Skip to content

We're upgrading our operations to serve you better. Orders ship as usual from Laval, QC. Questions? Contact us

Bitcoin accepted at checkout  |  Ships from Laval, QC, Canada  |  Expert support since 2016

NERDQAXE_HYDRO_BUBBLE Warning

NerdQAxe++ Hydro – Air Bubble in Coolant Loop

Air bubble trapped at high points of the coolant loop (cold plate or radiator) starves the BM1370 array of flow. Pump runs healthy but cold plate saturates; ASIC + VREG temps climb gradually over 5-15 minutes; firmware trips Overheat Mode at >=75 C ASIC. Different fix path than a dead pump — bleed first, parts later.

Warning — Should be addressed soon

Affected Models: NerdQAxe Hydro, NerdQAxe+ Hydro, NerdQAxe++ Hydro (BM1370 x4 water-block edition); any community NerdQAxe build running a self-mixed liquid cooling loop

Symptoms

  • Recent event: the loop was filled, refilled, topped up, drained, moved, tilted, or transported within the last 48 hours
  • Pump runs and sounds healthy (faint 50-200 Hz impeller whine present) but VREG temp climbs faster than ASIC temp on the dashboard
  • Dashboard shows Overheat Mode or temp_shutdown at >=75 C ASIC within 5-15 minutes of power-on, not the 30-90 seconds you would see with a fully dead pump
  • Clear tubing section shows intermittent bubbles cycling through, occasional gurgles, or a section that periodically goes empty for 2-5 seconds
  • Audible cavitation tick from the pump — a high-frequency rhythmic clicking that tracks pump speed (not motor whine)
  • Reservoir level keeps dropping over the first 30 minutes after fill even though there is no visible leak — air being driven out is leaving void behind
  • Inlet line to the water block warms up faster than expected; outlet stays cooler than inlet (a healthy loop has outlet 5-10 C warmer than inlet under load)
  • Coolant temperature at the reservoir is normal (<30 C) but ASIC and VREG temps climb anyway — the bubble is between the reservoir and the cold plate
  • Touch test on the radiator: the radiator core stays at near-ambient temperature even though the miner has been hashing for 5+ minutes — heat is not making it out of the cold plate
  • Hashrate falls progressively (not instantly) over the first 5-10 minutes from nameplate (~4.8 TH/s stock, ~6.0 TH/s OC) toward 0 H/s as thermal-capping kicks in
  • Bench-test of the pump (inlet in jug A, outlet in jug B at 12 V) shows healthy flow — eliminates pump as the failure
  • Tapping the chassis with a fingertip changes the dashboard temperature trend within 30 seconds (bubble dislodging into reservoir → flow restored briefly)

Step-by-Step Fix

1

Kill miner power; leave the pump running. Pull the 12 V input from the miner board only. The pump must keep circulating during the bleed — that is the entire point. If your wiring shares one rail between miner and pump, swap to a dedicated 12 V / 2 A adapter for the pump now and use this incident to fix the rail-share. Never try to bleed a loop with the pump off and gravity alone — the bubble does not move.

2

Tilt-and-bleed cycle. With the pump running, tilt the chassis to roughly 30 degrees in each of four orientations (front-up, back-up, left-up, right-up) for 30 seconds each. Bubbles migrate toward the reservoir cap as gravity and pump pressure conspire. Listen for gurgles through the cap as gas escapes. Top up coolant as the reservoir level drops. Repeat the full four-position cycle 2-3 times, or until the cycle no longer drops the reservoir level.

3

Reservoir top-up to MAX. Distilled water or non-conductive PC-loop coolant only. Never tap water — minerals deposit in the cold plate and never flush out. Never automotive coolant — wrong ion package, wrong viscosity, fluorescent dye that stains tubing. Top up slowly while tilting; this drives small bubbles out through the cap. The fill is done when the level holds steady through a full tilt cycle.

4

Visual confirmation: zero bubbles for 2 continuous minutes. With the pump running and the miner unpowered, watch the clear section of tubing under good light for two solid minutes. No 1-5 cm gas slugs traveling. No cyclic gaps. No rhythmic gurgles from the pump. Two minutes — set a timer, do not eyeball it. This is the gate before re-energizing the miner.

5

Re-energize the miner and watch the dashboard. Plug the miner's 12 V back in. Open AxeOS. Watch the temperature graph for 15 minutes. Healthy bleed: ASIC ramps to ~48-55 C and stabilizes; VREG follows by 5-10 C and stabilizes; coolant in reservoir reaches 28-38 C and stabilizes; hashrate locks to nameplate. If anything climbs past Overheat Mode thresholds (ASIC >=75 C), kill miner power and restart at Step 1.

6

Pump-only bench test before further bleeding. If the bleed did not clear after the first cycle, eliminate the pump as a variable. Disconnect the pump from the loop, drop inlet into one jug of distilled water, outlet into a second jug, energize at 12 V. A healthy 800 L/h - 1500 L/h pump transfers the first jug to the second in 2-5 minutes with visible force. Weak or non-existent flow → pump is dying, replace it (this stops being a bubble fix and becomes a pump-not-circulating fix).

7

Multimeter check on pump rail under load. Probe the pump's 12 V barrel connector while the pump is running. Expect >=11.8 V sustained. Below that, the supply is sagging or wiring is undersized — under-voltage is a cavitation contributor and a bubble accelerator. Swap to a known-good 12 V / 2 A adapter for the pump if rail voltage is low.

8

Leak hunt with paper towel and torch. Pump running, miner cool. Work the loop systematically — every fitting, every barb, every quick-disconnect, the reservoir cap, the pump body seal. A <1 drop per hour leak will return wet within a minute of wiping. The reservoir cap gasket is the most common culprit on community builds; pump-body O-rings are second.

9

Re-bleed with the chassis elevated higher. Some NerdQAxe Hydro chassis variants have an internal high point that does not bleed at 30-degree tilt — try 45 degrees for 1 minute per orientation. Add coolant slowly through the reservoir cap as bubbles escape. The rule of thumb: keep bleeding until a full four-orientation cycle stops dropping the reservoir level.

10

Drain-and-refill if the bleed does not clear after 30 minutes. Some bubbles, especially in old loops with biological growth or a clogged radiator, will not bleed without a full coolant replacement. Drain into a clean container (catch every drop — distilled coolant for inspection). Inspect for cloudy coolant, green algae, or sediment — if you see any, dispose and refill with fresh distilled. Refill slowly through the reservoir, run the pump-only bleed cycle from Tier 1 step by step.

11

Flash the latest stable shufps/ESP-Miner-NerdQAxePlus firmware. v1.0.29 introduced PID thermal control for hydro variants; current stable is v1.0.34 / v1.0.35 class. Use the shufps/nerdqaxe-web-flasher browser tool over the ESP32-S3 USB-C port. After flashing, set target ASIC temperature to 55 C, target VREG to 60 C, let PID handle the fans. If you have been chasing a 'bleed did not help' with pre-1.0.29 firmware, the firmware was the issue — air-cooled fan curve on a hydro board.

12

Install a YF-S201 flow sensor inline. Cut the tubing between pump outlet and cold plate inlet. Insert the sensor in flow direction (arrow on housing). Wire the 12 V and ground to your pump rail. Wire the signal pin to a 12 V automotive relay coil; route the relay's normally-closed contact through the miner's 12 V rail. When flow drops below the relay's pull-in threshold, the relay opens and cuts miner power before thermal damage starts. Total cost $8-25 CAD. Saves a $350-600 CAD board.

13

Replace fittings flagged in Step 8. A weeping fitting will not tighten its way to a permanent fix — it will re-leak in a month. Drain the loop, replace the offending fitting (G1/4 or barb to match your build), reassemble with PTFE tape on threaded fittings or a fresh O-ring on compression / quick-disconnect fittings. Refill, re-bleed.

14

Replace coolant with PG-25 or distilled + corrosion inhibitor on a long-life build. Plain distilled water is fine for 6-12 months but slowly grows algae if any light reaches the reservoir. PG-25 (25% propylene glycol, 75% distilled water) gives you longer service life, light algae resistance, and a slightly higher boiling margin — at the cost of ~5% lower thermal capacity than pure distilled. For a set-and-forget solo lottery miner, PG-25 wins. For a tuned overclock, distilled + an inhibitor (Mayhems X1 class, single-drop dose) is the compromise.

15

Reseat and re-paste the cold plate to the BM1370 chips. If thermals are still 5-10 C higher than expected after a clean bleed, the water-block-to-chip TIM has degraded. Drain the loop, remove the cold-plate mounting screws, lift the cold plate straight off without shearing sideways (a sideways lift tears BGA pads on BM1370). Clean old TIM with 99% isopropyl. Apply Arctic MX-6 in a small cross pattern across all four die. Reseat with even torque on all four corners. Refill and re-bleed.

16

Stop DIY when any of these is true: (a) ASIC temp sensor reads physically impossible values (negative, >150 C, or fixed at a single value) — sensor or solder joint damaged from a thermal event; (b) one BM1370 chip position reads 0 H/s at stock while the other three hash normally — that chip is dead; (c) visible scorching, lifted BGA pads, or discoloration on the PCB; (d) the cold plate or any fitting leaks coolant onto the board — secondary damage to the power stage starts within hours; (e) a third bleed in 30 days — there is a structural problem (micro-crack, dissolved-gas oversaturation, pump cavitation root cause) needing bench equipment.

17

D-Central bench process for NerdQAxe Hydro bubble events: hydro-specific bench jig with sealed test loop and calibrated flow sensor, leak-detection trace, pre-filled cold-plate integrity test, per-chip BM1370 hash verification, VREG regulator sanity (TPS546 family), firmware reflash to current stable shufps/ESP-Miner-NerdQAxePlus. Chip replacement uses NOS or graded refurb BM1370 stock when needed. Post-repair burn-in: 24 hours at nameplate with monitored thermals on the bench loop before return shipping.

18

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. Either send the cold plate still bolted to the chips (if you have not disturbed it) or remove it entirely with care — a half-removed cold plate shreds BGA pads in transit. Note on the shipping slip: observed symptom, recent fill / move events, firmware version, whether the miner ran past Overheat Mode for any sustained period, contact info. Saves diagnostic hours, which saves you money.

When to Seek Professional Repair

If the steps above do not resolve the issue, or if you are not comfortable performing these repairs yourself, professional service is recommended. Attempting advanced repairs without proper equipment can cause further damage.

Related Error Codes

Still Having Issues?

Our team of Bitcoin Mining Hackers has been repairing ASIC miners since 2016. We have seen it all and fixed it all. Get a professional diagnosis.