Goldshell – IP Address Not Found / find.goldshell.com

Power-cycle the miner and wait a full `90` seconds before retrying discovery. Unplug AC at the wall for `60` seconds, plug back in, wait `90` seconds for the control board to boot, the network stack to bring up, and DHCP to complete a lease. Retrying `find.goldshell.com` before this window closes is the single most common false-negative cause. Goldshell BOX miners boot in roughly `45-60` seconds; KD5/KD6/LT5/LT6/CK6 units take `75-120` seconds because of their heavier firmware stack. Don't start clicking rescan until the status LED has settled into its steady-state blink pattern.

Move your discovery laptop to the same physical LAN as the miner — wired, not WiFi. `find.goldshell.com` uses browser-side broadcast / mDNS that does not cross VLAN, subnet, or "guest isolation" boundaries. Plug the laptop Ethernet directly into the same unmanaged switch or router port group the miner is on. Turn WiFi off entirely so the laptop can't choose the WiFi interface for its broadcast. This single change resolves the majority of "no IP" false alarms and should be tried *before* any other step. If your home setup keeps WiFi and wired Ethernet on different subnets by default (most ISP-supplied gateways do), this is the only step that matters.

Check your router's DHCP client list directly. Log into the router admin UI (typically `http://192.168.1.1` or `http://192.168.0.1`; see the label on the router for credentials). Find the DHCP leases / connected devices page. Scan the MAC address column for a prefix of `18:2A:7B` (current Goldshell OUI), `5C:CF:7F`, or `24:C9:A1` (early BOX units). Any match = that's your miner's IP. Load `http://<that-ip>/` in a browser — if the Web UI appears, you're done. Set a DHCP reservation on that MAC so the IP won't change across reboots.

Disable browser extensions and retry in an incognito / private window. uBlock Origin, Privacy Badger, NoScript, and most VPN browser extensions silently block the UDP broadcast / mDNS calls `find.goldshell.com` depends on. Open a fresh Chrome or Firefox private window with no extensions active. Load `https://find.goldshell.com` and rescan. If the miner appears, add `find.goldshell.com` to the extension's whitelist for future use. Brave's Shields feature specifically blocks LAN-discovery JavaScript APIs and must be disabled per-site. Safari on macOS has the strictest default and often works better in a clean Chrome profile.

Try the offline `Goldshell_find.exe` / Mac / Linux native build from the Goldshell support portal. Run as admin on Windows (`Right-click → Run as administrator`) or `sudo` on macOS / Linux so raw-socket scans aren't blocked by the OS. The native build bypasses browser sandboxing entirely and routinely finds miners the browser tool misses. On macOS, `System Preferences → Security & Privacy → Local Network` must allow the app.

Run an `arp -a` scan after a `ping` sweep of the subnet. Windows PowerShell: `1..254 | ForEach-Object { Test-Connection -ComputerName "192.168.1.$_" -Count 1 -Quiet }` followed by `arp -a`. macOS / Linux: `nmap -sn 192.168.1.0/24` (replace with your actual subnet) then `arp -a`. Scan the output for MAC prefixes `18:2A:7B`, `5C:CF:7F`, or `24:C9:A1`. This finds the miner even when `find.goldshell.com`, router DHCP UI, and discovery tools all fail — the ARP cache is populated by the raw ping sweep and is network-layer-truth that can't be hidden by a browser.

Set a DHCP reservation on the miner's MAC, then factory-reset the miner to force a fresh lease. From the router admin UI, reserve a known IP (e.g. `192.168.1.50`) to the Goldshell MAC address you found in Step 6 or Tier 1. Save the config. Then on the miner itself, press-and-hold the `RST` button for `5-10 seconds` until LEDs cycle (BOX) or amber (KD6/LT6/CK6). Release. The miner reboots with factory DHCP, grabs its reserved IP, and you bookmark `http://192.168.1.50/` permanently. This also clears any stale static-IP config left over from a previous network.

Try `http://goldshell.local/` via mDNS. On Windows install Apple Bonjour Print Services (provides `mDNSResponder`); macOS and most Linux distros have mDNS built-in via `mDNSResponder` / `avahi-daemon`. Some Goldshell firmware revisions advertise the hostname `goldshell.local` via mDNS. `ping goldshell.local` to confirm resolution; if it returns an IP, browse to `http://goldshell.local/` or the resolved IP. Not all firmware versions expose this — try it but don't count on it.

Swap the Ethernet cable and switch port. CAT5e cables older than `~7` years, CAT6 with kinks / crushed jackets, or switch ports where the LEDs are inconsistent are common silent-failure paths. Swap the cable with a known-good under-`5`-year-old one. Swap the switch port. If possible, swap the switch entirely — a $`30` dumb gigabit switch from your closet is a fine diagnostic tool. This isolates Step 3 failures where the router "sees" the port as link-up but doesn't actually receive packets.

Check Windows / macOS firewall for mDNS / UDP broadcast blocks. Windows: `Control Panel → Windows Defender Firewall → Allow an app` — ensure Chrome / Firefox / Edge are allowed on both Public and Private networks. Start the "Bonjour" service (`services.msc → Bonjour Service → Automatic → Start`). macOS: `System Preferences → Security & Privacy → Firewall → Firewall Options` — allow incoming connections for your browser. Linux: ensure `avahi-daemon` is running (`sudo systemctl status avahi-daemon`). Broadcast-blocking firewalls silently drop the exact packets `find.goldshell.com` needs.

Direct-connect the miner to your laptop with a single Ethernet cable. Disconnect the miner from your network. Connect one Ethernet cable end-to-end between laptop and miner. Manually configure the laptop's Ethernet adapter: IP `192.168.1.100`, mask `255.255.255.0`, no gateway, no DNS. Power the miner. After `90` seconds, try these in order: `ping 192.168.1.1`, `http://192.168.1.1/`, `ping goldshell.local`, `http://goldshell.local/`. Modern Goldshell firmware falls back to `192.168.1.1` as a static-IP self-assignment if DHCP fails for `N` seconds. If Web UI loads, you've confirmed the miner is alive — the fault is entirely in your LAN configuration or a dead switch port. Reconfigure network from the Web UI, then reconnect to the main LAN.

Capture an Ethernet link-state snapshot with Wireshark during miner boot. If Step 11 gets no response, put Wireshark on the laptop's Ethernet adapter, start capture, power the miner, and watch for any Layer-2 frames from the miner's MAC prefix (`18:2A:7B` family). DHCP DISCOVER, ARP Request, or random garbage frames all confirm the PHY is alive. No frames at all for `120` seconds after boot = dead Ethernet PHY. A handful of corrupt frames = marginal PHY or cable issue — swap cable and retry. Wireshark's default "Ethernet" capture filter is all you need.

Serial console into the control board (BOX-series). On older KD-BOX, HS-BOX, ST-BOX, CK-BOX units there's a USB-TTL UART header on the control board near the Ethernet jack. Open the chassis (two Phillips screws on the bottom). Identify TX / RX / GND on the silkscreen (varies by rev — older BOX uses 4-pin `GND/VCC/TX/RX`). Connect a CH340 / CP2102 / FT232 USB-TTL at `115200 baud`, capture the boot log. If the log reaches a login prompt and network-stack-init messages, the control board is alive — the Ethernet PHY or magnetics are the fault and it's a hardware repair. If boot stalls before network init, the control board itself is failing. On newer KD6 / LT6 / CK6 units the serial access is via the CM4 module's GPIO header — the same approach applies but the pin-out is documented in the Raspberry Pi CM4 datasheet, not on the Goldshell silkscreen.

Check for IP collision / duplicate-MAC on the subnet. Rare but real: a static-IP configuration saved on the miner matches the IP another device (NAS, print server, old router) is actively holding on the new network. Symptoms: miner sometimes shows in ARP, sometimes doesn't; connection attempts to the IP occasionally work but are flaky. Resolve by factory-resetting the miner (Step 7) so it pulls a fresh DHCP lease that the router won't hand out to anything else, or by manually assigning an IP from a range you know is unused.

Stop DIY and book a repair when the Ethernet PHY is confirmed dead and the control board tests alive on serial. Ethernet PHY replacement on BOX-series is a hot-air SMD rework job on a small QFN package — not home-bench territory for most owners. D-Central's bench carries graded-salvage BOX control boards for swap-out, or can replace the PHY and magnetics in place. Cost is typically CAD `$95 – $180` depending on which board and whether the magnetics module survived. [Book ASIC Repair →](https://d-central.tech/services/asic-repair/) with a note that serial console boot reached login prompt but Ethernet link never came up.

Stop DIY when the control board fails to reach a login prompt on serial. A control-board boot failure on a Goldshell is a board-swap job, not a component repair — the SoC is BGA and the firmware is locked / signature-checked. D-Central keeps parts-graded BOX control boards and CM4 modules for the newer KD6 / LT6 / CK6 chassis. Typical out-of-pocket CAD `$140 – $280` including testing and burn-in. Ship the entire miner including PSU — a marginal PSU that brownouts during boot can present as a dead control board.

Ship the miner properly. Anti-static bag for each hashboard if you've removed any. Double-box the chassis with `5 cm+` of foam padding on every side. Include the PSU, the Ethernet cable you've been using, a printed note with observed symptoms, serial number, MAC address (if known), router DHCP client list screenshot, and contact information. The MAC and DHCP screenshot alone save D-Central `30-60 minutes` of re-diagnosis at the bench. Canada-wide shipping standard; US / international welcomed.