ASIC Miner Error Codes and Basic Troubleshooting for Whatsminer and Antminer

Most miner faults announce themselves twice: once as a symptom you notice on the pool or the dashboard, and once as a code or log line the firmware has already written down. This guide works from both ends. First the symptom side — what to check, in order, for the common complaints — then the code side, for Whatsminer and Antminer separately.

Common issues and where to start

Work through the checks left to right. Most of them are a login and a glance; the point is to rule out the cheap causes before anyone pulls a board.

If the checks point at a board or a PSU, that is a bench job — see the end of this guide. Login and IP problems are covered in more depth in our firmware guide, and fan, cable and PSU swaps in the spare parts guide.

Whatsminer error codes

MicroBT publishes a reference document for its error codes — WhatsMiner Error Code Description — on the Documents Download tab of the Whatsminer download page. The codes appear in two places: the Error Code column in WhatsMinerTool, and the Errors table on the miner's own status page, each with a cause and a timestamp.

Codes cover fans, electrical faults, environment, pool, firmware, PSU, control board and hash boards. The number tells you the category and, where it applies, the slot. A typical log from a unit that arrived on our bench:

Read together, those four lines say: the SM0 board is dead or disconnected, and SM1 has a chip-level fault. That is a two-board repair, not a firmware problem. The slot numbering is explained in our hash board slots guide.

• A 1970-01-02 timestamp means the error was logged before the miner had synced its clock — usually at boot, which is exactly when board-detection errors happen.
• Codes that show up on every miner at once are usually the tool, not the miners: 8000 is an old WhatsMinerTool talking to newer firmware, and 201-8410 after an SD recovery means the firmware needs upgrading. Both are covered in the firmware guide.

Antminer: reading the kernel log

Antminer firmware does not use numbered codes in the same way. The Miner Log (kernel log) tells the story in plain lines, and the ones that matter are the chain-detection lines at startup. Here is the sequence from an S19j Pro, trimmed:

• Initializing the power, please wait, this may take up to 2 minutes — the PSU is being brought up in steps; chain avg vol drop and rise lines follow as voltage settles.
• Chain[0]: find 126 asic, times 0 — board 0 found all 126 chips on the first try. That is what healthy looks like.
• Chain[2]: find 3 asic, times 0 then Chain[2]: find 126 asic, times 1 — board 2 found only 3 chips at first, retried, and found all 126. It is working, but a board that needs a retry is worth watching.
• !!! reg crc error — a corrupted register read on the chain bus. One or two are noise; a cluster of them alongside a short chip count points at the board.
• Chain[3]: find 108 asic, times 0 … 117, times 1 … 109, times 2 — board 3 never found all 126 across three attempts.
• Chain 3 only find 109 asic, will power off hash board 3 — the firmware gives up and shuts the board down.

Which physical board is chain 3 depends on the control board and how the data cables are plugged — see the CH0–CH3 layouts before pulling one.

Antminer: the hardware-error column

On the dashboard, each hash board has a Hardware Error count. Those are ASIC-level malfunctions — hashes that came back wrong. Two rules:

• Uniform counts across all boards are normal. Every board throws some hardware errors over time; if the three numbers are in the same range, nothing is wrong.
• An outlier must be diagnosed. On an L7 we looked at, boards 1 and 2 showed 2,230 and 1,643 errors; board 3 showed 111,500. Its real-time hash rate was still at spec, so nothing on the pool side would have flagged it — but a board throwing fifty times the errors of its neighbours is failing, and it will not stay at spec for long.

That is the case for looking at the dashboard and not only the pool: the hash rate lags the fault.

Keep a record

For every unit you touch, record the serial number, the codes or log lines, what you checked, and what fixed it — a spreadsheet is enough to start; cloud-based tools scale better. Six months in, that record is what tells you a particular batch has a pattern, and it is what a repair centre needs from you to quote accurately.

Codes pointing at a board?

Send us the model, the error table or the chain-detection lines from the log, and the board's serial number. We will tell you whether it is a settings or cable fault you can fix on site, or a board that needs the bench. We repair Whatsminer and Antminer hash boards, control boards and PSUs in Chicago — contact us.