A Bitaxe stuck at 0 GH/s — or hashing far below its rated terahash — almost always traces to one of six faults: a dead or partially-dead BM1370/BM1368 ASIC, a missing or out-of-spec core-voltage rail, a cracked solder joint under the chip, an over-aggressive AxeOS overclock, a sagging 5 V supply, or a stalled clock. Here is how to tell them apart before you reach for a soldering iron.

Start by reading what AxeOS is actually telling you

The single biggest diagnostic advantage the Bitaxe has over a sealed Antminer is that the firmware (AxeOS, built by the open-source Bitaxe community — credit to skot and OSMU) reports everything you need on the web UI and the on-board OLED. Before touching the hardware, browse to the miner’s IP and write down these readings:

• ASIC model / chip detected — AxeOS enumerates the chip at boot. A single-chip Bitaxe expects exactly one ASIC. If the dashboard shows the chip model, the UART link and clock are alive. If it shows “no ASIC” or a zero chip count, the chip never enumerated — a deeper fault.
• Hashrate (GH/s) — the headline number. Compare it to the expected figure in the table below.
• ASIC temperature — a valid, climbing temperature means the chip has core voltage and is drawing current. A reading pinned at ambient (or an obviously bogus value) points at a power or chip-death problem.
• Core voltage vs. target — AxeOS shows both the requested and measured core voltage. A large gap means the regulator is not delivering.
• Input voltage — should sit near 5 V on most single-chip Bitaxes. A drooping input is a power-supply story, not a chip story.
• Frequency, fan RPM, power (W), and accepted/rejected shares — high rejects or “hardware errors” with a non-zero hashrate is a very different problem from a flat 0.

One important distinction: AxeOS speaks Stratum V1. A pool or network problem shows up as shares not being accepted while the chip still produces a local hashrate — that is not a true 0 GH/s and not a board fault. A genuine 0 GH/s means the silicon itself is not returning nonces.

The six causes of 0 (or low) GH/s

1. Dead or partially-dead ASIC (BM1370 / BM1368)

The Bitaxe Gamma carries a single Bitmain BM1370 (Antminer S21 Pro silicon, TSMC 5 nm); the Supra runs a BM1368 (S21-class, also 5 nm); the Ultra uses a BM1366 (S19 XP). These are big SHA-256 dies with roughly 1,280 small cores internally. Two failure shapes show up: a fully dead chip never enumerates (AxeOS reports zero ASICs), while a partially-dead chip enumerates and reports a temperature but returns far fewer valid nonces than it should — you see maybe 40–70% of rated hashrate with a stubbornly elevated hardware-error count. ESD, a power surge, or sustained thermal abuse are the usual killers.

2. Core-voltage fault (TPS546 power domain)

The Bitaxe regulates one core-voltage rail to the whole ASIC through a TPS546 PMBus buck converter, typically commanding around 1.1–1.25 V. Crucially, voltage is controlled per power domain, not per individual core — there is no way to “switch off” one bad internal core to rescue the chip; the rail feeds the entire die. AxeOS polls the TPS546 for over-current, over-voltage, under-voltage and over-temperature faults. If the measured core voltage reads near zero (or wildly off target) while input voltage is healthy, suspect the regulator, its feedback components, or a cold joint on the TPS546 — the ASIC simply cannot hash without its rail.

Worth noting: the S21-class BM1368 and S21 Pro BM1370 are no-PIC designs. Unlike older Antminer hash boards, there is no PIC microcontroller to reprogram, which removes a whole class of failure — but it also means a chip that won’t talk is talking to you about the silicon or the power rail, not a corrupted calibration controller.

3. Cold solder / cracked joint under the chip

The BM1370/BM1368 sits on a dense array of underside pads. Thermal cycling, mechanical shock during transport, or a marginal factory reflow can crack one of those joints. The tell is intermittency: the board hashes fine cold, then drops to 0 once it warms and the joint expands open — or a gentle press on the chip changes the hashrate. Per D-Central’s hash-board diagnostics, a cracked joint and a cold joint present identically to a dead chip from the firmware’s point of view, because a single broken signal pad stops the chip forwarding clock or UART. This is the most repairable of the serious faults.

4. Frequency / voltage set too aggressive in AxeOS

AxeOS does not ship a closed-loop autotuner — the frequency and voltage you type in are applied at runtime, not validated against your specific chip. A value that is rock-solid on one Gamma can hard-fail on another because every die has a different voltage/frequency curve. Push frequency too high (or starve it of voltage) and the chip either drops to 0 or hashes low with a flood of hardware errors and rejects. The fix is free: in AxeOS, return the chip to stock frequency and voltage (or click the conservative preset for your model), reboot, and confirm it stabilizes before tuning back up in small steps.

5. PSU sag / undersized 5 V supply

Most single-chip Bitaxes run from a 5 V barrel-jack or USB-C supply, and the TPS546 has a hard VIN turn-on threshold around 4.8 V and a turn-off near 4.5 V. A thin USB cable or a no-name 5 V brick droops under the chip’s current draw; when input dips below the threshold the regulator simply shuts the ASIC off, and you get 0 GH/s or a board that boots, hashes for a few seconds, then dies. AxeOS will show input voltage sagging under load. Cure it with a quality supply and a short, thick cable. (The Gamma Turbo and Hex variants instead run from a higher-voltage input — around 11 V — so confirm which supply your board expects.)

6. Stalled clock

The BM1370/BM1368 needs its 25 MHz reference to do anything at all. If the on-board oscillator is dead — cracked crystal, cold joint — the chip never enumerates and AxeOS reports zero ASICs, looking identical to a dead chip. A scope on the clock line distinguishes the two: present and clean means the chip is the problem; absent means the clock source is.

Expected vs. actual hashrate

Board	ASIC	Process	Rough stock hashrate
Bitaxe Gamma	1× BM1370 (S21 Pro)	TSMC 5 nm	~1.0–1.2 TH/s
Bitaxe Supra	1× BM1368 (S21)	TSMC 5 nm	~0.5–0.7 TH/s
Bitaxe Ultra	1× BM1366 (S19 XP)	TSMC 5 nm	~0.4–0.5 TH/s
NerdQAxe++ (multi-chip)	4× BM1370	TSMC 5 nm	~4–4.8 TH/s

If you are seeing roughly half of the figure above, you are likely looking at a partially-dead chip, thermal throttling, or an unstable overclock — not a fully dead board. A flat 0 with no temperature rise points at power or clock. Cross-check your symptoms against the ASIC Fault Finder and confirm your tuning targets against the ASIC Power Profiles database before changing anything.

Ordered diagnostic walkthrough

1. Check the supply first. Swap in a known-good 5 V supply and a short, thick cable. Watch AxeOS input voltage hold steady under load. This rules out the cheapest, most common cause.
2. Reset AxeOS to stock. Drop frequency and voltage to default, reboot, and see if hashrate recovers. Aggressive tuning is the second-most-common cause and costs nothing to eliminate.
3. Confirm the chip enumerates. If AxeOS shows the ASIC model and a sane temperature, the clock, UART and core rail are alive — your problem is tuning, a partial chip, or a joint. If it shows zero ASICs, you are chasing power, clock, or a dead chip.
4. Read core voltage vs. input voltage. Input healthy but core voltage near zero = TPS546 / power-domain fault. Both healthy but still 0 = clock or chip.
5. Test for intermittency. Run it cold, let it warm, and watch for a drop to 0 — or gently press the chip. A hashrate that changes with heat or pressure is a cracked/cold solder joint, which is reflowable.
6. If it still won’t hash and power/clock check out, the silicon is the prime suspect.

Reflow vs. chip-replace — making the call

Reflow makes sense when the symptoms scream a joint, not dead silicon: the board hashes cold and dies warm, hashrate responds to pressure on the chip, or it started misbehaving after a knock. A controlled hot-air reflow with proper flux — and fresh thermal interface on reassembly — can recover a cracked-joint board cleanly. This is low-risk, high-reward work if you have a hot-air station, magnification, and a steady hand.

Chip replacement is the answer when the chip is genuinely dead — it never enumerates, power and clock are confirmed good, and a reflow doesn’t bring it back. Replacing a BM1370 or BM1368 means desoldering a 5 nm S21-class die, cleaning and re-tinning the pads, fitting a verified replacement, and reflowing to a controlled temperature profile. On a single-chip Bitaxe the ASIC is most of the board’s value, so a chip swap only makes economic sense done right, the first time — and only with a chip confirmed good on a BM1370/BM1368 tester before it goes down. After any rework, treat it like the factory does: re-verify enumeration, then a stable hashrate at stock, then ramp.

When to send it to D-Central

If you’ve ruled out the supply and the AxeOS settings, and you don’t have a hot-air rework station, microscope, and a way to source and verify a replacement BM1370/BM1368, this is the point to stop and ship it. D-Central repairs Bitaxe and open-source boards in-house — we’ve done component- and chip-level ASIC repair in Laval since 2016, with the reflow stations, chip testers, and parts to bring a dead or underperforming board back rather than scrapping it. Start a repair here: D-Central ASIC repair.

If the board turns out to be beyond economical repair — or you simply want a known-good spare while yours is on the bench — the Bitaxe is hand-built and tested before it ships. For tuning references, firmware notes, and the rest of the open-source mining knowledge base, the Bitaxe Hub is the place to start. And if your symptom is something other than a flat 0 GH/s, work through the sibling guides on a Bitaxe that won’t power on or one that’s overheating and throttling.