Antminer S21 Error Codes & Troubleshooting

An S21 throws an over-max-temp error and drops to 0 TH/s when chip or PCB temperatures cross the firmware’s thermal ceiling, almost always because dust has clogged the heatsink fins and fan blades, the thermal paste on the chips has dried out, or intake air is above roughly 30°C. Start by powering down, blowing out the fin stacks and fans with compressed air, and confirming intake temperature is at or below 30°C before restarting. If it still overheats after a paste refresh, a deeper fault may be present — see our S21 temperature-too-high guide.

A fan-speed error means at least one of the four chassis fans is reporting RPM below the firmware’s safety floor, so the miner halts hashing to protect itself. The most common cause by far is a 4-pin fan connector that has jostled loose, followed by dust on the blades, a worn bearing, or a broken tachometer (FG) wire where the fan spins but reports zero RPM. Power-cycle at the wall, reseat all four fan connectors until they click, and vacuum the blades; if a fan still will not report, follow our S21 fan-speed error guide to swap in a matched 12V fan.

When the dashboard shows “0 ASIC” on a chain, the control board cannot enumerate that hashboard’s serial chip chain — usually because the data ribbon is partially unseated, the power-connector pins are darkened from heat, or a single dead chip has broken the chain so every chip after it becomes invisible. Power-cycle, then reseat the ribbon and power connectors with the ZIF latches fully closed; this resolves the majority of cases. If a slot-swap shows the fault following the board, the chain needs bench diagnosis — see our S21 hashboard-not-detected guide or book S21 repair.

The standard S21 should hold around 200 TH/s, so a sustained shortfall usually means a 9-chip voltage domain has dropped offline (on the standard S21 a chip count short of nameplate in a multiple of 9 points to a domain regulator rather than a single chip; the S21 Pro and XP use the BM1370 with different counts), the PSU is sagging under load, or the unit is still on factory TEST firmware earning nothing. A rising HW% above roughly 2% instead points to over-aggressive tuning, heat, or marginal silicon. First hard power-cycle, confirm you are not on Bitmain TEST firmware, and revert any tuning to stock; for per-domain mapping see our S21 low-hashrate guide and S21 HW-error guide.

PSU overcurrent protection (ERR_OCP or ERROR_POWER_LOST) is the APW17 power supply shutting itself down because current demand exceeded its safe limit, after which it tries to recover every 30–120 seconds and trips again. It is triggered by an autotune profile requesting too much power, mains voltage sagging below spec under load, aging bulk capacitors in the PSU, or a shorted chip pulling excess current on a hashboard. Cold-cycle at the breaker, revert to stock settings, and confirm your mains holds 235–245V under full load; if it persists, work through our S21 PSU overcurrent guide.

If the PSU fans spin but the control board never publishes an IP, shows a dark or solid-red LED, or boot-loops every 5–15 seconds, the cause is typically low input voltage, corrupted firmware on the eMMC storage, a dead Ethernet PHY, or a downstream hashboard short dragging the management bus down. Power-cycle for 60 seconds, swap the Ethernet cable and switch port, then disconnect all hashboards and try to boot the control board alone — if it boots, reconnect them one at a time to find the short. For eMMC recovery and UART steps, see our S21 control-board boot guide.

A melting cord or scorched C19/C20 connector is caused by electrical resistance generating heat, almost always from a loose or oxidized connection somewhere in the power path, or from running an undersized cord on the S21’s 20A load. The S21 requires a genuine P13 20A cord with 12 AWG (2.5mm²) conductors and a 105°C jacket — reusing a 10A or 16A S19-era kettle lead is a fire risk. Kill power at the breaker immediately (not the miner switch), let the cord cool, and replace it with the correct rated cord on a dedicated circuit; full safety steps are in our S21 power-cable melting guide.

Many S21 faults are field-fixable: reseating ribbon and power connectors, swapping a fan, refreshing thermal paste, replacing an undersized power cord, and reverting tuning to stock all resolve a large share of tickets without opening the chips. A board belongs on a repair bench when an entire 9-chip voltage domain goes silent, the same chip position fails across two boards, a second BGA reflow fails within 30 days, or you see bulging capacitors or a burnt-component odor — these need a test fixture, hot-air station, and per-chip diagnostics. We have repaired this hardware in-house since 2016; if you have reached that point, book Antminer S21 repair.