ASIC Troubleshooting Database

A miner that reports 0 TH/s while still powered usually has a control board that booted but failed to enumerate any hashboards, or a PSU that tripped a protection circuit and stopped delivering high-current power to the boards. Start with a hard power-cycle: kill power for a full 30 seconds (not a quick tap), restore it, and give the unit about 3 minutes to re-enumerate the chip chains — cold-boot glitches clear on their own a surprising amount of the time. If it still reads zero, check that the PSU's main DC connectors are fully seated and that the data ribbon cables between the control board and each hashboard are detected before assuming chip damage.

ASIC chips on a board are wired in a daisy chain, so when one chip dies "open" it stops forwarding the clock and UART signals and every chip after it becomes invisible — the firmware then reports a partial chain or "0 ASIC." The break point is the chip immediately after the last one detected, which is why a board reading "29 of 108" points the repair tech straight at chip 30. First reseat the signal ribbon cable and power connectors and re-test; if the count is still short, the board needs bench diagnosis. We diagnose the exact failed chip and reflow or replace it in our in-house ASIC repair shop rather than guessing.

An over-temperature warning means an on-board sensor (usually the air-outlet sensor) crossed the firmware's thermal limit, most often because of dried thermal paste, a clogged heatsink, or insufficient airflow into the room — not a faulty chip. Power down, clear dust from the fins and intake, and confirm the room is supplying cool air and exhausting hot air separately. If the reading stays high after cleaning, the heatsink may have detached or the thermal interface gel needs reapplying; see the full temperature too high guide before running it again, because sustained overheating cracks solder joints.

A fan-speed error means the control board's tachometer reads a fan as stopped or spinning below its minimum RPM, so the firmware stops hashing to protect the chips from overheating — it is a safety interlock, not the chips failing. Check that both fan connectors are fully plugged into the control board and that nothing is jamming the blades; a single seized or unplugged fan will trip the whole miner. If the connectors are good and the fan still won't spin or reads zero, replace the fan — never bypass the error or blank the tach signal, as the board will overheat. Note a separate PSU fan failure can also halt the unit.

Most "PSU won't stay on" cases are a protection circuit doing its job: the APW-class supplies trip on overcurrent (a shorted hashboard pulling excessive amps), undervoltage (wall voltage sagging below ~180V), or short-circuit, and these hardware protections cannot be disabled in firmware. First confirm you're feeding it the correct input voltage — most high-power Antminer PSUs need a true 200–240V circuit and will not reach full power, or won't start, on a 120V outlet. If voltage is correct and it still trips immediately on connection, suspect a shorted board or a 12V rail sag under load rather than the PSU itself.

A pool connection failure means the miner never completes the TCP handshake with the pool, which is almost always a configuration or network issue rather than a hardware fault. First verify the pool URL is copied exactly from the pool's own documentation — a trailing space, a typo, or the wrong scheme (it must be stratum+tcp://, stratum+ssl://, or a stratum2+ variant) will silently block the connection. If the URL is correct, confirm the miner has a valid IP and can reach the internet (DNS resolution and the pool's port not being firewalled); the full cannot connect to pool guide walks through reading the miner's own logs to tell DNS, timeout, and TLS errors apart.

Low realized hashrate usually means one or more chips dropped out of a chain so the board is hashing with fewer ASICs than nameplate, while a high HW-error rate means chips are returning bad nonces — often from running too hot, an unstable voltage rail, or an over-aggressive overclock. Start with a 30-second hard power-cycle and let the miner re-enumerate, since enumeration glitches clear roughly one time in eight on cold boot. If hashrate stays down or HW errors climb, check temperatures and the power feed before touching tuning — and remember custom firmware autotuner values are calculated at runtime per board, not fixed presets, so a marginal chip will surface here. See the low hashrate guide for the chip-by-chip breakdown.

If the miner powers on but never appears on your network, the control board either failed to boot, didn't pull a DHCP lease, or the Ethernet port itself failed. First scan your router/DHCP table or use the manufacturer's discovery tool to find the assigned IP, and try a different cable and switch port to rule out the network. If it still won't appear, a factory reset via the recessed button often restores a board stuck on bad network settings; a confirmed-dead Ethernet port or a board that won't boot at all needs control-board repair or replacement.

As a rule, repair when the fault is isolated — a single dead fan, one failed hashboard, a control-board Ethernet port, or a handful of bad chips on an otherwise healthy board — because those are economical fixes that restore full hashrate. Lean toward replacement when the machine is several generations old and far less efficient than current models, or when multiple boards have widespread chip failure that costs more to rework than the unit is worth. We've repaired this hardware in-house since 2016, so when you're unsure, a bench diagnosis tells you the real cost before you commit; book it through ASIC repair.

The miner's own status page reports per-chain chip counts, so the chain showing fewer chips than the others is your failed board, and the position of the break tells you roughly where on that board the bad chip sits — the chain goes dark at the chip right after the last one detected. Beyond that, pinpointing the exact chip needs bench tools: a thermal camera shows a dead chip as a cold spot among hot neighbors, while a test fixture re-enumerates the chain and flags the break. On modern boards remember voltage is regulated per voltage-domain (a group of chips), not per individual chip, so an "abnormal domain voltage" reading points to a whole domain's chips or its regulator. This level of fault isolation is what our repair bench is set up for.