Whatsminer M56S – Temperature Too High

Power off at the PDU for 5 minutes and restart. Not a soft reboot from the web UI — a full cold shutdown. This clears wedged driver state and lets sensors reset. If TEMP_OVER returns within 20 minutes under the same ambient conditions, you have a real thermal fault and need to continue. If it runs clean for hours after, file it as a one-off firmware glitch and set a monitoring alert on temperature trending so you catch a recurrence before it becomes a shutdown.

Measure intake air temperature with a handheld thermometer at the miner's intake grille — not in the middle of the room, not at the exhaust. Air-cooled M56 target: below 30 C. Hydro M56S dry-cooler intake target: below 35 C. If you are above spec, fix the room before touching the miner: open a door, deploy a portable AC, redirect exhaust, or move the miner to a cooler bay. Highest-yield Tier 1 move in summer months.

Check for exhaust-into-intake recirculation — a classic container or basement layout mistake where hot exhaust gets pulled back into the intake within the same enclosure. Feel the intake: if it is noticeably warmer than room air 2 m away, redirect exhaust with a duct, add an intake plenum, or physically reposition the miner. No tools, no cost, massive thermal win — and fixes 'mystery TEMP_OVER' events in enclosed sites.

Verify coolant pump is running and all radiator fans are spinning (hydro M56S). Walk the loop. Ear to the pump housing — you should hear steady humming, not clicking or silence. All dry-cooler fans spinning. Inlet coolant warm is fine, hot-to-touch (above ~50 C) is a problem. No pump flow = shut down the miner immediately to protect the silicon; hashing with no coolant destroys BM1398 chips in minutes.

Via WhatsminerTool (Monitor -> Error Log) or directly via the API on TCP port 4028, capture the exact error code(s): 600/610, 350-352, 233-235, 243-245, or 275. Different codes point at different subsystems. Write them down before any other action — they are the first evidence trail and they disappear from some log views after a restart.

Clean intake grille, radiator fins, and all filters. Power off at the PDU. Use compressed air (short bursts, can upright) and a soft brush. Dust on a hydro radiator drops heat rejection capacity by 20-40% after a year in a dusty environment. Wipe the intake grille with a damp microfiber cloth. This is your 90-day maintenance baseline for any Whatsminer in a non-clean-room site.

Swap the button-board environment sensor if it is reading wrong. The sensor sits on a small 'button board' near the control board — not on the control board itself. If your handheld thermometer disagrees with the reported environment temp by more than 3 C, replace the button board. Community-confirmed fix for ~30% of 'cold room, high temp alarm' cases (Zeus Mining). Takes 10 minutes, costs under 30 CAD.

Re-torque all copper busbars between PSU and hashboards. Power off, cool 20 minutes, expose the copper joints. Inspect each for discoloration or blackening — any browning means replace the bar, not just re-torque. With a torque screwdriver, re-seat each M2.5/M3 busbar screw to ~8-10 Nm (community-measured; MicroBT does not publish this). About 20% of M56S TEMP_OVER events trace to loosened busbars after thermal cycling.

Clean busbar contact surfaces with isopropyl alcohol 99% and lint-free wipes. Oxidation on a copper busbar interface creates resistance, which creates I2R heating at the joint, which trips codes 233-235 and eventually TEMP_OVER. IPA 99%, wipe, let dry, reassemble. If the copper itself is pitted or burned, replace the bar — do not sand or file. Pitted copper never seats right and the failure mode accelerates.

Reduce power-mode or disable high-performance mode via WhatsminerTool (Pool -> Power Mode). Drop from high-performance to normal or low-power. Observe TEMP_OVER behavior for 2 hours. If it stops, you have confirmed the miner is thermally marginal at high-performance — either accept reduced hashrate, or fix the underlying thermal issue before returning. Right move for summer operation in a site not scaled for full-power M56S load.

Update or roll back firmware via WhatsminerTool. Power off, connect via the tool, cross-reference current version against community release notes (Bixbit mirror is most reliable English-language source). If on a known-buggy build, roll forward or back one version. M56S signature verification is less aggressive than M60S (which hard-bricks on downgrade) but still verify the build is authorized — recovery from an unsigned flash requires JTAG.

Replace TIM between ASICs and cold plate on the suspect hydro hashboard. Pull the cold plate. Clean both chip tops and plate surface with IPA 99% and lint-free wipes. Re-apply fresh TIM — Arctic MX-6 or Thermal Grizzly Kryonaut. Avoid conductive pastes near the hashboard. Uniform thin layer. Re-torque the cold plate per the original crossed pattern, hand-tight then quarter-turn. 12-18 months is the practical TIM refresh interval on continuous-duty hydro M56S.

Swap hashboards between slots to isolate a bad board. Label the 3 slots 0/1/2. Move the suspect board (thermal outlier from Step 6 of diagnostics) to a known-good slot. Run 30 minutes. If the hot behavior follows the board, the board is the problem. If it stays in the slot, the control board / cable / connector path is the problem. Same swap-to-isolate technique that works on every ASIC ever made.

Replace the SM0/SM1/SM2 on-board temperature sensors. These are TMP75-class I2C parts. If diagnostics showed a stuck sensor (pinned at -40 C or 127 C), desolder and replace the individual sensor. This is SMD rework — hot-air station territory, not iron and solder wick. If you are not comfortable with 0603-class SMD work, skip to Tier 4 and let the bench handle it.

Inspect and re-torque all PCB-mount power connectors and the PSU output terminals. Vibration and thermal cycling loosen the heavy-current connections between PSU and hashboards. Re-seat the entire power chain: PSU -> DC bus -> hashboard input. Any connector that shows blackening on disassembly needs replacement, not a re-seat. Hydro M56S is particularly prone because site-level pump vibration propagates into the miner enclosure.

Run a 30-minute soak on a single-board config via MinerTool's board-disable feature. An undocumented field workaround (BitcoinTalk, community-documented): disable individual hashboards via the MinerTool interface and keep the miner running at 1/3 or 2/3 hashrate. Use this to isolate the thermal culprit, or as a stopgap to keep earning while you wait for a repair slot. Not a permanent fix — diagnostic and stopgap only.

Stop DIY and book a D-Central repair slot when: (a) a hashboard is hot-outlier after TIM refresh and slot-swap, (b) more than one sensor is stuck and you are not SMD-equipped, (c) a copper busbar is burned beyond surface-clean, (d) the PSU reports over-temperature warnings (code 275) after a cold-start soak, or (e) per-chip status shows dead chips at 0 MHz. Any of those is test-fixture territory. Book ASIC Repair at d-central.tech/services/asic-repair/.

Ship safely to D-Central. Anti-static bags on hashboards, double-boxed with at least 5 cm of foam on every side. If shipping the full hydro unit, drain the coolant loop first per MicroBT's published drain procedure — shipping with coolant in it is how radiators crack in transit. Include a note with exact error codes, firmware version, observed symptoms, and ambient/coolant conditions when it tripped. Every line of that note saves bench time, which saves you money. Typical turnaround: 5-10 business days.