The 19-series power supply (Bitmain’s APW12, the brick shipped with the S19, S19 Pro, S19j, S19j Pro and T19) has two AC inlets, not one. Both must be energised for the PSU to deliver full power, and the order you switch them matters. Ideally you power both cords on and off at the same instant. If your setup can’t do that, power on the outer AC1 first, then the inner AC2, and reverse it on shutdown (inner AC2 off first, then outer AC1). Do it wrong and you leave one PFC stage cold-started under load, which is exactly the abuse that kills input stages over time.
Why the 19-series PSU has two inputs
The APW12 is a dual-input, single-output brick. Each of its two trapezoidal AC inlets feeds an independent EMI filter and Power Factor Correction (PFC) stage; both PFC stages then share a common DC bus that drives the LLC converter and the main hashboard rail. Splitting the AC feed in two is how Bitmain gets ~3600 W out of the unit without pushing a single inlet and cord past sane current limits.
- Main output (OUT1): the adjustable 12–15 V rail, up to 233 A, that powers the hashboards.
- Auxiliary output (OUT2): a fixed 12 V / 15 A rail for the control board and fans.
Because the two AC inputs feed separate front ends, running on a single cord is not “half power” — it is an unbalanced, out-of-spec condition. The miner will typically refuse to reach target hashrate, throw hashboard or power faults, or trip out under load. Feed both inlets, always.
Voltage and circuit requirements
The APW12 is a 200–240 V unit. It reaches its rated 3600 W output only on a 200–240 V supply; a running S19 pulls on the order of 3250 W, so a single 15 A / 120 V household circuit cannot supply it regardless of how the PSU is wired. Provision proper 200–240 V circuits, and size the branch breaker and cords to the real draw — roughly 15 A at 240 V, closer to 17 A at 208 V. Use two power cords rated 10 A or greater, one per inlet.
Tools and prerequisites
- Two IEC power cords (C19 to your outlet/PDU) rated ≥10 A, in good condition — no cracked jackets, no scorched or discoloured connector pins.
- A 200–240 V source: a rack PDU with switched outlets, or two properly rated branch circuits.
- The miner powered off and cool before you start plugging.
Power-on procedure
- Seat both cords first, unpowered. Plug each cord fully into its APW12 inlet and into the outlet/PDU before any switch is thrown. A loose or half-seated connector arcs on inrush.
- Energise both inlets together. Switch both cords live at the same moment using a shared PDU switch or breaker. Simultaneous energising lets both PFC stages start together, which is the healthiest case for the input caps.
- If you cannot switch both at once, apply power in this order: outer AC1 first, then inner AC2. Do not leave a long gap between the two — energise the second inlet within a second or two.
- Let the control board boot. Fans spin up, the miner’s status LED comes alive, and the unit begins its self-test. Give it 5–10 minutes to ramp to full hashrate before judging it.
Power-off procedure
- Stop the miner cleanly if you can — use the firmware’s shutdown/reboot rather than yanking power mid-hash, which is easier on both the flash and the power stages.
- Switch both inlets off together at the PDU/switch.
- If you cannot switch both at once, reverse the power-on order: inner AC2 off first, then outer AC1.
- Only then unplug the cords. Never unplug a live inlet under load — break the current at the switch first, then pull the plug.
How to confirm it worked
- Both fans spin up and stay running; no immediate spin-down/spin-up cycling.
- The miner’s dashboard shows both hashboards present and the unit climbing to its target hashrate within the warm-up window.
- No power-related fault codes in the miner log — a PSU running on one live inlet, or an under-volt condition, usually surfaces as a hashboard or “power” fault. Cross-check any code against the ASIC fault finder.
- Input current on each cord/PDU outlet is roughly balanced. A large imbalance means one inlet isn’t carrying its share — recheck that cord and connection.
Common mistakes
- Running one cord. The most common field error. The miner may boot on the auxiliary rail but will not sustain full hashrate. Both inlets, every time.
- Unplugging a live inlet. Break current at the switch first. Hot-unplugging draws an arc across the connector pins and pits them.
- Undersized 120 V circuits. The APW12 is a 200–240 V unit; do not try to force a full S19 onto household 120 V.
- Tired cords and PDUs. Discoloured pins, warm plugs, or a buzzing outlet under load mean a bad connection — replace the cord or move to a known-good outlet before it fails hot.
Safety notes
Never open the PSU shell. The lethal hazard lives inside the sealed case: the mains-side PFC bulk capacitors hold roughly 400 V DC and stay charged after the cords are pulled. There are no user-serviceable parts in there and no reason to open it for a power-on/off task.
On the output side, the 12–15 V DC rail is not a shock hazard — but it can push enormous current. A dropped tool or a slipped strand across the DC output can vaporise metal and throw an arc instantly. Keep the output side de-energised and tools clear whenever you work near the DC lugs.
When to escalate
If the PSU trips repeatedly with both inlets correctly fed on good 200–240 V circuits, hums or clicks without starting, shows scorching around an inlet, or the miner logs persistent power faults after you’ve ruled out the hashboards, the brick likely has a failing input or output stage. That is a bench job, not a switch-order problem — start a repair rather than keep power-cycling a unit that is telling you it’s hurt.
Related: Match your unit against the D-Central manuals library, decode any power or hashboard fault with the ASIC fault finder, and source a replacement brick or connectors from ASIC repair parts.
