A Whatsminer (MicroBT) M30, M50 or M60 hashboard that reports no chips, a low chip count, or “loss of balance” has almost always lost part of its series-connected ASIC matrix, its onboard regulation, or its board EEPROM. You can localize the fault to a board, a string, or a region with a multimeter and MinerTool, but the actual fix is BGA-level rework on a proper bench.

How a Whatsminer hashboard is built

MicroBT boards look superficially like Bitmain boards but are wired differently, and that difference drives every diagnostic decision. A complete miner is one Allwinner-based control board, three hashboards, an integrated power supply module, and the cooling assembly. Each hashboard carries MicroBT’s own K-series SHA-256 ASICs — an older 12/8nm-class die on the M30 generation, 5nm-class silicon on the M50 and the newest node on the M60.

The defining architectural trait is the series-parallel chip matrix. Where Bitmain splits a board into many small LDO-regulated voltage domains, MicroBT runs its chips in parallel columns of series strings sharing a common board rail (the firmware exposes this as chip_num chips per board and chip_column_num parallel rows — an M20S, for reference, carries 105 chips per board). The practical consequence is identical to every other ASIC: voltage is per-string/per-domain, never per-chip, and one dead chip opens its string and stalls the whole board. There is no partial-board mode.

• Voltage path. The PSU delivers a high-voltage DC rail to the board input; higher-power boards (M30S++, M50/M60) add onboard buck-boost regulation stages before the chip matrix. That rail then feeds the series-parallel string down to per-chip core voltage. Lose the rail, a regulation stage, or balance across the columns and the board goes dark.
• Onboard control. MicroBT hashboards use a dsPIC-class microcontroller for power sequencing and board management (the analogue of Bitmain’s PIC on older boards), plus on-die and on-board temperature sensors read over I2C.
• Board identity. Each board stores its calibration and SKU in an EEPROM — MicroBT’s CHIPSENIMA layout, which on current units is OP-TEE protected. This is why a Whatsminer board cannot simply be cloned or hot-swapped between units the way some boards can.
• Control board. Whatsminer runs an Allwinner SoC (H3 on early M2x, H6 on the M30 generation, H616 on M50/M60) under Linux/OpenWrt with btminer — not the Zynq or Amlogic boards Bitmain uses. A Whatsminer fault tree therefore never maps cleanly onto an Antminer one.

How to diagnose a dead or weak Whatsminer board

MicroBT’s own factory method — btminer’s built-in board self-test reading the CHIPSENIMA calibration data — is the right starting point, and we respect it: it is fast, model-aware, and it is what the firmware uses every boot. Begin there before you ever touch a probe.

What MinerTool and btminer report

Query the miner over MinerTool or the btminer API (TCP 4028 on V2 boards, 4433 on V3). The controller reports detected chip count per board, the SM0/SM1/SM2 board temperatures, and a numeric error code that already narrows the fault for you:

Code / symptom	What it usually means
530 hashboard not found / 0 chips on one board	Dead string, blown regulation stage, broken signal chain, or ribbon/power connector fault
540 chip-ID read failure	Enumeration breaks mid-chain — an open chip or a cold joint on its data/clock/reset output stops the count there
550 bad chips detected	Chips enumerate but fail pattern/frequency — degraded or marginal dies
560 loss of balance	Voltage/current imbalance across the parallel columns — a partial short or weak string
410 hashboard EEPROM read	CHIPSENIMA EEPROM unreadable — board not identified, calibration lost
300-302 / 350 temp sensor / high-temp protection	Failed I2C sensor, dried thermal interface, or a hot-spotting chip
202 / 233 / 263 PSU protection or comms	Power-side fault upstream of the board — clear the PSU before condemning the hashboard

Each of these has a dedicated walkthrough in our troubleshooting library, for example error 530 hashboard not found, 540 chip-ID read failure, 550 bad chips, and 410 EEPROM read. Our ASIC fault finder lets you start from the code or symptom and work toward the board-level cause.

On the bench: voltage and signal-chain checks

With the board off and capacitors discharged, work an unpowered pass first — it is the safest and catches most catastrophic faults:

1. Input and rail impedance. Measure resistance from the board power input to ground. A near-dead short points to a failed regulation stage, a shorted chip, or a cracked decoupling capacitor pulling the rail down.
2. String / domain taps. Compare the voltage taps along the series-parallel columns against a known-good board of the same model and revision. A column reading well below its neighbours is a partial short (failed chip or capacitor); one reading high/open is a broken string or lifted joint. Keep the black probe on a true board ground, never on the heatsink — touching the radiator can short the board.
3. Signal chain. Walk clock, data-in/out and reset along the chain with a meter in diode mode (and a scope where you have one) to find where the chain stops forwarding. Whatsminer does not publish per-pin diode values, so we baseline against our own known-good boards per revision rather than a universal table — the qualitative method is the same one Bitmain’s factory diode/voltage reference formalizes.
4. Thermal cross-check. Under a brief powered test, a cold chip among hot neighbours is a dead chip; a single hot-spotting chip or component is the short. This is faster than chasing one die across 100-plus positions.

Common Whatsminer failure modes

• Dead or degraded chip. ESD, thermal cycling, or a power event opens a die or drifts it off-spec. Opens break enumeration (540); degraded dies pass enumeration but fail as bad chips (550).
• Broken trace or cold joint. Vibration and thermal cycling crack BGA balls and snap signal traces at column boundaries — the classic intermittent board that drops out warm.
• Regulation / balance fault. A failed buck-boost stage or shorted filter capacitor pulls a column down and trips loss-of-balance (560) or a short-circuit fault.
• PSU and connector faults. MicroBT’s integrated PSU modules talk to the control board over a framed serial/I2C protocol; a PSU comms or protection trip can masquerade as a board fault. Reseated, oxidized, or hot-plugged ribbon and power connectors cause the same. Always clear the power path first.
• EEPROM corruption. A lost or mismatched CHIPSENIMA EEPROM (410, type-mismatch) makes a physically healthy board fail to come up.

Hydro and immersion: corrosion and coolant damage

Water-cooled models — M30S+ Hydro, M53, M56, M63, M66 — add failure modes air boards never see. A weeping quick-connector seal, a cold-plate microleak, or condensation under the plate puts coolant onto a live, high-current board. The result is electrolytic corrosion of BGA balls, pads and traces that often outlives the leak itself: you fix the seal and the board still degrades for weeks. Watch for the water-leak alarm, abnormal water-velocity faults, and wrong air/water firmware flags. Corroded hydro boards need full disassembly, ultrasonic cleaning, and inspection under magnification — not a quick reflow — and many are past economical repair.

Component-level repair reality — and where DIY ends

Honest version: most owners should diagnose to the board, then send the board to a bench. Here is what the actual repair involves.

• Reflow vs replacement. Reflow can recover a genuine cold joint, but it does not fix a dead die. A confirmed bad chip means hot-air removal at roughly 350-380°C, wick and clean the pads, reball a known-good chip with 0.4mm solder balls, and reflow on a controlled profile.
• Validate before you solder. A per-chip K-series tester confirms a salvaged or new chip works before it goes on the board — otherwise you can chase the same fault twice.
• Re-test from the start. After any rework, re-run the full board self-test, re-seat the heatsink with fresh thermal interface, and confirm balance across all columns. A board that passes enumeration can still fail under frequency and heat.
• Tooling reality. This needs a hot-air rework station, BGA reballing kit, microscope, a board test fixture, and a known-good reference board per model revision — remember MicroBT ships dozens of board revisions for a single model, each with its own calibration.
• Know when to stop. Multi-chip damage, lifted pads, a failed regulation stage, OP-TEE-locked EEPROM issues, or any hydro corrosion are bench work. Pushing a hot-air iron at a 5nm BGA without the right profile usually turns a one-chip repair into scrap.

Get your Whatsminer board diagnosed by D-Central

D-Central has run in-house, board-level ASIC repair in Laval since 2016. If your board is past the multimeter stage, our bench will find the failing component and tell you honestly whether it is worth fixing: send it in through D-Central ASIC repair. If you are repairing yourself, we stock the parts the job needs — replacement hashboards, ASIC chips, control boards, and other miner parts.

For model context and the per-model repair workflow that complements this board-level guide, see the M30S, M50S and M60S profiles, our Whatsminer M30 repair, M50S repair and M60S repair pages, and the broader M30S hashboard-not-detected walkthrough.