siblings) is dead. A screaming-hot chip (15+ degrees above) is shorted - power down immediately if you see this. Mark the cold/hot chip position.

Reflow the cold chip. Remove the heatsink, clean old paste/pad with IPA, apply flux, preheat bottom side to 150 degrees C, top-side hot air at 310-330 degrees C for 30 seconds on target chip. Cool naturally (do NOT quench). Re-paste with Arctic MX-6 or Kryonaut, reassemble. Retest enumeration. Reflow is the lowest-risk chip-level repair on Whatsminer boards but reflowed chips often fail again within 6-18 months.

Adapter-board component replacement. If the fault stays in one slot with two different hashboards, the adapter PMIC or an I2C pullup is suspect. Requires a donor adapter or a soldering-iron + hot-air setup with exact replacement parts. PMIC part numbers are not published by MicroBT; D-Central maintains an internal cross-reference. For DIY, photograph the adapter, identify the suspect IC by package and pin count.

EEPROM reflash (expert only). Codes 410-412 / 420-422 / 440-442 on a board with no visible damage may be recoverable by re-flashing the EEPROM with a matching image. Requires: CH341A programmer with SOIC-8 clip, a confirmed-good image from a donor board of the exact same bin and revision, and the EEPROM pinout on the M50S hashboard. MicroBT publishes none of this. If you have access to a working M50S, dumping its EEPROM is straightforward; if not, ship to D-Central.

Cross-flash custom firmware (M50S does NOT use DCENT_OS). Whatsminer hardware runs Vnish, Asicdip, or HiveOS-compatible firmware - not DCENT_OS (that is D-Central's Antminer-only firmware). Custom firmware on M50S can expose per-chip diagnostics the stock firmware hides, but MicroBT codes 100001-100003 are anti-tamper integrity checks that custom firmware can trigger. Cross-flash at your own warranty risk.

Stop DIY. You are at Tier 4 if: two physical hashboards fail in the same slot; burnt smell or visible scorching; EEPROM reflash required without a donor image; PMIC component-level repair required; thermal imaging isolated more than two dead chips on the same board. Each scenario requires bench equipment (programmable load, test fixture, hot-air, BGA rework, chip inventory) that costs more than the repair ticket.

Ship safely to D-Central. Pack the full miner or just the affected hashboards (if you're confident with disassembly) in original shock packaging. If original packaging is gone, double-box with at least 5 cm of foam on every side, anti-static bags on all boards. Include a note with observed symptoms, exact numeric error codes, firmware version, diagnostic history, and contact info. This note typically saves 30-60 minutes of diagnostic labor.

D-Central bench process. Incoming inspection with photo log, per-hashboard test-fixture run with programmable load, numeric-code capture, thermal-camera sweep, component-level diagnosis (PMIC, cap, MLCC, busbar), chip-level reflow or replacement with graded ASIC inventory, EEPROM reflash with donor-matched image where applicable, post-repair 24-hour nameplate burn-in, return shipping. Turnaround 5-10 business days Canadian, 7-14 US/international.