Most “dead hashboard” tickets that land on a repair bench are not chip failures at all — they are bad power connections. A copper blade that never fully seated, a 6-pin plug pulled loose when someone carried the miner by its harness, or a busbar bolt left finger-tight. The power path between your PSU and hashboards carries hundreds of amps at low voltage, so any high-resistance joint turns into a heat source, and heat plus current burns copper, connectors, and eventually the board itself. This guide covers how to make those connections correctly across the three schemes Bitmain-class miners use, how to verify them, and when a joint has already done damage.
First, understand what you are connecting
An ASIC PSU (Bitmain APW3/APW7 on S9-class, APW9 on S17-class, APW12 on S19-class, APW17 on S21-class) delivers a low DC voltage at very high current — roughly 12–13 V on S9-generation gear, 14.5–21 V on S17-generation gear, and 12–15 V on S19/S21-generation gear, at 130 A to 260+ A depending on model. Two safety facts follow from that:
- The DC output rails are not a shock hazard. 12–21 V will not hurt you. The danger on the DC side is current: short a busbar with a ring, a screwdriver, or a dropped tool and it will arc, weld itself in place, and vaporize metal instantly. Remove watches, rings, and bracelets before working near live copper.
- The lethal voltage lives inside the sealed PSU shell. The mains-side PFC bulk capacitors hold a dangerous charge long after the cord is pulled. Never open a PSU case. All the work in this guide is external — plugs, blades, and busbars only.
Always power down and unplug the PSU from the wall before touching any connection. Do not hot-plug hashboard power. Make and check every joint cold, then power up.
Voltage note for 120 V shops
Bitmain APW9-class PSUs are rated 200–240 V and deliver full output there. They will run on 110/120 V, but at reduced output — and a single 15 A/120 V circuit cannot supply a large miner’s full draw regardless of the connection quality. The stock APW12 and APW17 on S19/S21-class gear are 200–240 V input only; the documented 120 V path for that class is to swap in an APW3++ rather than run the stock unit off a 120 V circuit. If a miner brown-outs or hash drops under load on 120 V, that is a supply-circuit limit, not a loose connector. (Older APW3/APW7 on S9 run fine on 120 V at their lower wattage.)
Tools you actually need
- Correct hex/Allen or Phillips driver for your busbar bolts — a driver that cams out rounds the head and guarantees an under-torqued joint
- A torque driver if you have one; otherwise “firm and even,” never gorilla-tight (over-torque cracks lugs and strips threads)
- Needle-nose pliers for re-tensioning worn copper contacts
- Isopropyl alcohol + lint-free wipe, and contact/DeoxIT-type cleaner for oxidized copper
- A multimeter (continuity + DC volts) and, ideally, an IR thermometer or thermal camera for the post-power-up check
- Good light and magnification to inspect for burn marks and bent pins
The three connection schemes — know which one your miner uses
Do not generalize across families. Which scheme you are working with depends on the miner generation:
- Per-board wire connectors (4-pin / 6-pin): S9-generation and similar. Each hashboard is fed by its own plug on a wire harness.
- Copper blade plug-and-socket + bolted busbar: S17-, S19-, and S21-generation. Power reaches the board through a copper contact and is clamped by a screwed busbar. The S9 does not use the bolted busbar scheme — that is exclusively the newer generations.
1. Wire connectors with 4-pin / 6-pin plugs (S9-class)
- Check orientation before you push. These plugs are keyed by the shape of the latch and the pin profile — a squared body with a snap latch on one side, trapezoidal versus square pins. Seat them by feel and shape, not by force. Reversing polarity or forcing a plug in backwards will burn the board’s power input the instant you apply power.
- Seat every plug fully until the latch clicks home. An S9-class hashboard typically takes three 6-pin plugs; all of them must be fully inserted. A partially seated pin carries current through a fraction of its contact area, overheats, and discolors.
- Never carry the miner by the wires. Lifting or dragging a unit by its harness loosens, bends, or deforms the connectors and slowly ruins the crimp. Carry by the chassis.
- Inspect the pins and housings for browning or melt marks. A discolored plug has already been running hot — replace the lead rather than re-seating a damaged one.
2. Copper blade plug-and-socket (S17/S19/S21-class)
- Confirm the blade is fully embedded in its slot. Handling and vibration during transport routinely leave a copper blade sitting proud of its socket. Press it fully home, or pull it and re-seat it, until it bottoms out squarely.
- Install the retaining lock/clip correctly. The lock is what keeps the blade from backing out under thermal cycling — do not skip it.
- Re-tension worn contacts. After many insert/remove cycles a socket loses its spring and no longer grips the blade tightly, causing high-resistance contact. Gently close the slot gap with needle-nose pliers until the blade seats with firm friction. Clean any oxidation off the copper first — a dark, dull contact face is resistance you can wipe away.
3. Fixed copper busbar (S17/S19/S21-class)
- Every bolt present, every bolt tight. The busbar carries the full board current through its bolted joints. A missing screw, a cross-threaded screw, or a busbar that does not sit flat against its landing pad concentrates current through less metal and creates a hotspot.
- Mate flat, clean copper. The two faces must sit flush with no gap, no grit, and no oxide film between them. Wipe both faces before bolting.
- Torque firm and even, not maximum. A snug, even joint has the lowest resistance. Over-torquing cracks the lug or strips the thread and can leave you worse off than under-torque.
How to confirm the connection is good
- Cold check first. With the PSU still unplugged, tug-test each plug and blade — nothing should move. Confirm every busbar bolt is tight and no screws are missing.
- Power up and watch the first minutes. Bring the miner up and let it reach hashing load. A good connection stays close to ambient; a bad one heats fast.
- Feel/scan for heat at the joints. Run an IR thermometer or thermal camera over every plug, blade, and busbar bolt. Any connector running noticeably hotter than its neighbors is high-resistance — shut down immediately and re-make that joint. A connector that is too hot to touch is on its way to a burn.
- Look for discoloration. After a shutdown, inspect for browning, melted plastic, or blackened copper. Discoloration means a joint has already been overheating and the affected part should be replaced, not reused.
- Confirm all boards hash. Check the miner’s status page: all three chains should be present and hashing at expected rate. A chain that reads zero or intermittent right after a reconnection points back to that board’s power (or signal) connection.
Common mistakes that kill boards
- Reversing positive and negative on a wired connector — instant, unrecoverable damage on power-up.
- Re-seating a plug or blade that is already burnt/discolored instead of replacing it — the damage is done and it will only get hotter.
- Leaving one busbar bolt finger-tight “to come back to” and forgetting it.
- Hot-plugging power to save a reboot — arc pitting on the contacts raises resistance permanently.
- Working over live copper with rings, watches, or loose tools nearby.
When to escalate
If a board still reads zero after you have verified clean, tight, correctly-oriented power on all three schemes, the fault is likely past the connector — a boost/LDO circuit, the signal ribbon, or the chips themselves. Stop swapping connectors and move to structured diagnostics.
Related
Trace a dead or intermittent board with the ASIC fault finder, source replacement plugs, blades, and busbar hardware from ASIC repair parts, pull the exact model manual from the manuals library, or hand the unit to the bench through start a repair.
