New Multi-Option Firmware: A Comprehensive Guide

Bitmain’s multi-option firmware (sometimes called “universal” or “security” firmware) is the factory image that adds a Working Mode drop-down to the S9-family web interface, letting you trade hash rate against power draw without touching the hardware. Bitmain released four of these images — one each for the S9, S9i, S9j and T9+; you flash the one that matches your model, pick a mode — overclock for more terahash, underclock for lower watts — and the firmware retunes the boards on the next reboot. It does not support the small run of S9 units built on the C5 control board: flash a C5 board with this image and the upgrade is skipped, the version string stays put, and nothing changes. This guide covers what each mode does at the silicon level, the power-supply reality overclocking demands, and how to tune safely.

What the modes actually change

Every S9-class miner — S9, S9i, S9j and T9+ — runs Bitmain’s BM1387 ASIC: 63 chips per hash board, three boards, 189 chips in a full unit, wired into 21 voltage domains per board in series. The BM1387 is rated near 75 GH/s at roughly 98 J/TH stock. Voltage is regulated per domain, not per chip, so every “mode” is really a coordinated change to two knobs: the PLL clock the chips run at, and the domain rail voltage feeding them.

• Normal — factory frequency and voltage, untouched. The baseline you always start from.
• HashRate +0.5 TH/s to +2.0 TH/s — raises the PLL frequency above the factory ideal to lift hash rate. Voltage tracks up to keep the chips stable, so current per domain climbs and so does heat.
• HashRate Unchanged — Lower Voltage — trims domain voltage (and slightly the frequency) while holding the ideal hash rate, shaving watts for the same terahash. The efficiency-first choice for a heater or a tight power budget.
• HashRate -0.5 TH/s to -3.0 TH/s — drops voltage and frequency together to cut both hash rate and power consumption.
• Low Power Enhanced Mode (S9/S9i/S9j only) — the deepest undervolt. Hash rate falls by roughly 4.5 TH/s on the S9 and ~3.5 TH/s on the S9i/S9j, in exchange for the lowest draw the firmware offers.

Pre-flight: do not flash a sick miner

The multi-option firmware assumes healthy hardware. Retuning frequency and voltage on a board that is already marginal turns a warning into a dead chain. Do not upgrade or overclock if any of the following is true:

1. Real-time hash rate deviates more than 5% from the ideal for the model.
2. The miner is missing a hash board, or is showing missing/dead chips (“asic=0” or an x on any chain).
3. Hash rate is unstable or swinging.
4. The power supply is below 1400 W or its output is unstable.

Confirm chip count and chain health first — the kernel log should report the expected 63 chips on all three chains at a steady temperature. If a chain is short, fix the hardware before you touch firmware; overclocking will not resurrect a dead domain, it will only stress the survivors.

The power supply reality overclockers miss

The stated PSU numbers are a trap if you ignore input voltage. Bitmain’s own guidance is: for S9/S9i/S9j a +0.5 TH/s overclock needs a 1400 W supply, while +1.0 / +1.5 / +2.0 TH/s need 1600 W. For the T9+, a 1600 W supply carries up to a 12 TH result after over-frequency, and you need 1800 W once the target reaches 12.5 TH or more.

Here is what the tables omit: the APW3++/APW7 supplies these miners ship with only reach their high wattage on a 220–240 V feed. On 110/120 V they run at reduced output — roughly the ~1200 W class — not enough headroom for any overclock option. If you are on North American 120 V household power you cannot hit these numbers; move the unit to a 240 V circuit before selecting anything above Normal. Even inside the recommended range, variance between supplies and boards means a nominally-adequate PSU can still fold under load — the wattage rating is a floor, not a guarantee.

Tuning procedure — climb, don’t leap

Overclocking is iterative. Set a mode, let it soak, read the result, then move one step:

1. Baseline in Normal. Run the miner in Normal until the hash rate is stable and the boards are at steady temperature. If it is not stable here, stop — it will not be stable overclocked.
2. Step to +1.5 TH/s. Select the +1.5 TH/s option and let it soak. Watch board temperatures, per-chain hash rate, and hardware-error rate for at least a couple of hours.
3. Push to +2.0 TH/s only if +1.5 held. If +1.5 is rock-solid and your power supply and room temperature have margin, you may try +2.0 TH/s.
4. Back off if it wobbles. If +1.5 is unstable, drop to +1.0 or +0.5 TH/s. Downgrade any time the PSU can’t hold the option, on any power fault, on over-temperature protection (real-time hash rate reads 0), or if boards/chips drop out.

Apply the same mode across a fleet with a batch configuration tool — but tune one representative miner to a known-good option first, then push that setting out.

Verify the result

After an over-frequency option, expect the landed hash rate within roughly ±0.5 TH/s of target — normal, not a fault. A 14 TH miner set to +2 TH/s may settle anywhere from about 15.5 to 16.5 TH; either is fine. Confirm a clean result: all 63 chips per chain present, hardware-error rate low, board temperatures in spec, and the accepted-share rate at the pool matching the new hash rate over a full day — instantaneous hashboard numbers lie; the pool-side average is the truth.

Common mistakes

• Overclocking on 120 V. The single most common failure: the PSU simply cannot deliver, the miner throttles or resets, and the operator blames the firmware.
• Flashing a board with dead chips. A short chain fails the pre-flight for a reason. Retuning kills it faster.
• Flashing a C5 control board and expecting change. The image self-skips; that is by design, not a failed download.
• Undervolting past chip stability. Low Power Enhanced Mode is the floor — below a domain’s minimum stable voltage you get rising hardware errors, not free efficiency.

When to escalate

If a miner drops chains, refuses to hold any mode above Normal, or trips over-temperature protection at stock, the firmware is not your problem — the hardware is. That points to a failing hash board, degraded thermal paste, a weak PSU, or a bad domain — a bench job, not a settings job. Diagnose the chain and chip fault before you flash again.

Related: Trace a dead or unstable chain with the ASIC fault finder, pull the correct signed image and model documentation from the miner manuals hub, and if a board won’t hold stock after tuning, start a repair and let the bench take it from there.