Bitaxe – Overclock Errors High / Must Reduce Frequency

Reduce ASIC frequency by 50 MHz from the current setting in AxeOS Settings → ASIC Frequency. Save and wait 5 minutes for thermal and statistical settling. Watch the HW Errors field for 30 minutes. The single most-effective fix when overclock has gone past the chip's silicon-lottery ceiling. If HW Errors falls below 1.5%, you are inside the silicon lottery again; from here walk frequency back up in 25 MHz steps to climb closer to the ceiling without crossing it.

Reset to your variant's factory tune and observe for 30 minutes for a fresh baseline: Supra/Hex 485 MHz at 1.25 V, Ultra 485 MHz at 1.20 V, Gamma/GT 525 MHz at 1.20 V. From here, change exactly one variable at a time. The most common OC-troubleshooting failure is moving frequency, voltage, and fan curve simultaneously — you cannot isolate which change fixed or broke anything.

Update AxeOS to the latest stable release. Download the correct .bin for your chip family (BM1366 for Ultra, BM1368 for Supra/Hex, BM1370 for Gamma/GT) from the ESP-Miner GitHub releases page. OTA-flash via AxeOS Settings → Firmware Update. ESP-Miner ships V-F tuning improvements on most releases; some v2.4.x builds had a confirmed regression where HW Errors reporting was over-counted versus actual chip behavior.

Verify fan RPM in AxeOS dashboard — anything below ~3000 RPM under load on the stock fan indicates a bad fan. Visually inspect that the heatsink is flush against the chip; a tilted or pad-shifted heatsink is a common cause of phantom HW% increases that look exactly like silicon-lottery overrun. Reseat with fresh thermal paste if any doubt.

Disable any tuning extras: presets, dynamic tuning, autotune, community auto-tune scripts. Revert to the manual frequency and voltage fields only. Some autotune scripts hunt aggressively for hashrate without HW% guardrails and will keep climbing past the chip's ceiling. If you use one, configure its HW% back-off threshold (typically 1.5%) before re-enabling.

Multimeter on DC, probes at the barrel jack (Supra/Ultra/Gamma) or XT30 (GT/Hex) while the miner hashes at full tune. Supra/Ultra/Gamma must read ≥ 5.10 V; GT/Hex must read ≥ 12.0 V. Below = PSU sag. The same PSU that boots the miner can choke under sustained OC current draw — that brownout-adjacent condition mimics the silicon ceiling. Swap to a meatier PSU (60 W+ Gamma, 65 W+ GT, 200 W+ Hex) and retest.

Run the bench-sweep grid: spreadsheet with frequency on rows, voltage on columns. Hold each combination for 30 minutes. Record Average Hashrate, HW Errors, and wall-side power draw. Plot it. The cell with the best efficiency (W/TH) at HW% under 1.5% is your tune. There is no shortcut — every hour spent on this grid is a year of stable mining you would not otherwise have.

Tune voltage in 5 mV steps, not 25 mV. AxeOS allows fine voltage control on TPS546 rails. Your chip might do 588 MHz at 1.215 V cleanly while needing 1.225 V only at 600 MHz. The 5 mV granularity matters at the margin — enough to flip a chip from marginal HW% to clean HW% without burning the extra power that 25 mV demands.

Improve airflow before adding voltage. Adding voltage to fix HW% is correct only if thermals allow it; on a 70 °C chip, more voltage just pushes you to 75 °C and triggers thermal throttle. Order is non-negotiable: cool the chip first (mesh stand, better fan, lower ambient, fresh paste) — then add voltage if HW% still demands it.

Mount the Bitaxe on a D-Central Bitaxe Mesh Stand. D-Central manufactured the original Mesh Stand — we were the first shop to make them. Elevated airflow under the PCB drops PCB and ESP32 surface temps by 5-10 °C measurably, which translates to 25-50 MHz of free frequency headroom on most chips at the same voltage. Heat is the enemy of silicon-lottery margins.

Replace stock thermal paste with Arctic MX-6 or Thermal Grizzly Kryonaut. Stock paste degrades over months of thermal cycling; a fresh layer drops chip junction temp 3-8 °C measurably. Same effect as airflow improvement — more frequency margin at the same voltage. Clean old paste with isopropyl 99% before reapplying a thin uniform layer.

On Hex/GT, identify the weak-link chip using AxeOS v2.5.x+ per-chip HW% visibility. Run your tune for 60 minutes and observe which chip dominates board HW%. Lower board-wide frequency to that chip's ceiling, or — if the weak chip is wildly out of family — accept its HW% as a tax. On a Hex with one weak chip out of six, a 50 MHz board-wide reduction can bring HW% from 4% down to 0.8% with only ~7% hashrate loss.

For advanced V-F tuning, build ESP-Miner from source with a custom voltage table. Clone bitaxeorg/ESP-Miner, modify chip-family tables in main/asic/, rebuild, OTA-flash. This lets you tune below the AxeOS web UI's voltage floor or above its ceiling. High brick risk — back up your NVS partition first. Worth it only if you've identified that your specific chip wants a voltage the stock UI does not expose.

Run a 24-hour stability burn-in at your final tune. A 30-minute test passes most days; a 24-hour test catches heat-soaked failures — the chip clean at 55 °C that drifts at 65 °C summer-evening conditions. Log /api/system/info every 5 minutes via a Pi or laptop cron. If HW Errors stays below 1.5% across 24 hours including the warmest part of your day, the tune holds. If it climbs in the evening, back off 25 MHz for seasonal margin.

Use a smart-plug power logger (Kasa KP125M, Shelly Plus Plug S) to measure efficiency at the wall. Watt-hours over 24 hours divided by your average TH/s × 24h gives your true W/TH. This is the number that matters for pleb-mining economics — not dashboard hashrate, not nameplate. Real-world W/TH × your power cost = the actual cost of every block-finding lottery ticket.

If stock-tune HW% is above 1% on a fresh Bitaxe, or if HW Errors climbs without any tune change over weeks, treat as hardware — not OC. Book a Bitaxe diagnostic with D-Central. We've been in the Bitaxe ecosystem since unit #1: manufactured the original Mesh Stand, built the first Bitaxe and Bitaxe Hex heatsinks, maintain the full ESP32-S3 / BM1366 / BM1368 / BM1370 diagnostic chain in-house. Include serial log dump, AxeOS firmware version, and your bench-sweep grid notes — saves 30-60 minutes of bench time and reduces repair cost.

Ship safely. Anti-static bag, double-boxed, foam on every side, ≥5 cm of crush protection. Include a one-page note: variant, AxeOS version, stock-tune baseline HW%, the OC tune you tried, the symptom, and your contact. Canada-wide, US, and international shipping accepted. Bitaxe diagnostic turnaround 5-8 business days.