Antminer Won’t Turn On: The Bench Isolation Ladder

Dead miner. No lights, no fans, no web UI. Before you conclude the worst: this symptom has
exactly four suspects — wall power, the PSU, the control board, and a shorted hashboard dragging
the rail to zero — and a ladder that isolates them in about ten minutes with a multimeter and a
screwdriver. Work it top-down; each rung is cheaper than the one below it, and most
“dead” Antminers we see at the bench fall at the first two rungs.

This page is for a unit that shows nothing at all. If the fans spin but the hashrate is
zero, that’s a different fault family — go to
fans spin, no hashrate. For any other symptom,
start at the decision tree.

Tools: a multimeter, a Phillips screwdriver, and ideally a known-good PSU or a second identical
miner. Nothing on this page requires soldering.

Rung 0 — The two traps that look exactly like a dead miner

The 200–240 V APW trap

Bitmain’s APW12 and APW17 power supplies are 200–240 V input
only. Plugged into a 120 V household circuit they do not start — no fault light, no
fan twitch, nothing. A miner that “died” after moving to a new room, a garage, or a
different outlet has very often just moved from a 240 V circuit to a 120 V one. Confirm
the outlet voltage before touching anything else. If 120 V household mining is the goal, that
is a hardware-selection question, not a repair — see
220 V vs 110 V power
derating.

The dual-cord gotcha

17-series and 19-series all-in-one units take two AC cords. With only one
seated, behavior ranges from fully dead to a unit that starts booting and dies at hash-start.
Reseat both cords firmly and check both breakers before opening the case. It reads as too obvious
to check; on a busy bench it is the first thing we check anyway, because it keeps being the
answer.

Rung 1 — Prove the input power

Action: measure AC at the outlet with the multimeter.
You should see your nominal line voltage (120 V or 200–240 V,
depending on the circuit) held steady.
What it proves: a sagging or dead outlet — tripped breaker, overloaded circuit,
failed PDU port — takes the miner out of the suspect list entirely. A breaker that re-trips on
power-up is itself a finding: it points at a short downstream (Rung 4), so don’t keep
resetting it. If several miners share the circuit, check whether the others sagged at the same
time. While you’re there, confirm grounding —
proper
grounding practice — because a floating ground produces exactly the kind of intermittent
faults this ladder can’t catch. Undo: nothing to undo, this is a measurement.

Rung 2 — Prove the PSU

Action: disconnect the PSU outputs from the boards (or unplug the PSU-to-board
connectors on all-in-one units, following the
all-in-one PSU
procedure for your series), power the PSU, and measure its DC output.

You should see:

• APW7 class: 11.60–12.60 V. Below 11.4 V or above
  13.0 V is a failing supply.
• APW121215a (S19j Pro): a shared rail around 13.5 V in normal running;
  the design range is 11.96–15.20 V. This 3068 W supply speaks a framed I2C protocol
  with the control board, so its output setpoint is firmware-commanded — but a supply that shows
  nothing at all with mains present is dead regardless of firmware.

What it proves: in-spec voltage on every output rail moves suspicion to the
boards. Zero output with good AC input is a failed PSU — the single most common root cause we see.
Do not trust the PSU’s own green LED; supplies can light it while delivering out-of-spec
voltage. Full procedure with probe placement:
testing a PSU with a
multimeter; APW-series specifics:
APW9/APW12
problems and
APW PSU not powering on.
Removing the unit for a swap:
PSU
removal guide. On non-integrated setups, follow the
power-on/off
sequence for separate PSUs; for 19-series integrated PSUs the
19-series
power-on/off guidance applies.

The fastest decisive version: if you have a known-good identical PSU, swap it.
Lights back and fans spinning = done, the PSU was the fault. Undo: reconnect the original after
testing; no configuration changes are involved.

Rung 3 — Prove the control board alone

Action: disconnect all three hashboards — data ribbons and
power — leaving only the control board powered (fan headers connected). Power on and watch the
control-board LEDs and your router’s DHCP table.

You should see the control board boot: LED activity, and the miner requesting
an IP address (find it with the
IP Reporter
procedure). Most models will boot to a reachable web UI with zero hashboards attached.

What it proves: a control board that boots alone but not with boards attached
points at a hashboard or its power feed dragging the system down — go to Rung 4. A control
board that stays dark with a proven PSU is itself the fault: see the model-specific pages for
S9,
S19, and
S21 control boards
that won’t boot;
NAND boot-chain
corruption when the board powers but never finishes boot; physical damage at the input stage
is covered in
power-connector
damage. Undo: power off and reconnect the boards — connectors only, no settings touched.

Board-dependent from here. A dead control board’s recovery path depends
on which of the four board families you have — Zynq boards recover over a jumper-selected SD boot,
BeagleBone over the internal SD card, Amlogic over micro-USB OTG (date-locked on recent stock
firmware), CVITEK effectively not at all. Sixty seconds with the
control-board identifier tells you which
procedure — and which spare board — you actually need, then the
recovery playbook takes over.

Rung 4 — Find the offending hashboard

Action: with the control board proven, reconnect hashboards one at a time —
power and data for board 1 only, boot, observe; then board 2 alone; then board 3 alone.

You should see the miner boot normally with the healthy boards and fault (or
die) with exactly one of them. That board is your culprit.

What it proves: a hashboard with an internal short can pull a power rail down
hard enough to look like a dead machine. On the bench, the classic confirmations on the culprit
board are a near-zero resistance reading across its power input (a shorted power MOSFET — measure
across the MOS pins; continuity between any two of pins 1/4/8 means it has failed) and a boost
circuit that no longer produces its 19–25 V output. Those are component-level findings:
route them through the hashboard fault localizer or the
hashboard short-circuit page
for the repair path. Undo: reconnect all boards; isolation changes nothing in firmware.

Where the ladder ends

By this point you have one of four named outcomes: bad input power (fix the circuit), bad PSU
(replace it), bad control board (identify the family, then
recover or
replace
it), or a shorted hashboard (component-level repair). For different symptoms — the miner powers on
but misbehaves — start from the right door of the
troubleshooting decision tree. And if the
fault has gone past multimeter territory, our Montreal bench does
component-level ASIC repair, from the S9 era through current S21
machines.

Why won’t my Antminer turn on after moving it to a new outlet?

Check the outlet voltage first. APW12 and APW17 PSUs are 200–240V-only; a 120V household
outlet won’t start them at all. Also confirm both AC cords are seated on dual-cord models —
a move is exactly when one cord gets forgotten.

Can a single bad hashboard stop the whole miner from powering on?

Yes. A shorted hashboard — typically a failed power MOSFET — pulls the shared supply down
or trips PSU protection, and the whole machine plays dead. The board-only boot test (Rung 3)
followed by one-at-a-time reconnection (Rung 4) isolates the culprit in minutes.

My PSU’s LED is green — doesn’t that mean it’s fine?

No. A PSU can hold its indicator green while delivering out-of-spec voltage. The verdict
comes from a multimeter on the DC output: 11.60–12.60 V for APW7-class units. Measure
before you spend money on any other part.