Whatsminer M53S+ – Thermal Runaway Shutdown

Drop the miner to `Low-Power` mode immediately via WhatsMinerTool (Remote Ctrl → Power Mode → `Low`) or pull the network cable so the firmware idles. This buys minutes to diagnose without further chip stress. If chips are already past `95 °C`, hit the breaker and let the loop pull residual heat for `5-10 minutes` before re-energising.

Confirm the primary pump is running and not cavitating. Walk to the pump skid; listen for steady hum, not 'marbles in a can.' Verify discharge pressure gauge sits at the normal operating value. If the pump is dead, hit the standby cutover. If neither pump is running, breaker the M53S+ before the loop fully stagnates.

Inspect the sight glass for steady non-aerated flow. Bubbles, foam, or stop-and-start patterns mean air ingestion or a pump suction problem. A solid, steady column means the loop is at least moving.

Verify ambient at the dry cooler / primary heat exchanger is in spec. If facility ambient has climbed (heatwave, AC failure, neighbouring equipment dumping heat), that is the upstream cause. Restore HVAC or add ventilation before re-loading the miner.

Bleed the M53S+ cold plate if the chassis was recently moved, freshly commissioned, or the loop was opened. Crack the outlet quick-connect a quarter-turn with the pump running until non-aerated coolant flows steady for `~30 seconds`. Re-seat. Air in the serial three-section cold plate triggers thermal events on whichever section traps the bubble.

Pull and clean the `100-mesh` main-line strainer. Isolate, drain, inspect — anything older than six months will have debris. If your loop never had a `100-mesh` strainer (or had a coarser mesh), install one before re-energising. Particulate in the cold-plate microchannels is one of the slowest, most expensive failures to undo.

IR-verify all temperature sensors against the dashboard with a `±1 °C` IR thermometer. Shoot inlet and outlet manifolds, cold-plate skin over each `SMx` zone, and the PMIC packages. Any sensor reading more than `5 °C` cooler than IR is lying — stop running the chassis until it is replaced or recalibrated. A lying sensor is more dangerous than an honest over-temp.

Quantify loop flow with a clamp-on ultrasonic flow meter at the M53S+ inlet. Target `≥10 L/min` per chassis, steady. Flow oscillation more than `±15 %` points at pump cavitation or a cycling valve. Hard-low flow points at pump degradation, valve closure, or strainer block.

Measure differential pressure across the miner — inlet pressure tap minus outlet pressure tap. Expected `~30-60 kPa` at rated flow on a clean cold plate. Elevated ΔP means blockage. Document the value before any cleaning so you have a baseline for next time.

Pull a coolant sample (`~250 mL` from the sample port). Original PG25 has a faint pink, yellow, or green tint depending on inhibitor brand. Brown, black, cloudy, or particulate-laden coolant means chemistry is exhausted — drain, flush, refill before this miner runs another full-power shift.

Reverse-flush the M53S+ cold plate if Step 9 showed elevated ΔP. Disconnect from the loop, connect a clean demin-water source to the *outlet* quick-connect, drain through the inlet to a waste container. Use a MicroBT-approved cleaning additive. Repeat until flush water runs clean. Reconnect and re-bleed (Step 5).

Inspect and torque every quick-connect to manufacturer spec. Cross-threaded or partially-engaged connectors cause flow-loss events that masquerade as thermal events. A drop of approved thread sealant or a fresh o-ring on a worn fitting takes a chronic intermittent fault out of your life.

Send a coolant sample for full chemistry analysis — pH, inhibitor reserve, dissolved iron / copper / aluminium, total dissolved solids, microbial count. Local industrial water-treatment lab does this in `48-72 h`. Use the result to schedule full coolant changeout (`12-24 months` for PG25 in well-maintained loops; `6-12 months` for under-inhibited or contaminated).

Service the primary heat exchanger. Plate-and-frame: crack and chemically clean plates, replace gaskets if hardened. Dry cooler: backwash fins top-down with low-pressure water (do not bend the fins), confirm every fan at full RPM, replace any fan with audible bearing wear. Cooling tower: blowdown, basin clean, fill inspection, makeup-water chemistry verification.

Replace the primary pump or its impeller if Step 8 showed cavitation that persisted after suction-side fixes. A cavitating pump ingests bubbles that erode the impeller, which cavitates worse. Spec the replacement for `+25 %` head margin over peak demand, not nameplate match.

Tune BTMiner power mode and frequency targets to match what the loop can actually reject. If primary-side capacity is `±10 %` of rated heat load and HPM operation is `+15 %` over Normal, drop HPM and run Normal year-round. The hashrate cost (`~10-15 %`) is cheaper than one runaway-induced repair.

Update BTMiner firmware to current stable; thermal-management logic gets revised version-to-version. Roll back one version if the runaway started immediately after a firmware update. Document firmware version and date alongside the fault timeline so you can correlate.

Add loop instrumentation if you do not have it: continuous inlet/outlet temp logging, pressure logging at pump discharge and miner inlets, flow-meter logging, coolant-level switch with low-level alarm, leak-detector cable along the floor under the rack. Instrumentation that catches a runaway at minute one instead of minute thirty pays for itself the first event.

Stop DIY when silicon damage is visible, IR confirms a lying sensor and you have no bench fix, runaway recurs within `30 days` of a Tier 1-3 fix, the cold plate is suspected blocked beyond reverse-flush recovery, or the chassis was over-temp for more than `~5 minutes`. The M53S+ silicon is not a forgiving platform — book D-Central hydro service.

D-Central hydro bench: chassis goes onto our calibrated hydro test loop with flow, pressure, and temperature instrumentation across inlet/outlet. We isolate each `SMx` section, IR-map the cold plate at full power, run BTMiner diagnostics with chip-level visibility, and replace any combination of cold-plate gaskets and seals, hydro flow / temp sensors, BM1362 chips, PMIC components, and full coolant flush + refill. Donor BM1362 inventory available for unrecoverable boards.

Ship safely. Drain the chassis fully (residual coolant in transit is fine but expect partial drain). Cap both quick-connects with supplied or industrial blanking caps. Anti-static-bag any pulled board. Double-box. Include a printed note with observed fault codes and timestamps, firmware version, operating mode (Normal / HPM), coolant brand and last changeout date, and your contact — every minute of saved diagnostic time at our bench is repair-cost saved on your invoice.