A Bitaxe that overheats, throttles, or drops to zero hashrate has one of two root problems: a thermal fault (failed fan, dried paste, starved airflow, hot ambient) or a power-delivery fault (a struggling VRM/buck regulator, an unstable input rail, or an over-aggressive undervolt in AxeOS). This guide separates the two.

Heat or power? Read the symptom first

Bitaxe boards built on the open-source ESP-Miner/AxeOS firmware (credit to skot and the OSMU community for the design and tooling) expose almost everything you need in the web dashboard: ASIC core temperature, voltage regulator (VR) temperature, input voltage, current, and power. The pattern in those readings tells you which fault family you are chasing.

Symptom	Most likely cause	Family
Hashrate slowly tapers as temps climb, then resets	Cooling shortfall — fan, paste, or airflow	Thermal
Chip temperature spikes within seconds of boot	Heatsink not contacting die / no paste	Thermal
Fan spins but temps still rise under load	Dried paste or undersized airflow for the overclock	Thermal
Board reboots, AxeOS logs an overcurrent/overvoltage/undervoltage fault	VRM (TPS546) protection tripping	Power
Input voltage sags, hashrate erratic, random drops	Weak USB-C cable/PSU or input rail instability	Power
Zero chips detected, VR cold, no core voltage	Buck regulator / VCORE rail dead	Power
Stable but low hashrate after an undervolt	Core voltage set below what the frequency needs	Power (config)

Thermal faults and how to chase them

The single Bitmain ASIC on most Bitaxe boards — a BM1370 on the Gamma (Antminer S21 Pro silicon, TSMC 5nm, roughly 1.0–1.3 TH/s at ~21 W) or a BM1368 on the Supra (S21-class, 5nm, no-PIC, ~625–775 GH/s at ~15 W) — dumps all of its waste heat into one small heatsink. There is no thermal mass to hide behind, so cooling problems show up fast.

Fan failure and degraded airflow

A stalled or dying fan is the most common cause of throttling. Check the fan RPM in AxeOS first. If RPM reads zero or is wildly below its setpoint, the fan bearing, the connector, or the fan header on the board is the culprit. Confirm the fan curve is enabled and not flat-lined at a low duty — a fan curve that never ramps will let temps run away under an overclock.

Dried thermal paste and heatsink contact

If the fan is healthy but core temperature still climbs, suspect the thermal interface. Bitaxe heatsinks are clipped or screwed to the ASIC, and the factory paste dries out over months of thermal cycling. A repaste is the highest-value DIY thermal fix: power down, remove the heatsink, clean the old paste from the die and heatsink base with isopropyl, apply a thin even layer of fresh paste, and re-seat with even clip pressure. Poor contact or a tilted heatsink shows up as a hot corner on a thermal camera.

Ambient and enclosure

These boards are sensitive to the air you feed them. A closed shelf, a hot room, or a 3D-printed shroud that recirculates exhaust will push core temps up even on a perfectly pasted board. Give the intake clean, cool air and a clear exhaust path before blaming hardware.

Thermal targets and what protection does

Aim to keep the BM1370/BM1368 core comfortably under about 70 °C in normal operation, and keep the VR (regulator) temperature within its rating under sustained load. AxeOS enforces a hard floor: when the chip crosses the firmware’s overheat threshold (in the mid-70s °C range) it holds the ASIC in reset and drops core voltage to zero, then only resumes once the chip cools back under roughly 45 °C, restarting at a reduced frequency and ramping up again. So a board that “mines for a few minutes then dies and recovers” is almost always a cooling problem, not a dead chip.

Power-delivery (VRM) faults

On a Bitaxe the core voltage is produced by an onboard regulator, not a series string of domains like a big Antminer hashboard. Newer boards (Ultra, Supra, Gamma, GT) use a TPS546 digital DC-DC buck converter; older boards used a DS4432U I²C DAC regulator, and an INA260 monitors current and power. The TPS546 is the part to understand because it both sets VCORE and protects the chip.

What the regulator protects against

The firmware polls the TPS546 status register for four fault classes: overcurrent, overvoltage, undervoltage, and overtemperature. When AxeOS logs a power fault and reboots the board, one of those four tripped. That is a feature, not a defect — the regulator is shielding a 5 nm ASIC worth more than the board around it.

Input rail instability (the cheap fix first)

Single-chip boards (Max, Ultra, Supra, Gamma) run from a 5 V USB-C input; the TPS546 turns on around 4.8 V, so a thin cable or an underpowered USB-C supply that sags under load will trip undervoltage faults and cause erratic hashrate and random drops. The Gamma Turbo (GT) runs from a 12 V input instead. Before opening anything, swap to a known-good cable and a supply that comfortably exceeds the board’s wattage, and watch the input-voltage reading in AxeOS for sag under load.

VRM / buck regulator failure

If the input rail is solid but the board shows zero chips detected with no core voltage and a cold regulator, the buck stage itself — the TPS546, its inductor, or the surrounding passives — has likely failed. Symptoms include the board drawing excessive current at the input (a short downstream of the regulator) or no VCORE at all (an open). This is the failure that genuinely needs a meter and, often, a hot-air station.

Checking VRM output

With the board powered, measure the core rail at the inductor output / chip core test point. A healthy single-chip Gamma or Supra sits near its commanded VCORE — the TPS546 default command is around 1.2 V, within a 1.0–2.0 V configured range — typically landing near 1.1–1.25 V depending on what the autotuner has computed. Cross-check against the live VCORE, current, and power values AxeOS reports. A core rail near zero points to a shorted regulator or downstream short; a rail that reads high or absent points to an open in the buck stage.

How AxeOS overclock and undervolt interact with thermals

In AxeOS you set two things: ASIC frequency and core voltage. They are coupled — a higher frequency needs a higher core voltage to stay stable, and both raise power and therefore heat. The autotuning logic computes its operating point at runtime against the chip’s real behavior; these are not fixed presets, and the same model can land at different stable settings unit to unit. That is why blindly copying someone else’s numbers backfires.

Two failure modes come straight from this knob:

• Overclock-induced thermal throttle. Push frequency and voltage up and you may gain hashrate right up until the cooling can’t keep pace; the board then hits the overheat reset described above and your average hashrate falls. The fix is more cooling or a more conservative profile, not more voltage.
• Undervolt set too aggressive. Drop core voltage too far for the chosen frequency and the chip becomes unstable: rising hardware-error rate, stale shares, low effective hashrate, or outright dropouts — even though temperatures look great. Raise voltage a step or lower frequency.

For sane starting points and a feel for the voltage/frequency/power envelope of S21-class silicon, work from our ASIC power profiles database rather than guessing, then let the autotuner settle. If you’re still not sure whether you’re looking at a heat problem, a power problem, or a chip problem, the ASIC fault finder walks the symptom tree with you.

Field diagnostic walkthrough

1. Read AxeOS. Note core temp, VR temp, input voltage, current, power, fan RPM, and any logged fault. The pattern usually names the family before you touch hardware.
2. Rule out the input. Swap to a known-good cable and an ample 5 V (or 12 V on GT) supply. Watch for input-voltage sag under load.
3. Verify cooling. Confirm fan RPM and an active fan curve; ensure clean intake and a clear exhaust path.
4. Repaste if temps run hot. Fresh thermal paste and even heatsink pressure fixes most “throttles after a few minutes” boards.
5. Back off the tune. Lower frequency/voltage to a conservative profile; if instability or faults clear, your earlier settings were the cause.
6. Measure VCORE. If chips still won’t enumerate and no fault is input-related, meter the core rail at the inductor output against the commanded VCORE.

DIY versus send it to a bench

Do it yourself: fan replacement, repasting, airflow and enclosure fixes, swapping the input cable/PSU, and dialing back an over-ambitious overclock or undervolt in AxeOS. None of that needs more than a screwdriver, isopropyl, paste, and patience.

Send it to a bench: a failed TPS546 or buck stage, a board that draws excessive current or shows a downstream short, a dead VCORE rail, suspected ASIC damage after a thermal-runaway event, or any fault that needs hot-air rework, microscope inspection, and component-level diagnosis. VRM-level repair on a single-chip 5 nm board is unforgiving — pull the chip’s voltage wrong and you can finish off a part the protection circuit was trying to save.

Get it repaired or replaced

D-Central repairs Bitaxe and open-source boards in-house — we’ve run an ASIC repair bench in Laval since 2016, so VRM rework, regulator replacement, and chip-level diagnosis on Bitaxe-class boards are routine here rather than a guess. If a board is beyond economical repair, or you simply want a fresh, dialed-in unit, see the Bitaxe. For setup, tuning, and the rest of the open-source mining picture, the Bitaxe Hub is the place to start. Chasing a different symptom? Cross-check our companion guides on a Bitaxe that won’t hash or submit shares, a Bitaxe that won’t power on or boot AxeOS, and low hashrate and chip degradation.