Goldshell LT5 / LT5 Pro – Hashboard Offline or Dead

True cold boot. Power off at the wall (not at the PSU brick) for a full 120 seconds. This drains the VCORE caps and clears any stuck DDR / network state on the control board. Re-energise. Watch the dashboard chip count for 5 minutes. Expected: full nameplate count returns (162 on LT5 / LT5 Pro). If the count is still wrong, proceed to Tier 1 reseat. About 25% of LT5 hashboard-offline tickets resolve on the cold boot alone.

Open the chassis with a Phillips #2 screwdriver. With the LT5 fully unplugged, unplug the flat data ribbon between the control board and each hashboard, then firmly reseat each end. Unplug and reseat both PSU output cables on each hashboard input. Inspect both connectors for darkening, bent pins, or visible corrosion. Reassemble, boot, and recheck the chip count. Expected: full chip count returns. If still offline, proceed to Step 3.

Visual inspection of the affected hashboard with the chassis open. Look for burnt components (black soot around an ICT560 or near a VCORE MOSFET), swollen capacitors, discoloration around chip positions, or any obvious physical damage. Photograph any findings for the repair file. Expected: no visible damage. If you find visible damage, jump straight to Tier 3 / Tier 4 — there's no point continuing the diagnostic, you already know the failure is hardware. If clean, proceed to Step 4.

Hashboard slot swap test (Tier 2 diagnostic). Pull both hashboards. Reinstall Hashboard A in Slot 2 and Hashboard B in Slot 1. Reboot. Watch which slot reports offline. Case A: the offline status follows the physical hashboard - the hashboard is bad, plan a Tier 3 repair on that board. Case B: the offline status stays with the slot - the control board or harness is bad, plan a Tier 4 control-board swap. This 15-minute test is worth running before any parts purchase.

Cable substitution (if you have a spare LT-series ribbon, or can harvest one from a sibling miner). Replace the data ribbon to the failing slot with a known-good ribbon. Reboot. Expected: chip count returns. If the chip count returns with a fresh ribbon, you've isolated to a bad cable - a CAD $15-$35 fix. If the chip count is still wrong with a new ribbon, the control-board interface itself is at fault and you're heading to Tier 4.

Cold resistance check on the failing hashboard's VCORE input pad. With the board fully powered down and unplugged, set a multimeter to ohms and probe the VCORE input against ground. Expected: a finite resistance in the kohm range. A reading of less than 1 ohm to ground indicates a shorted MOSFET on the VCORE switching stage - DO NOT power the board up again, that's a guaranteed second failure. A reading of OL / open-circuit indicates a lifted joint or a blown input fuse. Either result routes to Tier 3 component repair.

Per-chip HW% capture (advanced - requires a bench harness or a working LT5 chassis you can swap into for diagnostics). With the suspect hashboard installed, capture the per-chip hardware-error rate during a 15-minute hashing run. Expected: one or two specific chip positions show 100% HW% while the rest are at 0-2%. A single bad position means a chip-level reflow or replacement (Tier 3, single-chip repair). A contiguous block of dead chips means a VCORE domain failure (Tier 3, regulator repair). Document the chip position - the heatmap pays back over time.

Firmware sanity check. From the web UI, confirm the LT5 is on a Goldshell-released firmware for the correct model variant (not a sibling product's firmware - a non-trivial number of LT5s come in with LT-LITE or LT6 firmware accidentally flashed). Expected: about page reports the correct model and a current firmware version. If wrong firmware is installed, re-flash the correct LT5 firmware via Goldshell upgrade flow over wired ethernet (never WiFi). Reboot and recheck the chip count. Note: firmware fixes almost never resolve a true hardware hashboard-offline event.

Tier 3 - chip-position reflow. Pull the failing hashboard and place it on the bench. With a hot-air rework station (Quick 861DW, JBC HA-T, or equivalent) at a 350 C profile, flux-paste applied around the suspect ICT560 chip, run a controlled 8-10 minute reflow cycle. Allow the board to cool fully before touching. Reinstall in chassis, boot, and capture per-chip HW% again. About 30% of single-chip failures recover from a clean reflow alone. If the chip-position is still 100% HW% post-reflow, the chip is dead and needs replacement (next step).

Tier 3 - ICT560 chip replacement. With a known-good ICT560 donor chip (harvested from a scrap LT5 hashboard, sourced through Zeus Mining, or pulled from D-Central's parts pool), remove the dead chip with hot air, clean the pads with isopropyl and braid, position the donor chip with reflow paste, and run a careful reflow cycle. This is genuinely advanced bench work - mis-aligned chips, cold joints, and bridged pads are the most common failure modes for hobbyist attempts. If you don't have at least one prior successful ICT560 swap under your belt, route this step to D-Central.

Tier 3 - VCORE regulator stage repair. If Step 6 cold-resistance revealed a shorted MOSFET, identify the failed switching stage on the hashboard's power tree (compare to a known-good board, or use the failed-component cold-resistance signature). Desolder the failed MOSFET / driver / cap. Inspect the surrounding components for collateral damage. Replace with the correct part number from a parts pool. Power up on a current-limited bench supply first to confirm the rail comes up clean. Only then reinstall in the LT5 chassis.

Post-repair burn-in. Reassemble the LT5, reinstall both hashboards, boot, and run a 24-hour stable-hashing burn-in at nameplate workload before declaring the miner fixed. Capture: chip count every hour, per-board temperature curve, accepted-share rate, dashboard hashrate variance. A successful burn-in shows steady chip count, temperature within 5 C of pre-failure baseline, and hashrate within 2% of nameplate. Anything less and you have a residual issue you didn't catch.

Tier 4 - ship to D-Central if any of: Tier 3 bench skills aren't in your toolkit; cold-resistance shows a dead short you don't want to chase; control-board damage is visible; or the same LT5 has had multiple hashboard failures inside 12 months. Package: pull both hashboards, ship in foam separately from the chassis to avoid vibration damage in transit. Include: a printed note with the symptom, the firmware version, the failed hashboard slot, and any per-chip HW% data you captured. D-Central uses the data to route the bench workflow without re-running diagnostics you already did.

Tier 4 - control-board replacement. If Step 4 isolated the failure to a slot rather than to a board, the control board's hashboard interface is at fault. D-Central holds a limited stock of donor LT5 / LT5 Pro control boards harvested from past unrepairable units, plus a working relationship with Zeus Mining's stock for newer replacements. Control-board swap turnaround is typically 5-7 business days once parts are in hand. Bench rate plus the parts cost lands at CAD $185-$375 for a complete swap with verification.

Post-repair preventive checklist. Before the LT5 goes back into service: confirm the operating environment is in spec (ambient 15-30 C, dust-free intake, no direct sunlight on the chassis, dedicated circuit if at full power). Schedule a 6-month deep-clean reminder. Capture a baseline per-chip HW% snapshot for future comparison. Note the position of any reflowed or replaced chips in your maintenance log - those positions will fail again first if the underlying thermal stress isn't addressed. This step closes the loop on a real repair, not just a symptom fix.