Immersion Cooled Miner – Fluid Temperature Too High

Drop miner load. Use the miner UI, `bosminer` dashboard, or WhatsminerTool to reduce power to `50%` or shut down entirely. Bulk temp will begin dropping within 10-30 minutes if rejection is working at all. This is both a triage move and a diagnostic: if temp doesn't drop with `0 W` in the tank, your loop is dead (not just overloaded) and you go straight to Tier 2.

Visually confirm pump operation. Listen, touch the outlet hose, check any flow-meter display. Outlet hose should be measurably warmer than the tank surface during hashing; cold or identical = no flow. If you have a variable-frequency drive on the pump, confirm it's at its expected output.

Clean the dry cooler coil face. Power the dry cooler down, shine a flashlight through the coil from one side, photograph the dark patches. Compressed air from the clean-air side (blow *out* the same way air normally flows *in* — opposite direction collapses the fins), then a gentle water rinse if the unit is rated for it. Let it dry before restarting. A `15-minute` cleaning can recover `20-30%` of lost rejection capacity on a neglected outdoor cooler.

Check ambient at the dry-cooler intake. A coil behind a wall, under a deck, or inside a shed is pulling recirculated hot exhaust — its effective ambient is `5-10 °C` above outdoor ambient. Move any obstruction. Summer Canadian shops regularly hit `40 °C` at a poorly-sited coil intake.

Top up fluid level to the commissioning mark. Evaporation is slow but real; every tank loses `1-3%` volume per year depending on fluid type and lid seal. Low fluid = pump cavitation risk + reduced thermal mass + higher bulk temp. Use clearly-labelled commissioning-spec fluid only. Never mix brands without the vendors' blessing.

Measure pump motor current under load with a clamp meter. Compare against the motor nameplate and commissioning log. A mag-coupled pump drawing `40-60%` of rated current while running is almost always slipping the coupling. A pump drawing `>120%` is binding or running dry. Swap or service before continuing — a failing pump damages itself faster the longer it runs slipped.

Pull the pump suction strainer and clean. Gasket debris, chip offcuts from mining card assembly, and in older tanks, degraded fluid sludge all end up here. Strainer should be blown out with IPA, inspected for mesh damage, and re-gasketed with the manufacturer's replacement O-ring — never a random assortment-kit ring. `$5-$30 CAD` in parts, hours back of thermal performance.

Bleed air from the high point of the loop. Every immersion loop has at least one air pocket that accumulates over months. A bleed screw at the highest point is standard; if your loop doesn't have one, install one before next service. Air in the loop reduces effective flow by `15-40%` without changing pump current and is invisible to almost every instrument.

Check heat-exchanger water-side on liquid-to-liquid installs. Water inlet and outlet temperatures with a surface probe or K-type — healthy plate HX at rated load runs `4-8 °C` water-side delta. Low delta = plate fouled or water flow low. CIP with citric acid (`4%` solution, `60 °C`, circulated for `2 hours`) recovers most fouling short of mineral scale.

Inspect gasket / lid seal for fluid creep. Immersion tanks that lose fluid faster than normal almost always have a compromised lid gasket or a hose-fitting weep. Look for darker-coloured deposits on any fitting — they fluoresce under UV if your fluid has a tracer. Replace the gasket (manufacturer-spec, usually EPDM or Viton, `$20-$60 CAD`).

Sample and lab-test the fluid via the fluid vendor's program (most vendors offer a free oil-sample analysis for active customers). Key metrics: acid number (should be below vendor threshold, typically `<0.5 mg KOH/g`), dielectric strength (`>30 kV/2.5 mm` gap), viscosity at `40 °C` (within `±10%` of new-fluid spec), water content (`<100 ppm`). Fluid outside spec = schedule a change; don't wait for failure. Fluid change on a `200 L` tank is a `4-8 hour` job, runs `$800-$2,400 CAD` in fluid plus disposal fees.

Retune the PLC / control loop. Pull current setpoints: pump speed curve, dry-cooler fan ramp curve, alarm thresholds, PID P/I/D gains. Retune against current installed load, not commissioning load. For dry-cooler ramp: start fans at bulk `35 °C`, full-speed at `48 °C`, alarm at `55 °C`, shutdown at `60 °C` on hydrocarbon fluids; shift lower by `5 °C` on two-phase. If your PLC is vendor-locked and the retune tool isn't available, a control-system integrator or D-Central consulting hours are the fix.

Upsize the dry cooler or add a secondary heat exchanger if Step 5 of the diagnostic proved you're capacity-limited. Rule-of-thumb: for hydrocarbon single-phase immersion, you want `1.4×` miner wattage in rejection capacity at your `97.5th percentile summer ambient`. A tank pushing `12 kW` of miners at `32 °C` ambient needs a `17 kW` rejection loop. Undersized rejection will bake your fluid no matter how well you maintain it.

Replace the mag-coupled pump with a matched or upgraded unit. Key specs: max head at rated flow, chemical compatibility with your fluid (mineral oil vs fluorinated vs synthetic ester have different elastomer requirements), motor IP rating (at least IP55 for shop environments, IP66 for outdoor cabinets). Bench-test the replacement pump on a clean loop before putting the tank back into service.

Full fluid change-out with flush. Drain bulk, remove miners, flush tank with a vendor-approved flush fluid (usually a light hydrocarbon matched to the base fluid), pump out flush, inspect tank for sludge at the bottom, refill with commissioning-spec fresh fluid. Pressure-test all fittings before re-energizing. Schedule a `24-hour` loaded burn-in with continuous bulk-temp and hashrate logging before calling it production.

Stop DIY and engage a pro when: (a) cost-benefit of a DIY fluid change exceeds `$1,500` and you don't have the floor-space, PPE, or disposal setup, (b) you need a custom heat-rejection design because existing infrastructure is maxed out, (c) hashboards have visible fluid-attack damage (trace delamination, conformal-coat failure, salt deposits), or (d) your immersion build was a DIY-from-the-start retrofit and fundamental loop geometry is the real problem.

D-Central's immersion scope. Our shop doesn't run a dedicated immersion rebuild line the way we do for air-cooled Antminer/Whatsminer hashboard repair, but we do handle: (1) immersion-damaged hashboard assessment and repair on our bench — PCB rinse, re-flow, BGA replacement on water-or-fluid-contaminated boards, (2) ASIC-side firmware / sensor recalibration for miners running custom thermal profiles in immersion, (3) consulting hours for capacity audits, PLC retunes, and dry-cooler sizing for operators running `2-200 kW` immersion fleets. [Book via the ASIC Repair and Consulting service pages.](https://d-central.tech/services/asic-repair/)

Pack and ship any damaged hardware correctly. Drain and rinse hashboards with `99%` IPA, air-dry 4 hours, then pack in anti-static bags with desiccant. Include a service note with: tank fluid brand and age, coolant lab results if available, bulk-temp log for the `72 hours` prior to failure, and whether the miner was running when fluid temp alarmed. That metadata is what turns a `$600` quote into a `$300` quote because we skip diagnostic time we'd otherwise bill.