A non-integrated miner is any ASIC where the power supply is a separate box — an APW-series or third-party PSU wired to the chassis by DC cables — rather than a supply bolted inside the unit. On these machines the order in which you make and break connections matters: get it wrong and you arc a high-current DC lug or work near live mains-side capacitors. The rule is simple — never plug or unplug DC output cables while the PSU has AC applied. AC off first, DC connections last. Below is the safe power-on and power-off sequence, then the full diagnostic tree for a miner that shows no response at all.
The golden sequence: AC last on, AC first off
Every safe procedure on a non-integrated miner follows the same order. Making DC connections on a live supply is what kills lugs, control boards, and occasionally fingers.
Powering on
- With AC disconnected, seat every DC output cable — the hashboard power feeds and the control-board feed — and confirm each is fully home. On an S9 each hashboard has its own paired connectors; on S17/S19/S21-class hardware the hashboards land on a bolted copper bus-bar, so torque the bolts down — a loose bus-bar lug arcs and burns under load.
- Confirm your circuit can carry the load. High-power PSUs like the APW12 and APW17 require 200–240V and will not run on 120V at all. Older APW3/APW7-class supplies will run on 110/120V but at sharply reduced output — enough to brown out an S9 under full hashrate.
- Now apply AC. The control board asserts the PSU enable line and the supply ramps to the firmware-commanded voltage. Watch for the control-board LED and listen for the fans spinning up to full RPM during self-test.
Powering off
- If the miner is hashing, stop it cleanly first — use the firmware’s shutdown/reboot rather than yanking power — so the boards are idle when the rail drops.
- Cut AC at the switch or PDU.
- Wait. The DC rail collapses in seconds, but the mains-side PFC capacitors hold a charge far longer. Only after AC is off and the unit has sat should you break any DC connection.
What is actually dangerous inside a PSU
Know which hazard is real. The DC output side is low voltage — nominal 12V on S9-class (APW3/APW7), 14–21V on S17-class (APW9), 12–15V on S19/S21-class (APW12/APW17). Those rails will not shock you. What they will do is dump enormous current: a dropped tool across a bus-bar vaporizes and throws molten metal. Treat the DC side as a burn-and-arc hazard, not a shock hazard.
The mains side is the shock hazard. After the PFC stage the supply holds roughly 410–420V DC across its bulk capacitors, and those stay charged after you pull the plug. Do not open a PSU case and probe internal boards unless you know how to confirm the caps are discharged. Everything below stays on the outside of the supply — sockets, cords, and DC output lugs — so you never have to.
Miner shows no response at all: full diagnostic tree
Symptom: you power on and nothing happens — fans dead, no network-port LEDs, control-board light off, no display. Work this in order and stop at the first failure you find. You need a multimeter and a known-good spare PSU and power cord if you have them.
- Reseat everything first. Kill AC, then unplug and firmly re-seat every connection in the chain: wall/PDU, C13/C19 power cords, the AC feed into the PSU, and every DC output lead. A large share of “dead” miners are a single connector that vibrated loose. Re-apply AC and retest before going further.
- Verify the wall. With AC on, measure the socket with your multimeter. No voltage — or the wrong voltage for the PSU (a 240V APW12 fed 120V) — means the problem is upstream. Check the breaker/air switch and have an electrician verify the circuit. No meter? Prove the socket with a device you know works.
- Check the power cord. Socket good but miner dead: measure the output end of the power cord under load. No continuity or no voltage at the far end means a failed cord or connector — replace it. Cheap or damaged C19 cords are a common culprit on high-amperage machines.
- Bench the PSU alone. Cut AC. Disconnect every DC output lead from the miner. Re-apply AC to the PSU by itself and watch the internal fan. If the fan never spins, the supply is dead — replace it.
- Test the DC output. Fan spins but you want proof: with the supply running and outputs still disconnected from the boards, measure the main DC output terminals. Expect the nominal rail for that PSU (12V-class, or 14–21V on APW9). No output, or a rail that sags immediately, means a failed supply. Note: if you hard-enable a bare PSU by grounding its enable pin with no control board managing it over I2C, it comes up at its default maximum output — never bench a hashboard on a hard-enabled supply, and don’t rely on that reading as the miner’s normal operating voltage.
- Isolate the control board. Cut AC. Reconnect only the control-board power feed (on an APW12 this is the fixed OUT2 12V lead; on an S9 it is the control-board connector) and leave the hashboards disconnected. Apply AC. If the fans and control-board LED still do nothing, the control board is faulty — that is your failed part.
- Add the hashboards back. Control board lights up correctly: cut AC, reconnect all hashboard DC feeds, and power on. If the miner now drops dead again, you have a short on a hashboard pulling the rail down. That board needs bench repair — do not keep cycling power into a hard short.
If it runs, then quits after a while
A miner that boots fine but dies or restarts after minutes to hours is usually not “dead” — it is a load or thermal problem:
- Undersized PSU. Give the supply headroom: its rated output should sit comfortably above the miner’s full-load draw — roughly a 20% margin. A supply run flat-out against its ceiling will sag or trip under transient spikes. Match the PSU to the model (an APW3 cannot feed an S19).
- Thermal protection. If it drops out when hot, suspect airflow: clogged intakes, dead fans, a hot room, or firmware overtemp cutoff. Clear the dust, confirm all fans spin, and improve the intake air.
Note on hydro and immersion units: the fan and air-intake steps above apply only to air-cooled miners. Liquid-cooled and immersion hardware is fanless and cooled by coolant flow — there a “thermal cutout” points at pump, flow, or coolant temperature, not dust.
Common mistakes
- Making or breaking DC connections with AC still applied — the fault this whole sequence exists to prevent.
- Feeding a 200–240V-only PSU (APW12/APW17) from a 120V circuit, then condemning a miner that simply never powered.
- Reading a hard-enabled bench PSU’s maximum output and mistaking it for the normal operating voltage.
- Cycling power into a shorted hashboard instead of isolating it — you risk taking the PSU or control board with it.
When to escalate
Once you have isolated the failure to a control board or a shorted hashboard, you are past the field-serviceable line. Board-level repair — reballing, DC-DC converter replacement, EEPROM work — needs a bench, hot air, and the right parts. Bag the failed part, note which step condemned it, and send it in rather than guessing at components.
Related: Run a symptom-first diagnosis with the ASIC Fault Finder, source a replacement PSU, cord, or control board from ASIC repair parts, and when a board is condemned, start a repair to get it on the bench.
