IceRiver Error 350/351/352 Overheat Protection Triggered

Hard power-cycle the miner: switch off at the wall outlet for 60 seconds, then power back up. Some 350/351/352 events are transient - a brief ambient spike, a fan stall that recovered. A clean cold boot lets the chips settle to ambient and re-arms the firmware's protection logic. If the code clears and stays clear after 30 minutes of full-load hashing, no further action needed. If it returns, escalate.

Measure inlet air temperature 5 cm from the front grille with an IR thermometer or honest digital thermometer - not the room middle. Verify <=35 C Normal Mode or <=30 C Performance Mode (IceRiver's published spec). Confirm at least 15 cm clearance front and rear, no obstructions, no other miners exhausting into this miner's intake. Drop the room temp, open a window, or relocate the miner if ambient is out of spec - environmental fixes are the cheapest fix you will ever apply.

Visual inspection of intake grille and exhaust path without opening the chassis. Look for dust mat, lint, pet hair, or external blockage. Wipe the grille with a damp microfiber cloth. Move anything on top of, behind, or directly in front of the miner. Plenty of 350 trips clear here without ever cracking the case.

Open chassis, shop-vac the heatsink fins. Power off at the wall, undo the corner chassis screws, lift the cover. With a shop-vac on low suction (or compressed air at <=30 PSI), clean the heatsink fins on the offending board, the intake side of the board, and any visible dust accumulation in the chassis. Do not touch fan blades while the miner is powered on. Reassemble, power up, run a 60-minute load test. The most common cause of 350/351/352 after 12+ months of service is dust loading - this is the fix 60-70% of the time.

Verify fan RPM under load. Reboot, monitor the dashboard fan RPM fields while the miner ramps to full hashing. KS-class 12038 6000 RPM fans should hit 5500-6500 RPM under load. Fans below 5000 RPM or stalling intermittently are failing - bearing wear, blade damage, or driver IC degradation. Replace.

Replace the failing 12038 fan. Order the IceRiver-spec 12038 6000 RPM axial fan (DF1203812B2UN is the part number on most KS retail units - confirm against the sticker on your existing fan). Power off at the wall. Disconnect the failing fan's 4-pin PWM connector at the controller. Unscrew the fan from the chassis (typically four M4 screws). Mount the replacement fan in the same orientation (airflow arrow matches the original - critical, a backwards fan is worse than no fan). Reconnect, reassemble, power up, verify RPM on the dashboard.

Re-seat the hashboard data ribbon and power connectors on the offending board. Power off, open chassis, disconnect the data ribbon and power connectors on the suspect hashboard. Inspect the headers for bent pins, oxidation, blackening. Reconnect firmly until you hear/feel the click. Loose connectors do not cause 350-class trips directly, but a partially-seated power connector can starve the board of current and compound a marginal thermal situation.

Inspect the chassis-level airflow path with the chassis open. Anything blocking it (cable harness routed across a fin pack, tape residue from prior repairs, an aftermarket noise-damping mod that throttles airflow, foam packing material left from shipping) is hurting cooling. Clear obstructions, route cables away from airflow paths, document the chassis interior with a phone photo before reassembly.

Improve external ducting and exhaust if ambient is borderline. Build an exhaust duct: flexible insulated 6-inch duct from the rear of the miner to a window, vent hood, or adjacent unheated space. A simple duct drops effective ambient at the intake by 5-10 C because you are stopping the miner from re-ingesting its own exhaust. Highest-leverage modification for retail home miners running KS-class hardware in residential space.

Pull the offending hashboard for paste refresh. Power off, open chassis, unmount the hashboard (note the screw locations and orientation - phone photo before disassembly). Bring the board to an ESD-safe bench. Wear an anti-static wrist strap connected to chassis ground or an ESD mat. The 1004LV100 chips on KS5-class boards are ESD-sensitive.

Remove the heatsink and inspect the thermal interface. Unscrew the heatsink mounting screws / clips per the order documented in IceRiver / Zeus repair notes (typically corner-out then center, but verify against your specific board revision). Lift the heatsink straight up - do not twist or pry, the paste seal can lift pads. Inspect: paste pattern (should be even and slightly damp; pump-out shows as dry edges with a wet center, or patchy coverage), thermal pads on PMIC / voltage-domain ICs (should be even thickness and pliable; aged pads are compressed, cracked, or visibly shrunken).

Clean old paste and pad residue. Wipe the ASIC die surface and the heatsink base with 99% isopropyl alcohol and a lint-free wipe. Do not soak - IPA is fine on the die top, not great if it pools under nearby SMD components. For thermal pads, peel cleanly and discard; clean the IC tops with IPA. Goal: bare metal on every contact surface, no residue.

Apply fresh thermal paste with a pump-out-resistant compound: Arctic MX-6, Thermal Grizzly Kryonaut, or equivalent. Pea-sized drop in the center of each ASIC die (do not spread - heatsink pressure spreads it evenly). Do not use cheap no-name paste sold for under $5 per tube; cheap paste is the reason your old paste pumped out in 18 months.

Replace thermal pads on PMIC / voltage-domain ICs. Use the same thickness as the original (typical KS-class pads are 1.0 mm or 1.5 mm - measure the original before discarding). Pad material: silicone-based, 6 W/mK or better. Cut pads to fit each IC top with about 1 mm overhang on each side. Press into place, peel the protective film, mount heatsink immediately.

Re-mount heatsink with proper torque. Tighten mounting screws / clips in a star pattern - center first if applicable, then opposite corners alternating. Distributes paste evenly and prevents tilting the heatsink against the die. Do not over-torque - finger-tight plus a quarter-turn is the right ballpark on M3 fasteners (about 0.5-0.7 Nm with a torque driver). Reassemble the miner, run a 60-minute load test, watch Temp1/Temp2 to verify the offending board is now within 5 C of its peers.

Stop DIY and ship to D-Central when: full Tier 3 refresh fails to clear the trip within 30 days; trip follows the board through slot-swap with fresh interface; you see capacitor bulging or burnt-component smell; two or more boards trip simultaneously after refresh; or you do not have ESD-safe bench, IR thermometer, or pump-out-resistant paste. Book a D-Central ASIC repair slot at d-central.tech/services/asic-repair/.

D-Central bench process: test fixture runs the suspect hashboard at controlled load with thermocouple-instrumented chip junctions to confirm or rule out chip-level hotspot. Per-chip current monitoring isolates degraded ASICs on the array. Heatsink interface refresh with industrial-grade pump-out-resistant paste and properly-spec'd silicone thermal pads. Chip replacement (1004LV100 or KS3-equivalent) with graded inventory where applicable. Post-repair 24-hour burn-in at nameplate with continuous Temp1/Temp2 logging to verify peer-spread <=5 C before returning the board. D-Central is positioning as the Western English-language IceRiver repair authority - the only direct competitor on KS-series is Zeus Mining (China-based, trust-deficient).

Ship safely. Pack the suspect hashboard(s) - or the full miner if you cannot isolate - in anti-static bags, double-box with at least 5 cm foam every side. Include a printed note with: model and serial, observed error code(s) (350, 351, or 352), exact firmware version, board(s) affected, what you have already tried (cite the step numbers from this page), ambient conditions during the trip, and your contact info. The note saves diagnostic time, which saves you money.