Antminer S19 Hydro – Coolant Temperature Too High

Controlled shutdown and cool-loop observation. Power the S19 Hydro off at the PSU switch but leave the CDU pump running for `10 minutes`. Read loop inlet temp every minute. If inlet drops from trip temperature back under `55 °C` with only the pump running, your heat-rejection path is working — the trip came from the miner saturating the loop, not from a cooling fault. Record the cooldown curve; it is your baseline for every following step. If loop temp does not drop, you have a heat-rejection problem (Tier 2 and Step 2 of the diagnostic path) and no miner-side work will fix it.

Verify coolant level and top up with the correct fluid. Open the CDU reservoir and visually verify coolant is at the MAX line. If below, top up *only* with the coolant grade the loop was filled with — deionized water or the labelled propylene-glycol mix. Do not top up with tap water, distilled-but-not-deionized water, automotive antifreeze, or PC coolant dye. Close the reservoir, bleed air at the highest point of the loop with pump running until flow is bubble-free, and observe flow rate and loop temperature for `15 minutes`.

Clean the dry cooler / radiator air-side. Power off, shut the pump, let the HX cool. Inspect fins for dust, cottonwood fluff, leaves, spider webs, pet hair. With HX cool, use compressed air capped at `≤90 psi` to blow dust out in the opposite direction of normal airflow (exhaust-to-intake) to push debris back the way it entered. Comb bent fins with a fin-comb. A `20–30%` capacity loss from clogged fins is typical on a year-old residential install. Re-measure ambient-to-loop approach — should improve within `30 minutes`.

Confirm ambient air into the dry cooler is within spec. Place a trusted thermometer `5 cm` in front of the HX intake face. Sustained operation requires ambient `≤30 °C`; ideal is `≤25 °C`. Higher means an environmental problem: another miner's exhaust being recirculated, direct summer sun on the HX, intake blocked by furniture or snow. Relocate, shade, duct, or baffle until ambient at the HX intake is clean. No amount of fiddling with the miner compensates for an HX inhaling `40 °C` air.

Factory reset via the reset button. Bitmain's reset button is valid only `2–10 minutes` after boot; press-and-hold for `5 seconds`, wait roughly `4 minutes` for auto-restart. Clears any user-level tuning misconfigurations (aggressive autotune, custom frequency tables, mis-set thermal envelopes) that push chips past stock and dump more heat into the loop than the HX can reject. Reconfigure pool and worker settings after reset. If `ERR_WATER_TEMP` stops tripping on stock, your prior config was over-tuned for the cooling envelope.

Verify pump amperage and flow rate against spec. Measure pump draw with a clamp-on ammeter on the pump supply lead, CDU at normal duty. Compare against the pump nameplate and against the commissioning baseline. Confirm flow rate on the CDU display or in-line flow meter — S19 Hydro Pro/XP target roughly `10–12 L/min` per miner (verify against your specific CDU manual). Below spec = flow problem upstream; above spec with hot loop = HX capacity problem. High-and-drifting pump amperage is the signature of a failing circulator.

Check and clean any in-line coolant filter. Many hydro builds include a Y-strainer or bag filter upstream of the miners to catch debris. Shut pump, relieve pressure, open the filter housing, inspect the screen or bag. Rinse scale or particulate off; replace the element if torn, discoloured, or older than the manufacturer's interval. Refill, bleed the high point, restart. A choked filter can drop flow `20–40%` before anything else looks wrong.

Inspect all quick-disconnect couplings on the miner and manifold. With pump off, cycle each QD: disconnect, inspect the internal valve for debris or stuck-open/stuck-closed failure, wipe with a lint-free cloth moistened in `99%` isopropyl, reconnect firmly. Listen for the click / seat confirmation. A partially-engaged QD coupling looks fine externally but flows at `30–60%` of rated — exactly the signature of a hot, high-delta-T miner on an otherwise-healthy loop. Zero-cost, resolves more `ERR_WATER_TEMP` complaints on recently-moved units than any other single step.

Verify no other miner on the same loop is starving this one. On a shared-loop deployment (multiple S19 Hydros parallel-plumbed off a single CDU), a failed or throttled neighbour can hoard flow and starve the others. Check loop-wide flow distribution if you have per-drop flow meters, or temporarily isolate (close QD) each other miner and watch this miner's temps. If isolating a neighbour fixes this miner, you have a flow-balance problem — valve it back partially, not fully, and balance by delta-T.

Measure loop pressure and confirm it's inside the design window. Many hydro builds include a pressure gauge on the CDU. Low pressure = air ingress or slow leak; high pressure = blocked downstream path or stuck check valve. Target and tolerance depend on the specific CDU and loop head; verify against the commissioning record. A steadily-dropping pressure over days/weeks is a slow leak — find it before it finds your electrical equipment.

Full drain and flush of the loop with a coolant swap. Highest-value advanced fix on a loop more than 12–18 months old or after any wrong-fluid event. Power off, shut pump, drain to a clean container (dispose per local regulations — glycol is not storm-drain material, especially in Canada). Flush with deionized water at pump minimum duty until flush runs visibly clean (usually 2–3 full volumes). Refill to MAX with spec coolant: deionized water for DI loops, or propylene glycol at labelled concentration (typically `30–50%` — verify against the CDU manual). Bleed the high point for several minutes. Monitor the first 24 hours — fresh loops commonly release trapped air.

Replace or overhaul the circulator pump. If diagnostics identified a failing pump (high-and-drifting amperage, cavitation audio, low flow with clean HX and no air), swap it. Shut pump, drain enough loop volume to get below the pump inlet, close isolation valves, unfasten mounting, disconnect electrical, install matched-spec unit (verify head curve and flow rating against the original — mining hydro loops are not a place for 'close enough'). Refill, bleed, verify amperage and flow at commissioning baseline.

Descale internal coolant passages and hashboard cold plates. If a DI flush didn't restore HX performance, you're looking at scale on internal passages — the consequence of a past wrong-coolant event or long-term tap-water top-ups. Power off, remove hashboards (track cold-plate connections and torque sequence), isolate each cold plate, circulate a proprietary descaler (e.g. citric-acid-based HVAC descaler) per its instructions, rinse with DI until flush runs clean and neutral pH, reassemble. Multi-hour procedure — worth doing only once lighter-touch options are exhausted.

Reapply thermal interface material between cold plates and hashboard ASICs. Factory TIM on Bitmain hydro boards pumps out and dries over `12–18 months` of continuous service (faster than air-cooled variants because hydro runs chips harder under tighter envelopes). Power off, remove the hashboard, carefully separate the cold plate from the ASIC array (follow Bitmain's torque pattern in reverse, track insulating shims by position), clean every `BM1398` / `BM1362` die top and cold-plate contact patch with `99%` isopropyl until shiny, apply a uniform thin layer of Arctic MX-6 or Thermal Grizzly Kryonaut, reseat in original torque sequence in two passes. Expect `5–10 °C` chip-to-water delta drop after TIM cures under load.

Cross-flash DCENT_OS for per-chip thermal telemetry and tuning control. Stock Bitmain firmware on S19 Hydro exposes chain-rollup temperatures and loop inlet/outlet, but does not surface per-chip thermal data or let you tune per-domain voltage. DCENT_OS (https://d-central.tech/dcent-os/) — D-Central's own open-source Antminer firmware, source at https://github.com/DCentralTech/DCENT_OS — adds per-chip HW% visibility, autotune, per-domain voltage, and Stratum V2, making hydro thermal forensics dramatically sharper than stock `bmminer`. Alternatives on Antminer hardware: Braiins OS+, LuxOS, Vnish. DCENT_OS is the Mining Hacker default — built by plebs, for plebs, open-source, no licensing fees. Back up pool and worker config before flashing anything.

Stop DIY and book a D-Central bench slot. Ship to D-Central when any of these is true: hashboard cold plate is suspected internally fouled and flushing didn't restore it; miner reads sensor values that disagree with an independent thermocouple by `> 5 °C` at a cold boot (sensor / PIC fault on a hydro manifold — SMD rework on a dense board with hydro-specific wiring); a leak test shows coolant egress from inside the hashboard cavity (destructive to adjacent electronics — do not run this unit); capacitor bulging, burnt-component odour, or visible corrosion from a prior leak event. Book at https://d-central.tech/services/asic-repair/ — Canada-wide, US, and international accepted.

D-Central hydro-bench process. Test fixture with programmable load and an isolated loop; per-chip isolation with official Bitmain test binaries; cold-plate removal, internal flush / descale, and TIM refresh; chip replacement with graded `BM1398` / `BM1362` salvage or new-old-stock where available; leak-test at `1.5×` working pressure for `≥30 minutes` before re-release; `24-hour` burn-in at nameplate on a monitored bench loop before ship-back. Hydro boards are more fragile in shipping than air-cooled — expect more packing and a slightly longer bench cycle than an equivalent S19 air-cooled job.

Pack safely before shipping. Drain the hashboards of coolant (open QDs, let gravity plus gentle compressed air clear residual fluid — trapped coolant in transit is a hazmat classification nightmare and gets packages refused at the border). Cap every coolant port with the Bitmain-supplied plug or a clean elastomer plug. Pack each board in an anti-static bag, double-box with at least `5 cm` of closed-cell foam on every side, include a note with observed symptoms, firmware version, loop coolant type and age, recent maintenance history, and contact info. That note saves our bench an hour of diagnostic time, which saves you money on the invoice.