Goldshell KD-BOX – Not Hashing / Zero Hashrate

Full cold boot. Power off the KD-BOX at the wall or by unplugging the 12V barrel connector. Wait a full 60 seconds (not 5, not 10) to let PSU capacitors drain and force every ASIC register to reset from hardware defaults. Power back on. Watch the fans ramp, watch the green LED come solid, open the web UI, and wait 5 minutes for the hashrate to climb. If the miner was stuck in the 2.1.x warm-reboot hung-state bug or a transient chip-init stall, a cold boot recovers it about 40% of the time. Log wattage, ambient temp, and firmware version before moving on.

Switch the pool to Dxpool Kadena as a control test. Open the web UI, go to Pool Configuration, and screenshot current settings. Set Pool 1 URL to stratum+tcp://kda.ss.dxpool.com:5831, Worker to YOUR_DXPOOL_USERNAME.kdbox01, Password to x. Save. Cold boot the miner. Wait 10 minutes. Check the Dxpool dashboard for your worker. If shares start flowing, your original pool was outside Goldshell's supported stratum whitelist (Dxpool and F2Pool only per Zendesk KB 17169151093657).

Verify the worker name format. KD-BOX cgminer is picky. The worker field must be username.worker with exactly one literal dot, alphanumerics and hyphens only on the worker suffix. No spaces, no special characters, no emojis. Strip any copy-pasted zero-width characters by retyping the worker manually instead of pasting. Save, cold boot, wait 5 minutes, check pool. Invisible unicode characters from PDF copy-paste are a common silent failure.

Check ambient temperature and airflow. KD-BOX units are more prone to cold-start lockup below ~18 C ambient. If your mining space is a cold Canadian basement in January, move the miner to a warmer space for the first cold boot. Confirm the exhaust fans have 15 cm of clearance and the intake isn't choked with dust. A dust-clogged intake starves the hashboards of cooling, which can trigger firmware-level throttle-to-zero on some Goldshell builds.

Factory reset via the web UI. Navigate to Setting, then Reset. Wait for the reboot. Reconfigure pool, worker, and password from scratch. Do not restore from a saved config file. Zero-width characters and stale configuration entries accumulate across firmware updates. A fresh config on known-good firmware is the cleanest Tier-1 recovery. Screenshot your known-good settings after success so you have a text-based reference (not a binary config file) for future resets.

Hard-wire ethernet and disable WiFi for all firmware work. Goldshell officially requires ethernet for firmware upgrades and downgrades. WiFi bricks devices mid-flash at an unacceptable rate. If your KD-BOX is WiFi-only today, run a temporary ethernet drop before touching firmware. This is a non-negotiable rule in the Goldshell Zendesk firmware KB 16805936980633 and the single highest-ROI prevention rule on any Goldshell BOX miner.

Change the default admin / 123456789 password before anything else. The KD-BOX ships with a known default credential and is a documented target for pool-hijack botware if port-forwarded to the public internet. Any miner sitting on a home LAN with default creds is one misconfigured router away from having its hashrate redirected to an attacker wallet. Use the web UI password change or SSH passwd. Mining Hacker sovereignty baseline - non-optional.

Downgrade firmware to the last known-good version. Download KDBox-2.0.6 (or the current known-good for your model - for KD-BOX II consult the Goldshell GitHub firmware tree) from the official Goldshell mirror or the GitHub release. Upload via web UI, System, Firmware Upgrade, over ethernet. Wait the full flash cycle (5-10 minutes). Do not power-cycle mid-flash. After reboot, verify firmware version string. Cold boot. Wait 10 minutes. Hashrate should climb to nameplate. Fixes the 2.1.1 / 2.1.3 cgminer regression in ~70% of field cases.

Log-dive via SSH. On current KD-BOX firmware, SSH is usually enabled. Default creds are admin / 123456789 (change them per Step 7). ssh admin@<miner-ip>. Once in, run tail -n 200 /tmp/cgminer.log and dmesg | tail -n 100. Look for chain X init fail, ASIC detect failed, nonce out of range, and I2C errors. Paste anomalies into the D-Central support Discord or a GitHub issue. The Mining Hacker community will often pattern-match the log to a known firmware bug in minutes.

Swap PSU and 12V cable with a known-good set. Borrow a spare Goldshell brick or use a Mean Well LRS-350-12 with a DC barrel adapter. Confirm the 12V line reads 12.0V plus or minus 0.2V with a multimeter under load. A sagging PSU can cause chip-init failures that look like firmware issues. If the miner hashes with the known-good PSU, RMA or recycle the original brick. Do not daisy-chain PSUs or use non-rated barrel adapters - brownouts cause intermittent chip-init failures.

SD-card recovery flash. If the KD-BOX is stuck on bad firmware and won't accept a web-UI upgrade/downgrade, it's SD-card time. Email hello@goldshell.com to request the burn-kdbox-*.img recovery image (Goldshell does not publish these on GitHub). Write the image to a FAT32-formatted SD card with balenaEtcher or dd. Power off the KD-BOX. Open the case (four screws on the lid). Locate the microSD slot on the control board. Insert the recovery SD card. Power on. The miner auto-flashes - LED pattern documented in the James Chambers recovery guide.

Hashboard visual and thermal inspection. With case open and miner unplugged, inspect both ICT560-class ASICs and surrounding PMICs. Look for scorched pads, lifted components, swollen electrolytic capacitors, and cracked solder joints near thermal-stressed areas. Document with photos for any eventual repair ticket. Reflow a suspect chip only if you have a proper hot-air station and rework experience. Otherwise document findings and ship to a repair shop (D-Central or equivalent).

Hashboard swap between suspect and known-good units. If you have two KD-BOX units, swap hashboards between them (the hashboard-to-control-board ribbon is a small FPC connector - handle with care, it tears easily). Track which miner fails after the swap. This isolates failure to hashboard vs. control board in a single test. This is a diagnostic procedure, not a production fix. Re-seat the original hashboards once you know which side is bad, then plan the replacement.

Replace the control board. If isolation points to control board failure, replacement control boards for KD-BOX are available through Zeus Mining, BT-Miners, and similar parts specialists at approximately CAD $80-150. Swap is four screws and one ribbon cable. Document the firmware version before and after the swap - a fresh control board usually ships with factory firmware, which you may need to re-downgrade to 2.0.6 immediately on first boot before pointing at a pool.

Replace the hashboard / PMIC. A failed ICT560 or PMIC is a full hashboard replacement unless you have chip-level rework capability. Replacement KD-BOX hashboards run CAD $200-350 depending on supplier and revision. At that cost, compare against a used KD-BOX II market price (~CAD $450-700) before committing. For many owners, a used KD-BOX II is the more economical path than repairing an original KD-BOX with a dead hashboard.

Ship to D-Central ASIC Repair if Tier 1-2 fixes fail, SD-card recovery is required but you lack bench tools, the hashboard shows visible damage, or per-chip isolation suggests reflow territory. D-Central offers diagnostic plus repair for Goldshell BOX-series with flat-rate diagnostic fees and Canadian-dollar pricing - zero Canadian competitors for this repair vertical. Book a repair slot at d-central.tech/services/asic-repair/ or contact the team via ticket with your model, firmware version, symptoms, and cgminer log excerpt.