The Bitaxe is a single-ASIC, open-source Bitcoin miner. An ESP32-S3 controller drives one Bitmain BM1370 (Gamma) or BM1368 (Supra) 5nm SHA-256 chip through a TPS546 power stage, with an EMC2101-managed fan, an OLED, and USB-C input. This page maps every section, what it does, and how it fails.

Treat this as the wiring diagram behind the rest of our Bitaxe repair library. When a symptom guide says “check the VRM” or “the ASIC isn’t enumerating,” this page tells you which component that is, where it sits in the signal chain, and which sub-guide goes deep. D-Central has repaired ASIC hashboards in-house in Laval since 2016, and the open-source Bitaxe — credit to skot and the Open Source Miners United community — is one of the most service-friendly boards ever shipped. Almost every fault on it is diagnosable with a multimeter, a thermal camera, and patience.

The five functional blocks of a Bitaxe board

Every single-board Bitaxe (Ultra, Supra, Gamma) follows the same topology. Power comes in over USB-C, gets stepped down by a buck regulator to a low-voltage, high-current core rail, and feeds one Bitmain ASIC. An ESP32-S3 microcontroller runs AxeOS, talks to the ASIC over UART, manages the power stage and fan over I2C, and paints the OLED. Understanding which block owns a symptom is 80% of the repair.

1. ESP32-S3 controller — the brain

The ESP32-S3 is a dual-core Xtensa LX7 SoC running at 240 MHz with onboard Wi-Fi. It hosts the entire AxeOS firmware: the web dashboard, the Stratum client, the pool connection, and the device’s two serial roles. On the Bitaxe it drives two logical UART paths — a control channel for GPIO and I2C tunneling, and a data channel for direct ASIC serial traffic. The ESP also owns the ASIC’s active-low reset line (nRST): hold it low and the chip is held in reset; release it high to bring the ASIC out of reset before initialization.

Common failure modes: bricked or corrupt AxeOS flash (device won’t boot, no AP, no web UI), Wi-Fi association failures, USB enumeration problems on the programming port, and bad firmware flashes. Most ESP-side problems are recoverable in software — a clean AxeOS reflash over USB resolves the majority. Hardware ESP failures (a physically dead module) are rarer and usually trace back to a power or ESD event. Firmware recovery is covered in our Bitaxe firmware recovery guide.

2. The Bitmain ASIC — where hashing happens

The single SHA-256 chip is the heart of the board and the most heat- and current-stressed component. Depending on model it is a BM1366 (Ultra), BM1368 (Supra), or BM1370 (Gamma) — all TSMC 5nm dies harvested from the same silicon families as Bitmain’s S19 XP / S19k Pro, S21, and S21 Pro respectively. The BM1368 and BM1370 are “no-PIC” designs: unlike S9-era boards there is no separate PIC microcontroller gatekeeping the hashboard, so the ESP32-S3 talks to the chip directly over UART. The ASIC core runs at roughly 1.0–1.2 V and pulls the lion’s share of the board’s current; voltage is regulated per power domain, not per chip — the S21-class die stacks several series voltage domains internally and the regulator sets the core rail feeding them.

Common failure modes: the ASIC enumerates but produces zero or unstable hashrate, high hardware-error (HW) percentages, the chip running far hotter than the board reports, or no chip detected at all on the data UART. Chip-level faults — cracked solder balls under the BGA, a reflow needed, or a genuinely dead die — are the line between a DIY fix and a bench job. Diagnosis and the replace-vs-retire decision are in our Bitaxe ASIC chip failure guide, with stock and tuned operating points in the ASIC power profiles database.

3. VRM / power stage — TPS546 buck regulator

The Bitaxe uses a Texas Instruments TPS546-class buck regulator (the TPS546D24A on the Gamma) at I2C address 0x24 to convert the USB-C input down to the ASIC core voltage. It is a fully digital PMBus part: AxeOS sets the output voltage, current limits, and protection thresholds in firmware at runtime — nothing here is jumper-set. On the Gamma the firmware brings it up around 1.15 V with overvoltage protection at 125%, an overcurrent fault limit near 30 A, and an over-temperature fault at 145 °C. The same I2C path reads back input voltage, output voltage, output current, and regulator temperature, which is how AxeOS shows live power draw.

Common failure modes: the regulator faults and latches off (board powers but ASIC gets no core voltage), phantom over-voltage or CML communication faults on the PMBus, missing or wrong output voltage, and — in the worst case — a shorted power stage that trips your USB-C supply. The VRM is the single most common hard-failure point on heavily overclocked Bitaxes because it carries the full core current. Probing the core rail and reading PMBus faults is covered in the Bitaxe power & VRM repair guide.

4. Thermal management — EMC2101 fan controller + sensors

Cooling is managed by an EMC2101 fan controller on the I2C bus. It generates the fan PWM (a 6-bit duty register), reads fan RPM back through a tachometer input, and exposes an external temperature channel tied to the ASIC plus its own internal sensor. AxeOS polls these every few seconds, drives the fan curve from ASIC temperature, and will throttle or shut down on an over-temp event. The TPS546 contributes a second temperature reading from the power stage.

Common failure modes: a stalled or dead fan (no RPM tach, rapid thermal climb), a fan spinning at 100% regardless of load (controller or tach fault), dried-out or detached heatsink thermal interface causing the chip to overheat while the board reports a plausible temperature, and dust-clogged fins. Overheating is the number-one cause of premature ASIC wear, so treat any thermal warning seriously. Full thermal triage is in the Bitaxe overheating guide.

5. I/O and the rest — OLED, USB-C, level shifting, headers

The OLED is an I2C display showing hashrate, temperature, and pool status; it is purely informational — a dead OLED does not stop mining. USB-C is the power and programming interface; a single 5 V USB-C supply with adequate current is the usual input, and an undersized adapter is a frequent cause of brown-out instability that looks like an ASIC fault. Because the BM13xx chip I/O runs at 1.8 V while the ESP32-S3 GPIO is 3.3 V, the board includes level shifting on the ASIC serial lines — a subtle failure point if a line gets damaged. The board also carries the fan header and the programming/serial header.

Common failure modes: a blank OLED (usually cosmetic — bad display, loose connection, or an I2C address conflict), USB-C connector fatigue or a bad cable, brown-outs from a weak supply, and damaged serial/level-shifter lines that cause “no chip detected” even though the ASIC is fine. If the board powers but never hashes, start with our Bitaxe no-hashrate walkthrough before suspecting the ASIC.

Bitaxe ASIC chip reference table

Each Bitaxe generation maps to a specific Bitmain die and its industrial sibling. Hashrates below are approximate stock figures — AxeOS lets you tune voltage and frequency at runtime, so real-world numbers vary with cooling and the power profile you choose.

Bitaxe model	ASIC	Process node	S21-family lineage	Cores / chip	Approx. stock hashrate
Bitaxe Ultra	1× BM1366	TSMC 5nm	S19 XP / S19k Pro silicon	~894 small cores	~0.4–0.5 TH/s
Bitaxe Supra	1× BM1368	TSMC 5nm	S21 / T21 silicon (no-PIC)	~1280 (80 big × 16 small)	~0.7 TH/s
Bitaxe Gamma	1× BM1370	TSMC 5nm	S21 Pro silicon (no-PIC)	~1280	~1.0–1.2 TH/s
Bitaxe Gamma Turbo	2× BM1370	TSMC 5nm	S21 Pro silicon (no-PIC)	~1280 each	~2.0–2.5 TH/s
NerdAxe / NerdQAxe (community)	Multiple BM1366 / BM1370	TSMC 5nm	S19 XP / S21 Pro silicon	Varies (multi-chip)	Varies by board

Two accuracy notes worth pinning down, because the internet gets them wrong constantly: every one of these dies is TSMC 5nm, not 3nm, and the BM1368/BM1370 are no-PIC parts — there is no S9-style PIC heartbeat chip to defeat. Voltage tuning acts on the power domain via the TPS546, and AxeOS autotuner targets are computed at runtime from live telemetry, not loaded from a preset table.

Diagnosing from a symptom: where to go next

Once you know the block, route the symptom to the right sub-guide. No web UI or no boot → ESP32-S3 / firmware. Powers on but no hashrate or “no chip detected” → serial/level-shifting then the ASIC. Hot, throttling, loud, or fan not spinning → thermal/EMC2101. Board powers the supply but the ASIC gets no core voltage, or the supply trips → VRM/TPS546. For a fast, structured triage across every fault class, use our ASIC fault finder, and pair it with the ASIC power profiles database when you want known-good voltage and frequency targets for your chip.

DIY or send it in?

The honest line: AxeOS reflashes, fan swaps, thermal-paste re-pastes, cable and supply fixes, and OLED replacements are well within reach for a hardware-fluent owner with basic tools. Reballing or replacing a 5nm BGA ASIC, rebuilding a shorted power stage, or chasing an intermittent fault under load is bench work that needs hot air, microscopy, and a current-limited supply. If you’re past the easy fixes, D-Central repairs Bitaxe and open-source boards in-house — send it to us at our Laval shop and we’ll diagnose it properly. Start a repair at d-central.tech/asic-repair.

If a board is genuinely dead and not worth the rebuild, a fresh unit is often the better call — see the Bitaxe for a current build, and the Bitaxe Hub for the full library of guides, tuning notes, and the rest of this repair series. We stand on the shoulders of the open-source Bitaxe community; our job is to keep these little miners hashing.