NerdQAxe++ Hydro – Pump Not Circulating Coolant

Power off everything at the wall. Pull the 12 V inputs from both the pump and the miner. This is non-negotiable — every second under load with no flow accelerates thermal damage. Wait 5 minutes before opening anything so coolant in the water block has time to stop rising and the BM1370 domain can discharge safely. A hard breaker-level cut beats any AxeOS software reboot for clearing a thermal emergency.

Walk around the loop and inspect. Is the reservoir full to its MAX line? Is the clear tubing showing bubbles or still liquid? Are there visible kinks, pinch points, or obstructions? Is the radiator unobstructed and dust-free? A deliberate walk-around pass catches roughly 40% of no-flow events before anything more serious happens. Document anything you change so you can roll back if it makes things worse.

Check pump power delivery. Is the pump's 12 V adapter firmly seated in the wall outlet and in the pump itself? Is a separate fused power supply intact? If the pump shares a rail with the miner, an undersized 12 V / 4 A adapter sags under combined load — a dedicated pump supply eliminates that variable. Swap to a known-good 12 V / 2 A dedicated adapter for the pump if one is available on hand.

Top up coolant. Distilled water or a clear non-conductive PC-loop coolant only. Never tap water — minerals deposit in the water block and never come out. Never automotive coolant. Fill to the reservoir's MAX line. Tilt the chassis gently while adding coolant to flush small air pockets out through the reservoir cap. A half-full reservoir cavitates a pump within seconds of power-on.

Verify firmware is on a current stable build. AxeOS → System → Version. Cross-reference against shufps/ESP-Miner-NerdQAxePlus releases on GitHub. Pre-v1.0.29 builds do not include PID thermal control and will mis-manage a hydro board. OTA (or USB-C web-flasher at github.com/shufps/nerdqaxe-web-flasher) to the latest stable (v1.0.34 / v1.0.35 class at time of writing); betas (v1.0.36-beta2 and later) are optional for testers only.

Measure pump input voltage under load with a multimeter on DC. Probe at the pump's 12 V barrel connector while the pump is supposed to be running. Expect >=11.8 V sustained. Below that, the supply or the wiring is sagging and the pump cannot produce rated torque. Swap in a known-good 12 V / 2 A supply. If the sag persists, check for a partial short in the pump itself — a shorted winding pulls excess current and droops the supply, which is a pre-failure signature you can catch with the multimeter.

Pump-only bench test. Disconnect the pump from the miner's loop entirely. Drop the inlet into a jug of distilled water, the outlet into a second jug, and energize at 12 V. A healthy 800 L/h - 1500 L/h pump will transfer the first jug to the second in 2-5 minutes with visible force at the outlet. A dead or weak pump barely dribbles. This is the definitive pump diagnosis — do not skip it before deciding whether to replace the pump.

Full bleed with the chassis elevated. Run the pump standalone (still disconnected from the miner's 12 V input). Lift the chassis to a ~30-degree angle. Trapped air migrates toward the reservoir cap. Hold each orientation (front-up, back-up, left-up, right-up) for 30 seconds. Top up coolant as the level drops. The bleed is done when: zero visible bubbles in the clear tubing for 2 continuous minutes, reservoir level stable, no cavitation sound from the pump.

Inspect the loop for biological growth or corrosion. A loop running distilled water unattended for more than 6 months can grow algae, especially if light leaks onto the reservoir. Cloudy coolant, green tint, or slimy residue in the reservoir means: dump the loop, flush with 10% distilled white vinegar, flush twice with distilled water, refill with fresh coolant. Biological crud in a NerdQAxe water block is not economically removable, so prevention matters on a long-term install.

Flush the radiator. Disconnect the radiator from the loop. Connect a garden hose (low pressure) or a second pump to force distilled water + 1 drop of PC-loop detergent through it in both directions. Sediment, flakes, and rust will come out. Flush with distilled until the output runs clear. Reinstall, re-bleed following Step 8. A clogged radiator is the silent long-term killer of flow rate on NerdQAxe hydro loops that have been running more than a year.

Replace the pump if the bench test from Step 7 confirmed it dead. Source a 12 V DC aquarium or PC-loop pump rated 800 L/h - 1500 L/h (the minerbases.com setup guide's confirmed spec range for the NerdQAxe Hydro water block). Match the fitting size of your existing loop to avoid adapter plumbing. Bleed fully before connecting to the miner. Parts cost: $15 - $45 CAD budget, $60 - $120 CAD for a brand with documented MTBF (recommended for set-and-forget solo lottery miners).

Flash the current stable shufps/ESP-Miner-NerdQAxePlus firmware with explicit PID thermal control. v1.0.29 is the earliest build with PID; current stable (v1.0.34 / v1.0.35 class) is the right baseline. Use the NerdQAxe web flasher at github.com/shufps/nerdqaxe-web-flasher. After flashing, set target ASIC temperature to 55 C, target VREG to 60 C, let PID ramp fans automatically. Hydro thermals demand closed-loop control, not the fixed-fan-curve air profile.

Drop tuning to stock before re-testing flow. Core voltage back to stock (0.78 V - 1.15 V range, specific to your chip bin per diglucky.com's tuning guide). Frequency back to 485 MHz. Hashrate drops from ~6 TH/s back to ~4.8 TH/s — accept it temporarily. Run at stock for 30 minutes of stable hashing, confirm all thermals healthy, only then step overclock back up (+25 MHz per iteration, 20 mV per iteration, 20-minute stability observation each step).

Reseat the water block and re-paste the BM1370 chips if ASIC temps are still 5-10 C higher than expected with confirmed healthy flow. The NerdQAxe++ Hydro block is pre-filled at factory and not designed for field disassembly. Drain the loop, remove the block's mounting screws, lift the block straight off the four-chip array without shearing sideways (a sideways lift tears BGA pads). Clean old TIM with 99% isopropyl. Apply Arctic MX-6 in a small cross pattern across all four die. Reseat with even torque. Re-bleed the loop. This is the hardest Tier 3 step — skip to Tier 4 if you are uncertain.

Add a flow sensor or flow switch as an interlock. The tribal-knowledge Mining Hacker move: install a 12 V inline flow sensor (YF-S201 class, $8 CAD) between the pump and the water block, wired to a relay that cuts the miner's 12 V input when flow drops below threshold. The ESP32-S3 can read the sensor's Hall pulses via GPIO if you want firmware-level awareness; a simpler mechanical flow switch on the pump rail is the bulletproof option. Prevents repeat burnouts on loops running unattended.

Stop DIY when any of these are true: (a) ASIC temp sensors read physically impossible values (negative, >150 C, or fixed at a single number) — sensor damaged or solder joint open, (b) a BM1370 chip position reads 0 H/s at stock while the other three hash normally — that chip is dead, (c) visible scorching, discoloration, or lifted solder on the PCB near the BM1370 BGA, (d) the water block leaks onto the board at any mount point. Four-chip NerdQAxe boards with one dead chip are usable at 75% hashrate but are borrowed time.

D-Central bench process for NerdQAxe Hydro repairs: diagnostic on a dedicated bench jig, per-chip BM1370 hash verification, individual chip isolation, water-block integrity test on a sealed loop with flow sensor, VREG regulator sanity check (TPS546 family), firmware flash to the latest shufps/ESP-Miner-NerdQAxePlus stable. Chip replacement uses NOS or graded refurb BM1370 stock. Post-repair burn-in: 24 hours at nameplate with monitored thermals before return shipping.

Ship smart. Drain the loop completely before shipping — no residual coolant on the board. Bag the board in anti-static, double-box with >=5 cm foam on every side. Include the water block (if still attached to the chips and you have not disturbed it) or remove it only if you are certain it comes off cleanly — a half-removed block shreds BGA pads in transit. Note on the shipping slip: observed symptom, firmware version, whether the miner ran dry for more than 30 seconds, and your contact. Saves diagnostic hours, which saves you money.