Repair Guide · Antminer Z15 · BM1746 · Hashboard Repair

Antminer Z15K Hashboard Repair: Chip-Level Guide

The Z15K hashboard is a dense, high-current board that fails in predictable places: input power, the BM1746 chain, and the sense/telemetry path. This guide walks through field diagnosis and chip-level repair, and flags where the Z15 Pro differs despite sharing the same architecture.

By Frontline Hash Technical Team··3 min read
3
Hashboards per Z15K unit
BM1746
ASIC on Z15-series boards

The Antminer Z15K runs three hashboards built around the BM1746 ASIC, the same chip family used across the Z15 line. Boards fail for reasons that are usually visible on a meter before they are visible in firmware: a shorted input stage, a broken voltage domain, or a single ASIC pulling the whole chain down.

This guide covers field diagnosis and chip-level repair on the Z15K, and notes where the Z15 Pro diverges despite sharing the architecture.

What "chip-level" actually means here

Chip-level repair means working at the component and package level rather than swapping whole boards. On a Z15K that typically means:

  • Replacing MOSFETs, inductors, shunts and sense resistors on the power path
  • Reflowing or reballing individual BM1746 packages
  • Repairing the serial chain (TX/RX) between ASICs
  • Replacing control-side connectors, headers and the sense harness

Board-level vs chip-level

A board swap fixes the symptom and hides the cause. If a Z15K board died from an unstable PSU or blocked airflow, the replacement will die the same way. Always confirm the root cause before returning a unit to service.

First-pass diagnosis

Pull the board, log the fault, and work in this order. Do not start with hot air.

  1. Step 1

    Read the log, not the LED

    Capture the kernel log and note which chain and which chip index report first. A single failing chip index usually means that ASIC or its immediate neighbours, not the whole board.

  2. Step 2

    Check the input stage cold

    With power off, measure the input rail to ground. A low resistance reading points to a shorted MOSFET or capacitor on the input side before you touch the ASIC chain.

  3. Step 3

    Verify each voltage domain

    Power the board on a bench supply with current limiting and confirm every domain comes up within tolerance. A domain that sags under load is a power-path fault, not a chip fault.

  4. Step 4

    Scope the chain

    Probe TX/RX at the first and last ASIC. A clean signal at the first chip and a dead one at the last narrows the break to a specific link in the chain.

  5. Step 5

    Isolate the suspect ASIC

    Only after the power and signal paths check out do you reflow or replace a package. Replacing a chip on a board with a bad rail just destroys the new part.

Reading the failure pattern

Chain reports zero hashrate, board powers up

Serial chain break or dead ASIC

Scope TX/RX at first and last chip

Board draws high current immediately

Input MOSFET or bulk capacitor short

Cold resistance to ground on input rail

Hashrate drops progressively, then dies

Thermal degradation or sense drift

Thermistor readings and sense resistor values

One chip index fails repeatedly

That ASIC or its neighbour

Reflow suspect, then re-check neighbours

Board not detected at all

Connector, harness or control side

Continuity on harness and header pins

Power path repair

The input stage takes the most abuse. On a Z15K the high-current path runs through MOSFETs, inductors and shunts that all see continuous load. Look for:

  • MOSFETs with cracked packages or burned legs
  • Inductors with discoloured windings or lifted pads
  • Shunt and sense resistors that have drifted out of tolerance
  • Solder joints on the input connector that have gone cold under thermal cycling

Replace components with equivalents, not "close enough" parts. A MOSFET with the wrong RDS(on) will run hotter and fail again. After any power-path repair, bring the board up on a current-limited supply and confirm the rail holds under load before reinstalling.

BM1746 chain repair

When the power path is sound and the chain still fails, the fault is in the serial link or the ASIC itself.

Reflow vs reball vs replace

Reflow

Single chip with intermittent link, no visible damage

Low; may not hold if the package is genuinely degraded

Reball

Chip works but joints are cracked from thermal cycling

Medium; requires stencil and controlled profile

Replace

Chip confirmed dead after isolation

Higher; new part can be killed by an unfixed rail

Work one chip at a time and re-test the chain after each intervention. If two adjacent chips fail together, suspect the link between them before replacing either package.

Sense and telemetry

The sense path is easy to overlook. A drifted sense resistor or a damaged thermistor will make a healthy board report wrong voltage or temperature, and the firmware will throttle or shut it down. Verify sense readings against a known-good board before condemning the ASIC chain.

Z15K vs Z15 Pro: same architecture, different board

The Z15 Pro uses the same BM1746 architecture and the same general repair workflow, but the boards are not interchangeable. Treat them as siblings, not twins.

ASIC family

BM1746

BM1746

Repair workflow

Same diagnosis order

Same diagnosis order

Board layout

Model-specific

Model-specific

Sense circuitry

Verify against Z15K reference

Verify against Z15 Pro reference

Firmware expectations

Z15K firmware

Z15 Pro firmware

Interchangeable boards

No

No

Shared procedures

Power-path checks, chain scoping and ASIC isolation carry over between Z15K and Z15 Pro.

Model-specific parts

Sense components, connectors and board layout differ. Use the correct reference board for comparison.

Firmware discipline

Never run a board on firmware for the other model, even briefly. Wrong voltage tables damage ASICs.

Preventing repeat failures

A repaired board that goes back into a bad environment will come back. Before returning any Z15K to service:

  • Confirm PSU output is stable under load, not just at idle
  • Check airflow across all three boards and clear dust from heatsinks
  • Verify ambient temperature stays within the miner's rated range
  • Log the repair and the fault so a repeat failure is recognised quickly

When to stop repairing

If a board has multiple clustered ASIC failures, a burned substrate, or repeated failures after correct repair, the economics shift. At that point a replacement board is the honest recommendation, not another reflow.

Chip-level repair on the Z15K is straightforward once you respect the order of operations: power path first, chain second, sense third, ASIC last. The Z15 Pro follows the same logic with its own reference values. Get the diagnosis right and most boards go back into service without a full ASIC rebuild.

Frequently asked questions

Can a Z15K hashboard be repaired at chip level?

Yes. Most failures are power-path or single-ASIC faults that are repairable with reflow, reball or component replacement. Full-board ASIC replacement is only justified when multiple chips fail in a cluster.

Are Z15K and Z15 Pro hashboards interchangeable?

No. They share the BM1746 architecture and similar repair procedures, but board layout, sense circuitry and firmware expectations differ. Do not swap boards between models.

What causes repeated hashboard failures on Z15K?

The usual root causes are unstable PSU output, poor airflow across the board, and dust-driven overheating. Fixing the environment is cheaper than repeated board repairs.

Repair & maintenance

Get Your Miners Back to Full Hashrate

Chip-level diagnostics, component-level repair, and burn-in validation — on-site across the USA, no shipping required.