We model a 1 MW farm (approx. 280 M50 units).
Interpretation: While high-performance firmware appears most profitable, it fails sensitivity analysis when power >$0.08/kWh or BTC price drops >30%. Low-power custom firmware offers the best risk-adjusted return. whatsminer custom firmware
| Firmware | Avg Hashrate (TH/s) | Power (W) | Efficiency (J/TH) | Temperature (°C) | Rejection Rate (%) | | :--- | :--- | :--- | :--- | :--- | :--- | | Stock 2.5.4 | 70.2 | 3450 | 49.1 | 68 | 1.2 | | Vnish v5.0 (High Perf) | 84.5 | 4100 | 48.5 | 81 | 2.8 | | Asic.to v3.1 (Low Power) | 65.1 | 2800 | 43.0 | 62 | 0.9 | | LuxOS v23.10 (Auto-Tune) | 77.8 | 3540 | 45.5 | 74 | 1.5 | We model a 1 MW farm (approx
MicroBT’s Whatsminer dominates approximately 35–40% of the SHA-256 ASIC market (as of 2025). Stock firmware, while stable, prioritizes conservative thermal envelopes and fails to exploit silicon lottery variations. Third-party developers have therefore released custom firmware (e.g., Vnish, Asic.to, LuxOS for Whatsminer) that reconfigures the kernel-level control of the BM1397, BM1398, and BM1366 chips. This paper asks: Under what conditions does custom firmware deliver net positive ROI? $0.08/kWh or BTC price drops >