When an Antminer’s status page shows an ‘X’ next to an ASIC position, that chip is not answering the control board. Each hashboard reports its chips as a grid: a chip that boots, gets clocked, and returns nonces reads as an o (or shows its address/temperature); a chip the board cannot talk to reads as an x. A board that finds fewer chips than it should — a D3 board, for example, carries 60 ASICs — is telling you the serial chain that strings those chips together is broken somewhere. This guide walks the fault from cheapest cause to most expensive, so you rule out power, heat, and connections before you ever suspect silicon.
Why an ‘X’ Appears: How the Chain Actually Works
The ASICs on a hashboard are not wired in parallel. They sit in a single daisy-chained data path — the control board clocks a command down the chain and each chip relays it to the next through the board’s ribbon interface (an 18-pin cable carrying CLK, TX/CI, RX/RO, RST, I2C and low-voltage logic power). Power is delivered per domain, not per chip: groups of chips sit in series across a shared DC-DC converter, so the domain — not any single chip — is the smallest unit the voltage regulator drives.
Two failure patterns follow from this:
- Chips lost from a point onward. If one chip dies so hard it can no longer pass data through, every chip after it in the chain goes dark and reads
x. The board reports a low ASIC count, and the losses cluster at the tail. This is a chain break, and it points at one specific chip (or the joint next to it), not the whole board. - A whole domain drops. If a DC-DC converter or a series group fails, that block of chips loses its core voltage together while the chain around it may still partially enumerate. That is a power-domain fault, not a data fault.
Reading where the X’s fall — a clean tail, a mid-board block, or scattered singles — is the first diagnostic, and it costs nothing but attention.
Tools You’ll Need
- A known-good PSU rated for the miner, on a proper circuit (see the voltage note below)
- Multimeter (DC volts + continuity/diode mode)
- Compressed air or a low-pressure blower, and isopropyl alcohol (99%) for connector cleaning
- Access to the miner’s web UI and kernel log
- Optional but decisive: a thermal camera or IR thermometer to spot a cold (dead) or runaway-hot chip while the board is under load
Troubleshooting Procedure
-
Read the kernel log before touching anything. In the miner’s web UI open the status/kernel log and note the exact ASIC count per chain and where the X’s sit. “Chain 3: ASIC=57, X’s at positions 58-60” is a tail break; scattered X’s across all three chains usually means environment (heat, power, ground), not three simultaneous chip failures. Let the diagnosis follow the pattern.
-
Swap in a known-good PSU. A sagging or aging supply is the single most common cause of chips dropping out under load, because the board only misbehaves once it draws full current. Substitute a proven PSU of the correct rating and retest.
Power note: Bitmain APW-series supplies (APW3, APW8, APW9/APW9+) are rated for 200–240V. They will run on 110/120V, but they deliver only reduced output there, and a single 15A/120V circuit physically cannot supply a large miner’s full draw — starve the board and chips will drop out and read
x. If you are on 120V, confirm the miner’s total draw fits the circuit, or move it to a 240V feed for full rated output. -
Confirm the miner is properly grounded. A floating or poorly grounded frame lets electrical noise corrupt the low-voltage data chain, producing intermittent, wandering X’s. Ground the chassis to a known earth and retest.
-
Reseat and inspect the hashboard connectors. Power the unit fully down first. Unplug and reseat the ribbon/data cable and the board’s power connector at both ends. Look for oxidized pins, bent contacts, or a heat-discolored joint. A single high-resistance pin on the data ribbon breaks the chain exactly like a dead chip does — and it is free to fix. Clean contacts with isopropyl, let them dry, reseat firmly.
-
Rule out heat and dust. Chips throttle, misreport, or drop off the chain when they overheat. Check the board temperatures in the log against ambient, and clear any dust packed into heatsinks and between fins — a clogged board runs hot and sheds chips as it warms. Confirm both fans spin at full RPM and that intake air is cool; a hashcenter that has drifted above spec ambient will manufacture intermittent X’s across every board.
-
Check the network only if the symptom is stalling, not lost chips. Packet loss to the pool causes stale shares and rejected work, not a lower ASIC count — but it is worth clearing while you are in the UI. Go to Network → Diagnostics, enter your pool’s stratum host, and ping it. 0% packet loss with average latency under ~100ms is healthy; heavy loss or spiking max latency points at your network or the pool, not the hashboard.
-
Isolate the chip or domain. If a specific board still shows the same X’s after clean power, good ground, reseated connectors, and normal temps, the fault is on the board. With a thermal camera under load, a dead chip runs cold while its neighbors heat; a shorted one runs runaway-hot. That thermal signature localizes the exact ASIC or domain that needs component-level repair.
A Safety Word on the DC Side
Never open the sealed PSU shell. The real shock hazard in a miner lives inside that case — the mains and PFC bulk capacitors hold a lethal charge long after unplugging. The low-voltage DC busbars and rails feeding the hashboards (roughly 12–21V) will not shock you; their danger is the opposite — enormous current. A slipped tool or dropped ring across a live rail can arc and vaporize metal instantly. Power down and disconnect before you probe, and keep jewelry and stray metal off the board.
How to Confirm the Fix Worked
- Reboot the miner and reopen the kernel log. Every chain should enumerate its full ASIC count with no
xpositions. - Let it run under load for at least 30–60 minutes and re-check — heat-related chain breaks only reappear once the board is hot, so a cold-boot pass alone does not prove the repair.
- Confirm the hashrate has returned to spec and that the accepted-share rate is steady with no climbing hardware-error (HW) counter, which would signal a marginal chip still on the edge.
When to Escalate to Board Repair
If the X’s survive good power, a solid ground, reseated clean connectors, and normal temperatures, you are past field troubleshooting and into component-level hashboard repair — replacing a failed ASIC or the DC-DC parts of a dead domain, which needs hot-air rework, the correct donor chips, and a board diagram. That is a bench job, not a swap-a-cable job. Diagnose it against a fault tree first, then get the right parts or hand it to a repair queue rather than throwing good chips at a bad board.
Related
Walk the symptom through the ASIC fault finder to narrow the cause, source ASICs and DC-DC components from ASIC repair parts, keep your model’s manuals handy for the board layout, and when it’s a bench-level fix, start a repair and let the shop take the board.
