Whatsminer M50S – Temperature Too High

Hard power cycle for 10 minutes at the breaker or wall plug — not a soft reboot. Allows heatsink thermal mass to drop and clears the firmware's trip-state memory. If the miner comes back and holds under 75°C on all three board sensors for 30 minutes, you have enough headroom to work through the remaining steps without cascading another shutdown event mid-diagnosis.

Move the M50S out of enclosed space. Closets, cabinets, and compact racks without forced exhaust recirculate hot air back into the intake. Place the miner in an open garage or large room with at least 60 cm of exhaust clearance behind it. If TEMP_OVER stops on the same firmware, airflow is the cause — move to Tier 2 cleaning steps, not hardware replacement.

Pop the intake grille and visually inspect the filter and heatsink fins. Clogged filter — swap it or wash it with soap, dry fully before reinstalling. Dust-packed heatsink fins — note it for Tier 2 compressed-air work. Do not run a miner with a visibly clogged filter; you will re-trip TEMP_OVER within an hour and accumulate unnecessary thermal stress on the chips.

Verify intake ambient with an independent thermometer held at the grille, not the dashboard reading. Target: ≤30°C for performance mode, ≤25°C for a high-performance tune. If your probe says 22°C and the dashboard says 38°C, the button-board TMP75 sensor is drifting — skip directly to step 11, which is a cheap part swap, before you rearrange your ventilation.

In WhatsminerTool, temporarily switch to low-power mode (or drop the power-limit ceiling 10-15%). This is not a permanent fix but it buys stability while you work through cleaning, inspection, and hardware checks. It also prevents the thermal-cycle-recover-thermal-cycle pattern that accelerates cap and solder fatigue on a miner that is otherwise physically fine.

Full chassis blow-out with compressed air, outdoors, miner powered off. Intake side first, then exhaust, then through heatsink fins lengthwise. Immobilize each fan with a toothpick or finger so it does not free-spin under the air blast (damages bearings). Compressed lint and dog hair wedged between fins is invisible from outside and responsible for ~30% of M50S TEMP_OVER returns per D-Central bench data.

Verify both fans spin freely by hand with power off. Flick each fan blade — a healthy 12038 dual-ball-bearing server fan spins several rotations before stopping. A dying bearing grinds, wobbles, or stops after half a rotation. Either symptom means fan replacement before this miner runs again. Related: Whatsminer M50S Grinding Fan Noise. One bad fan halves CFM and guarantees TEMP_OVER on any warm day.

Export `miner.log`, `power.log`, and `system.log` via WhatsminerTool over TCP 4028. Identify the FIRST error code in the timeline. 2xx (233-235, 243-245, 275) = PSU tripped first, go to step 9. 6xx = intake sensor (real or drift), step 11. 35x = hashboard-specific, steps 12-14. The sequence matters more than the final code — the first event names the root cause.

Check PSU busbar screws under load with an IR thermometer. Expected: 50-60°C at full load. Hot spots above 80°C = loose screw = resistive loss = cascade into 233-235/275 and then into a hashboard-level TEMP_OVER. Power down, torque the copper-row screws to MicroBT spec, re-check. Loose busbar is the most-missed cause in the MicroBT PDF because the official text just says 'check copper row screw' without a torque value.

Swap the PSU with a known-good unit from another compatible miner. If TEMP_OVER stops with a different PSU on otherwise identical hardware and firmware, you had a PSU thermal cascade, not a hashboard problem. Related error: Whatsminer M50S Power Supply Failure. Treat the original PSU as suspect and bench-test or retire it — running it in another miner just delays the next repair event.

Replace the button-board temperature sensor (or the whole button board). If step 4 showed dashboard ambient reading more than 3°C above your probe, the TMP75 sensor on the button board is drifting. Order a compatible M50S button board (CAD $25-60), unscrew the front, disconnect the ribbon, swap, reconnect. 15 minutes, single ribbon cable. This is the most over-diagnosed M50S repair in the field.

Thermal imaging on a suspect hashboard. Miner off 10 minutes, lid off, power back up, run 60 seconds at load, shut down. FLIR ONE Pro or a bench thermal cam sweeping the full board lengthwise. Look for one chip position more than 5°C hotter than neighbours (paste failure or lifted heatsink) or a uniformly-hot board (airflow starvation at that specific slot). Document chip positions for step 13/14 reference.

Re-seat heatsinks and refresh thermal paste/pads. Remove the heatsink, clean old compound with 99% isopropyl and lint-free wipes, apply a rice-grain-sized dot of Thermal Grizzly Kryonaut or Arctic MX-6 per chip. Excess paste insulates — do not glop it on. Replace failed thermal pads between heatsink and voltage-domain ICs. Re-torque heatsink fasteners evenly in a cross pattern to avoid tilt.

Reflow the worst-performing chip if IR consistently isolates a hot position across multiple sessions. Preheat bottom-side to ~150°C, top-side hot air at 310-330°C for ~30 seconds, cool naturally, re-paste, reassemble. BM1385-family chips on the M50S tolerate one reflow cycle well. Two reflows on the same chip rarely hold — the second attempt usually means chip replacement, which is Tier 4 bench-only work.

Underclock via WhatsminerTool performance mode: drop power-limit 5-10%. Community consensus on BitcoinTalk's M50S thread is 1-3% hashrate loss for 8-12°C lower junction temperatures. On a second-hand M50S with unknown history, this is the highest-ROI stability tune you can make — it extends chip life meaningfully and eliminates marginal TEMP_OVER trips without any hardware replacement cost.

Verify firmware is an authorized MicroBT release. Avoid nightly or pre-release builds that shipped thermal-curve regressions (community reports on Whatsminer Medium flag several). Do NOT cross-flash Vnish/Asicdip/HiveOS on M50S — the MicroBT integrity check (100001-100003) trips, and M60S-generation hardware signature-verifies on boot and can hard-brick. DCENT_OS is Bitmain-only and does not apply here — stay on MicroBT firmware.

Stop DIY and book D-Central ASIC Repair when: paste refresh and button-board swap did not clear 350/351/352; IR isolates a chip position that recovers briefly after reflow and returns in days; you see capacitor bulging, PMIC discoloration, or burnt-component smell anywhere in the chassis; OR 600/610 persists in a verified-cool room after a new button board. At that point it is test-fixture, chip-replacement, or control-board bench work.

Shipping protocol: pack hashboards (or the full miner) in anti-static bags, double-boxed with at least 5 cm foam on every side. Include a written note — exact error-code timeline from miner.log, ambient conditions when the fault occurred, firmware version, PSU history, what you have already tried. Every hour of diagnostic saved is an hour off your invoice. D-Central turnaround is 5-10 business days; Canada, US, and international all accepted.