What Your Antminer Is Trying to Tell You

The Machine Keeps a Diary

Every Antminer writes a running log of what it sees: board temperatures, fan speeds, chip counts, voltages, network status. Many owners do not look at it until something breaks. Machines often show warning signs before they fail, sometimes for weeks. Here are the most common messages, in plain English.

The Common Errors, Translated

• Temperature faults. The firmware shuts a machine down when a board sensor passes its limit. When one machine overheats, the cause is often local to that machine: a failed fan, a clogged heatsink, or dried thermal interface material. When a whole row overheats at the same hour each afternoon, that pattern can indicate a site level factor, such as hot air finding its way back to the intakes, although other explanations are possible. Airflow design is covered in Hot Aisle Cold Aisle Containment.
• Temp sensor failures. Hashboards carry temperature sensors, and when the controller cannot read one it will typically disable the whole chain rather than run without that data. A board that hashes but reports sensor errors may need only a sensor repair. A disabled chain can look like a dead board on a spreadsheet when the actual fault may be a $10 sensor.
• Fan lost. The controller expects each fan to report its speed. When the signal drops or RPM falls below threshold, the machine alarms or stops to protect itself. Fans are among the most frequently replaced parts, and many operators treat them as consumables. Manufacturer warranties generally exclude fans.
• Missing chips. A healthy hashboard reports its full chip count. When the log shows a partial count, the board has lost contact with some of its chips. Causes range from a bad solder joint to heat damage, corrosion, or manufacturing defects. Creeping chip loss across many boards can indicate an environmental factor, such as dust, humidity, or corrosion, while single random boards tend to reflect ordinary attrition.
• Power faults. Input voltage out of range can point to the power supply or to the electrical system upstream. PSUs fail on their own, and a cluster of PSU failures after a storm or on one PDU may indicate an upstream factor. More in Mining Electrical Basics.
• Zero hashrate with everything green. This is often a network or pool issue: DNS failures, blocked ports, a dead switch, or a mistyped pool address. The machine may be healthy and hashing with nowhere to send its work. Monitoring hashrate at the pool as well as at the machine is a common way to spot it.

Common Practices at Sites

Errors happen at every mining site. How operators track and respond to them varies with fleet size, staffing, and contract terms. Common practices include:

• Ticketing. Many operators log alarms with the machine, serial number, error, action taken, part replaced, technician, and date. That history can show repeat failures on the same unit.
• Pattern review. Rebooting a machine clears the error but does not explain it. When many machines in one row throw the same error, operators often look for a shared cause such as airflow, power, or network.
• Spare parts. Fans, PSUs, and control boards are commonly stocked on site, since a machine waiting on a low cost part loses hashrate while it waits.
• Log retention. Machine logs can be overwritten or wiped by reboots and reflashes. Some operators export and archive logs on a schedule so the history is available later.

Machine logs and failure patterns are among the records reviewed when questions arise about hosted hardware. For more, see the ASIC hosting expert witness page.