Whatsminer Error 350-352 – Hashboard High Temperature Protection

Hard power-cycle at the PDU for a full 60 seconds — not a soft reboot. `btminer` caches per-board fault state across soft reboots, and a cold start is the only way to clear a stuck `Error 350` flag lingering after you fixed the root cause. Wait the full 60 seconds before powering back on so the PSU caps drain. After power-up, watch MinerTool for 15 minutes to confirm the flag cleared. This alone resolves roughly 8-12% of `350` tickets on D-Central's Whatsminer queue.

Verify intake ambient with an IR gun at the front grille — not room-middle. Target `<=30 C` at the front face. MicroBT's `0-35 C` spec is data-centre; in a residential install, intake can run `5-10 C` hotter than room-centre if the miner is wall-mounted, behind another miner's exhaust, or in a closed closet. Ambient over 30 C at the grille means the rest of this guide is academic until you fix the environment.

Clear the 30 cm zone in front of the intake grille — shelves, boxes, curtains, another miner, shop towels. Whatsminer is a blow-through design that chokes on restricted intake. Zero-dollar, two-minute fix that resolves a surprising percentage of `350` tickets that appeared after a workshop rearrangement.

Re-pull the fault log and per-board temps 30 minutes after power-up. SSH `status` or MinerTool remote monitor. Confirm whether `350 / 351 / 352` has cleared or returned, and at which board index. If the flag moved to a different board index, root cause is global (paste aging across the miner, or ambient) rather than board-specific.

Inspect the intake mesh for obstruction. Run a finger across the full surface — pet hair, paper shreds, packing foam residue, dead insects. Even 10% mesh blockage creates a microlocal airflow deadzone behind the obstruction, flagging the hashboard directly behind it. Clean with a soft brush or mesh-safe vacuum nozzle.

Compressor-blow the fin stack from the exhaust side. Real air compressor at `80-90 PSI`, miner off, blow from back (exhaust) through to front (intake). Reverse-direction blowing pushes dust *out* of the fins. Canned air cannot move sufficient volume. Spin each fan blade by hand while blowing to clear the hubs. Expect a visible dust cloud pass one; second pass after 30 s confirms clean. Dust carpet raises chip temps `5-10 C`, which is exactly the margin you lost to trigger `350`.

Compressor-blow the fans individually, off-chassis. If Tier 1 didn't clear the flag, pull side panels and blow each fan directly. Gunk on blades unbalances the rotor and drops RPM below spec without triggering fan-error. A clean fan at `100% duty` should sound like a jet spool-up; a gunked fan sounds muted.

Reseat the flagged hashboard's ribbon and DC power connectors. Power off, wait 5 minutes for caps to drain. Label slots 0/1/2. Inspect IDC pins for oxidation, blackening, or bending; wipe with `99% IPA` on lint-free wipe if needed. Reseat firmly until the latch clicks. A marginally-seated data ribbon feeds garbage NTC readings to the control board, triggering phantom `350` flags.

Swap-test the flagged board between slots. Highest-value diagnostic on Whatsminer thermal faults. Move the flagged board to a different slot, power on, run 30 minutes, pull fresh status. If `Error 350` follows the board (e.g. `350` in slot 0 becomes `351` in slot 1), it is the board itself — paste, NTC, or silicon. If the flag stays in the original slot regardless of which board sits there, it is control-board / harness. This 30-minute test saves hours of parts-throwing.

Replace any fan reading below `5500 RPM` at `100% duty`. M30S / M50S / M60S use standard `12038` or `12038S` axial fans — Delta, Nidec, Sanyo Denki OEM or GDStime / JSL-equivalent. Replace all four as a set, not one at a time — survivors are close behind on the aging curve, and a fresh single fan develops imbalance within weeks. Expect `$25-$60 CAD` for four decent OEM-equivalent fans.

Roll back `btminer` firmware one version if `Error 350` appeared after a recent upgrade. Download prior firmware from the Whatsminer firmware archive or community backups. Use MinerTool firmware update with the rollback image. Isolates firmware-threshold-tightening as a cause. If the flag clears on older firmware, report on the Whatsminer sub-forum — community pressure is the only mechanism that gets undocumented MicroBT changes reverted.

Full thermal paste refresh on the flagged hashboard. Remove board from chassis. Unscrew heat-spreader — multiple Phillips or Torx with spring-loaded retention. Lift spreader carefully to avoid tearing factory thermal pads. Clean old paste with `99% IPA` and lint-free wipes — zero residue on die or spreader. Apply `Arctic MX-6` or `Thermal Grizzly Kryonaut` in thin uniform film, rice-grain blob per chip region, let mounting pressure spread. Reassemble with even torque in star pattern. On 24-month-old boards running `350`, paste refresh alone clears the flag in the majority of cases.

Full thermal paste refresh on all three boards if Step 5 diagnosis pointed to global paste aging. Do all three in the same service window. Factory paste ages at roughly the same rate across the miner — single-board refresh means you are back in the chassis in 3-6 months for the next one. Budget a full afternoon, skip the regret.

Replace crumbled thermal pads on PMIC and voltage-domain components while the heat-spreader is off. Crumbled pads transfer heat poorly, raising nearby chip temps indirectly through the PCB. Measure old pad thickness with calipers before sourcing — MicroBT does not publish pad specs. `0.5 mm`, `1.0 mm`, `1.5 mm` pads cover the common cases.

Replace suspect NTC thermistor (advanced — SMD work). If NTC cross-check confirmed drift (NTC reports hot, in-die silicon temp reads cool), the fix is component-level thermistor swap. `10K NTC 0603` or `0805` parts from Digikey / Mouser match most Whatsminer sensor footprints, but confirm resistance curve (`B3380K` vs `B3950K`) against the board marking. Hot-air rework, flux, tweezers. If uncomfortable with SMD at this level, this is where DIY stops and D-Central takes over.

Replace the full fan bank with higher-static-pressure industrial fans for marginal environments. For Canadian garage installs running through summer, or for miners in poorly ventilated spaces, upgrading from stock `12038` to `Delta THA1248BE` or similar high-SP fans adds `1000-1500 RPM` headroom and `5-8 C` margin across the whole miner. Prevention play for operators who know their environment is marginal.

Stop DIY and book a D-Central Whatsminer repair when: same board flags `Error 350` after a clean paste refresh + fresh fans + verified clean ambient; visible capacitor bulging, cracked MLCCs, or burnt-component smell; swap-test shows slot-specific fault (control-board / harness); Tier-3 held less than 60 days before return; you lack hot-air rework and diagnosis points to chip-level work. Book at d-central.tech/services/asic-repair/ — Canadian bench, `5-10 business day` turnaround, Canada / US / international.

D-Central bench process on `Error 350`: test-fixture boot with programmable load and instrumented thermal probe. Per-chip thermal mapping via IR camera and in-die readings. NTC cross-validation against bench reference resistances. Confirmed chip-region root cause gets component-level repair with graded replacement chips, correct reflow profile, fresh TIM, fresh pads, and 24-hour burn-in at nameplate before return ship. PMIC / control-board / harness root cause gets component-level repair with matching-spec parts, not a blanket swap.

Ship safely. Hashboards in anti-static bags inside the chassis (whole-miner ship) or individually bagged (boards only). Double-box with `>=5 cm` foam every face. Include a physical note: full `btminer` status snapshot from Step 1, firmware version, install date, last paste-refresh date, ambient at install, and which Tier 1-3 steps you already ran. Documented ship-ins save real dollars.