Antminer – BHB42 BMMiner Version Mismatch

Hard power-cycle at the breaker. 60 seconds off. A soft reboot doesn't always reset `bmminer`'s EEPROM cache. Power-cycle forces a fresh hardware enumeration pass. If the mismatch was a transient `I²C` glitch, you'll know within 2 minutes of boot.

Screenshot everything before you change anything. Web UI → Miner Status, Configuration, Kernel Log, System Info. Save the `BHB42xxx` model string, the `bmminer` version, and the pool configuration. You'll need these to pick the right firmware download later, and when you eventually file a warranty or repair ticket they prove the as-received state.

Check the pool field. If the pool URL is locked to `TEST`, stop mining right now and do not let it run another minute — you're producing work for Bitmain's QA pool. If the pool is your real pool but hashrate is 0 or nonsensical, the version mismatch is blocking work submission, not the pool config.

Confirm the `BHB42` silkscreen. Power off, open the chassis lid, photograph the control-board silkscreen. The revision (`BHB42601` vs `BHB42602` etc.) matters for firmware selection. Write it on a sticky note taped to the chassis — you'll reference it a dozen times during this repair.

Download the exact-match firmware. From `support.bitmain.com/downloads`, find the firmware listed specifically for your `BHB42xxx` control-board revision. Confirm the published SHA-256 matches your downloaded file before flashing. Do not use a generic S17 blob on a `BHB42602` board and hope for the best — Bitmain's signature check will reject it or, worse, brick your control board mid-flash.

Format a microSD as FAT32. Use a fresh SD card, not one that has been in a camera or phone. Single FAT32 partition, no other files. Copy the firmware image to the root as Bitmain's documentation specifies (usually `download.bin` or similar — check the firmware's README).

SD-card recovery boot. Power off. Insert the SD card. Hold the IP button (or the recovery button per your model's rear-panel documentation), then apply power. Keep the button held for ~5 seconds after the power LED comes up. The miner's activity LED will flash a distinct pattern during reflash — typically 5-10 minutes. Do not power off during this window even if it looks frozen.

Wait for first boot and reconnect. After the flash, the miner reboots into the newly installed firmware. Give it 3-5 minutes to settle. Reconnect to the web UI, confirm the model banner now reads the real model name (e.g. `ANTMINER S17 Pro`) and that `bmminer` version matches the published firmware build.

Reconfigure the pool. Set your real pool and wallet. Confirm save persists. Watch 10 minutes of hashing — the version mismatch message should be gone from the status banner, hashrate should climb to nameplate, and `kern.log` should be clean.

Factory-reset if SD flash didn't take. Some late-rev `BHB42` boards respond better to: flash via SD, then immediately factory-reset via the hardware button (hold 10+ seconds while powered on). That forces the new firmware to wipe stale EEPROM cache on the control board. Reconfigure from scratch after reset.

SSH recovery flash. If the IP button and SD card both fail, SSH in (`root`/`root` on unsecured stock) and run the firmware upgrade from shell. Use Bitmain's official `upgrade` script where available, or `dd` the firmware image to `/dev/mtdblock0` if you know what you're doing. Wrong partition = bricked control board, so this is not a first-try move.

Cross-flash DCENT_OS. Bitmain's stock firmware hides the problem you're trying to debug. [DCENT_OS](https://d-central.tech/dcent-os/) — D-Central's own open-source Antminer firmware — exposes per-chain `bmminer` logs, per-chip EEPROM reads, and auto-detection of control-board revision against hashboard hardware version. It'll tell you explicitly if the pairing is wrong rather than silently falling back to `BHB42xxx`. [Source on GitHub.](https://github.com/DCentralTech/DCENT_OS) Alternatives: Braiins OS+, LuxOS, Vnish — all expose deeper diagnostics than stock.

Reflash the hashboard PIC. With a Bitmain hashboard test fixture + the appropriate PIC programmer, pull the PIC's current firmware, compare against the version the control board expects, and reflash if they don't match. This is vendor-tool territory; most home bench operators ship this out to D-Central. PIC chip replacement (PIC16F1704 / 1708) is $3-$8 in parts and a steady hand.

EEPROM rewrite from backup. If the hashboard EEPROM is wiped but you have a known-good EEPROM image for the same `BHB42` board revision, the fix is an I²C programmer + the original EEPROM image. D-Central keeps a library of EEPROM images for common `BHB42` sub-revisions — include your exact silkscreen in any repair ticket to speed triage.

Control-board eMMC full reflash. If SD recovery fails and the board won't boot SSH, the control-board eMMC itself is corrupted. Pull the SoC out via JTAG or eMMC programmer, reflash the full Bitmain production image for the `BHB42` revision, reseat. This is a 2-4 hour bench job.

When to stop DIY. If EEPROM is wiped and you don't have an image library, if the PIC chip is dead and needs replacement, or if the control-board eMMC is corrupt and SD recovery won't take — you're in test-fixture territory. [Book a D-Central ASIC Repair slot.](https://d-central.tech/services/asic-repair/) BHB42-series boards are in our wheelhouse; we maintain EEPROM image archives and PIC firmware libraries for this generation.

What D-Central does at the bench. EEPROM read/write via dedicated I²C programmer, PIC firmware reflash via Microchip PICkit or equivalent, control-board eMMC reflash via UART+JTAG where needed, signature verification against the correct `BHB42` sub-revision, and a 24-hour burn-in under nameplate load before the unit ships back. Canada-wide shipping, US and international welcomed.

Ship safely. If only the hashboards are affected, ship just the hashboards — it's cheaper and safer than shipping the whole chassis. Anti-static bags, double-box, 5 cm foam on every side. Include a note with: observed symptoms, the `BHB42xxx` silkscreen string, `bmminer` version if you could read it, and your contact. Every minute of bench diagnostic time you save translates to dollars off the final invoice.