Goldshell – Factory Reset Not Working (RST Button)

Identify your exact SKU from the rear sticker — not the marketing name on the front. KD-BOX, KD-BOX Pro, and KD-BOX II look almost identical from the front but have different reset behaviours. The rear sticker shows the precise SKU (e.g. `Goldshell KD-BOX Pro`, `Goldshell Mini-DOGE II`). Match this SKU to the per-model procedure table — picking the wrong procedure for your variant means your hold never registers and you'll spend an hour fighting a button that's working perfectly fine.

Get a real reset tool. A 1.5 mm SIM-eject pin (the kind that ships with iPhone / Pixel boxes) or a dental pick. The recessed RST button is 4-6 mm deep on most BOX models. A bent paperclip flexes under pressure and registers a tap (sub-100 ms) instead of a hold — the reset daemon ignores anything under the configured hold window. Push straight in, do not angle the pin. Confirm a tactile click before starting any timer.

Note the front-panel LED state before the reset and take a photo. Most Goldshell models do not give explicit reset confirmation feedback — the only way to recognize that something changed is to see the LED state shift between before and after. On chassis models (KD5/6, LT5/6, CK5/6, HS5, KD-MAX) watch for the `LED1` green triple-blink during the hold; on Mini-DOGE family watch for all LEDs to flash red once when the AC-power-cycle reset registers.

Execute the per-model procedure. Set a real timer using your phone stopwatch — counting in your head is wrong. KD-BOX / KD-BOX Pro / KD-BOX II / HS-BOX / CK-BOX / ST-BOX / KA-BOX / AL-BOX II = powered on, hold RST 5 seconds, release. KD5 / KD6 / KD-MAX / LT5 / LT5 Pro / LT6 / LT-LITE / CK5 / CK6 / HS5 = powered on, hold RST 10 seconds, release, watch for LED1 triple-blink confirmation. Mini-DOGE / II / III+ = power down first, hold RST, plug in AC while still holding, keep holding RST for 10 seconds total from AC plug-in. ST-BOX is the only model where RST is on the rear panel (not side or front). Push straight in, hold for the exact duration ±0.5 s.

Wait 60 seconds after release for the miner to reboot. The reboot is silent on most BOX models — no LED ceremony — and lasts 30-60 seconds while the firmware re-runs first-boot config. Do not interrupt the reboot. Do not re-press RST during this window. Do not power-cycle. The reset daemon writes the default config to flash during this period and any interruption can corrupt the write.

Run find.goldshell.com from a computer on the same LAN. Confirm the miner appears with a NEW DHCP-assigned IP (different from the pre-reset IP). If the miner appears with the old IP, the reset did not register — go back to step 4 with a proper SIM pin and a real timer. If find.goldshell.com returns nothing, plug the miner directly into your router with ethernet (skip switches and powerline adapters), cold-boot, wait 2 minutes, check the router DHCP lease table for a new device with a Goldshell MAC OUI (b8:d8:12, 5c:cf:7f, dc:a6:32 are common Goldshell vendor prefixes — confirm against your model).

Login at http://<new-IP> with the default credentials admin / 123456789. Confirm three things changed: the pool config is blank (no pool URL, no worker name, no password), the hostname is the model-default (e.g. `Goldshell-KD-BOX`), and the password is `123456789`. If all three are reset, the procedure worked — change the password to something strong before doing anything else. Goldshell miners with default credentials are scanned by botware constantly.

If the standard procedure fails on a Mini-DOGE, switch explicitly to the AC-power-cycle method. Power down the miner completely. Hold RST with one hand using a SIM pin. With the other hand, plug in the AC adapter while still holding RST. Keep holding for 10 seconds total measured from AC plug-in. All LEDs flash red once when the reset daemon registers the press. The runtime 5-second hold from BOX-series does not apply on Mini-DOGE family — they only read the button during the boot window.

If you're on an ST-BOX and can't find the RST button, look at the rear panel of the chassis. The ST-BOX is the only Goldshell model where RST is on the back, near the ethernet port, recessed under a small label. Push straight in with a 1.5 mm SIM pin, hold 5 seconds, release. Do not confuse it with the power button or the ethernet port itself.

Confirm the reset is registering but the config is corrupted (NVRAM corruption mode). Watch LED1 during the hold (chassis models) or count seconds carefully (BOX models). If you see the confirmation pattern AND the miner reboots AND it comes back with the same config — you're not failing to reset, you're hitting a corrupted NVRAM blob that survives the reset. This is rare (~3% of reset tickets in D-Central's bench data) but real. Do not retry — escalate to Tier 3 SD-card reflash.

Email hello@goldshell.com for the burn-<model>.img recovery image. Subject: `burn-<model>.img firmware recovery request — SN: <serial>`. Include the exact model from the rear sticker, the serial number, and a one-line summary (`reset registers via LED1 triple-blink but pool config persists across reboot — suspect NVRAM corruption, requesting recovery image`). Goldshell's response window is 24-72 hours typically, can stretch to 2 weeks during Chinese holidays. The recovery image is not on GitHub or the official firmware download page — email is the only documented channel. Once you receive yours, archive it locally for future use.

Acquire a quality microSD card. 8 GB or 16 GB, Samsung EVO Plus or SanDisk Industrial — Class 10 / U1 minimum. Avoid no-name cards; cheap cards corrupt during the recovery flash and stack a second brick on top of the first. Total cost: CAD $15-$25. Plus a microSD-to-USB adapter if you don't already have one (CAD $10).

Flash the burn-<model>.img to the microSD card with balenaEtcher. Open Etcher, click Flash from file, select the image Goldshell sent, click Select target, pick the SD card (TRIPLE-CHECK — picking your laptop's SSD ends very badly), click Flash. Etcher auto-verifies the write. A red X on verification means a bad SD card — get a different one. Total time: 2-5 minutes for an 8 GB image.

Insert the SD card into the Goldshell's microSD slot, cold-boot, wait 15 minutes. The SD slot is on the back, side, or under a rubber flap depending on model — early KD-BOX and HS-BOX need 4 Phillips screws to expose it. Power on with SD inserted but ethernet unplugged. The bootloader prefers SD over eMMC and will rewrite the eMMC userspace partition (clearing the corrupted config blob). Watch for LED activity; the recovery flash typically shows a slow blink or alternating pattern for 5-15 minutes, then the device either powers down (LEDs go dark) or reboots into the recovered state. Do not interrupt this. Pulling power mid-recovery produces a brick that requires a control-board swap, not another reflash.

Verify the recovered miner is on a clean configuration. Remove the SD card, cold-boot one more time, run find.goldshell.com to locate the miner's new IP, log in at http://<IP> with admin / 123456789. Immediately change the default password (Settings → System → Password). Check System Info for the firmware version. Run a 24-hour idle burn-in before declaring the recovery successful — if the config corruption was caused by an eMMC wear failure, the new config will corrupt again within 24-48 hours and you'll need Tier 4 hardware repair.

When to stop DIY and ship to D-Central. You are at Tier 4 when: (a) the RST button itself is mechanically broken — no tactile click, button stuck depressed, button rocks loose in the chassis hole (microswitch failure, requires switch replacement); (b) 3+ SD-card recovery attempts across 3 different quality cards all fail to clear the config corruption (eMMC failure, requires control-board swap or eMMC chip replacement); (c) post-recovery the miner cannot accept any config save — every save is ignored — suggesting an eMMC flash-write failure rather than a userspace corruption; (d) you see capacitor bulging, scorch marks, or any physical damage near the control-board flash chip; (e) the miner was port-forwarded to the public internet at any point — the pool-hijack botware can reinstall its config override after every reset until the eMMC is wiped at the chip level. Book through D-Central's ASIC Repair service at https://d-central.tech/services/asic-repair/.