ASIC PSU Sizing Guide: Power Supply Math for Miners

A PSU that is too small trips offline under load. A PSU that is too large wastes capital and runs at lower efficiency. Getting asic psu sizing right is the difference between a setup that hums for years and one that throws thermal faults every August. The math is not complicated — rated power times a headroom factor, divided by available circuit capacity — but it depends on details that miner-spec sheets do not always make obvious. This guide walks the calculations, the voltage trade-offs, and the redundancy choices for current-generation Bitmain, MicroBT, and Canaan units.

The basic sizing formula

Start with the manufacturer’s rated wall-draw, not the nameplate hashboard wattage. A Bitmain S21 rated at 3500 W at the wall is the number to size against, not the 3300 W or so that the hashboards alone consume. Multiply rated wall-draw by 1.15–1.20 to give the PSU headroom. A 3500 W miner sized at 1.20 headroom wants a 4200 W PSU; at 1.15, a 4025 W PSU. The headroom matters because real-world draw exceeds nameplate during cold-start and during firmware-driven boost modes, and because PSUs lose efficiency as they approach 100 percent of rated capacity.

For miners with integrated PSUs (most current-gen Bitmain and MicroBT units), the manufacturer has already done the sizing — the PSU shipped with the unit is matched. The math matters when sourcing replacement PSUs, when running miners on aftermarket supplies, or when sizing the upstream circuit. Aftermarket PSUs from Bitmain (APW12, APW171) and from MicroBT (P21G, P29) are sold separately and need to match the miner’s required input voltage and amperage profile.

Why 240V matters: efficiency and amperage

Most modern Bitcoin ASICs accept both 200–240V and lower voltages, but the efficiency curve is meaningfully better at the higher end. A Bitmain APW12 PSU is rated around 93 percent efficiency at 220V; at 110V the same supply drops to 87–90 percent. The difference is not academic — a 3500 W miner running at 90 percent versus 93 percent efficiency wastes an extra 100+ watts as heat, every hour, all year.

The amperage math also favors 240V. The same 3500 W draws roughly 14.6 A at 240V or 31.8 A at 110V. A standard residential 20 A circuit at 110V cannot support a single S21 at full power; at 240V it has room for one S21 with margin to spare. The dedicated guide on 240V circuit sizing for ASIC miners covers the electrical-side detail and code references.

Operators wiring a home or garage setup almost always benefit from a 240V circuit dedicated to mining. A standard 240V/30A circuit supports up to two current-gen Bitcoin ASICs comfortably; a 240V/50A circuit supports three or four with a PDU. Larger setups move to sub-panels, three-phase, or both — covered in sub-panel installation and three-phase power for mining farms.

Reading the PSU spec sheet honestly

PSU ratings can be confusing. A “3600 W PSU” might be rated at 3600 W only when fed 240V; the same unit might be rated 2800 W on 200V or 2400 W on 110V. The Bitmain APW12 datasheet, for example, shows different maximum output curves depending on input voltage. Read the rating that applies to the voltage you will actually feed it, not the headline number.

Also check the certification. 80 Plus Platinum or Titanium ratings mean the PSU hits a defined efficiency at specific load percentages. Mining PSUs are usually not 80 Plus certified — they are purpose-built and run at different efficiency curves. The manufacturer’s own efficiency chart is the source to trust for sizing, not generic 80 Plus tier comparisons.

Single PSU versus dual PSU configurations

Some larger ASICs ship with two PSUs in parallel. A hydro unit like the S21 Hyd or an immersion-ready S21 XP Immersion needs the dual-PSU configuration to meet peak draw. The two supplies share load, and most configurations tolerate the failure of one PSU at reduced hashrate rather than tripping the whole miner. For air-cooled single-PSU units, a failed PSU takes the whole miner offline immediately.

Headroom: why 1.15–1.20 and not 1.50

Oversizing PSUs costs money up front and drags efficiency. A 5000 W PSU running a 3500 W load is at 70 percent of capacity — usually the efficiency sweet spot for the unit, but the capital cost premium over a 4200 W PSU is not always justified. A 1.15–1.20 headroom factor balances: it covers cold-start spikes, sustained boost-mode operation, and the manufacturer’s nameplate-versus-real-world variance, without paying for capacity that will never be used.

The exception is operators who run custom firmware with aggressive overclock profiles. A miner pushed to 110 percent of nameplate hashrate may draw 115–120 percent of nameplate power, and on top of that needs PSU headroom. In that case sizing PSUs at 1.30–1.40 of nameplate is the safer call. The trade-offs are unpacked in the broader piece on custom ASIC firmware risks.

Redundancy: when N+1 PSU is worth it

For single-unit home setups, PSU redundancy is rarely worth the cost. If the PSU fails, the miner sits offline for the day it takes to swap in a spare. Annual revenue loss from one PSU failure is small relative to the cost of running two PSUs continuously. Keep one spare PSU per cluster of five units instead of running redundant supplies on each one.

For commercial setups where every hour offline matters — hosted operations, large farms with SLAs — N+1 PSU configurations make sense. Multiple PSUs feed a common DC bus, and any single supply can fail without taking miners down. Rack-mount mining shelves often integrate this. The cost premium is meaningful — typically 30–50 percent over single-PSU configurations — and is justified only when downtime cost exceeds the redundancy cost.

Common PSU mistakes that take miners offline

The recurring failures fall into a small set of categories.

Undersizing on cold start: a PSU sized exactly to nameplate trips when the miner cold-starts at 110 percent of nominal for a few seconds. Headroom prevents this.

Wrong input voltage: a PSU configured for 220V fed 110V will refuse to start or will start at derated output. Verify the input voltage selector switch on PSUs that have one.

Mixed PSUs on a dual-PSU miner: two PSUs of different models or firmware versions can fail to share load correctly. Match exactly.

Daisy-chaining beyond rating: some operators run multiple miners off a single PSU using cable adapters. This breaks the manufacturer’s design assumption and overloads connectors. Each miner should have its own PSU, sized appropriately.

Cable gauge errors: the DC cables from PSU to hashboard are sized for the manufacturer’s load. Substituting smaller gauge cables creates resistive heating and voltage drop. Always use the manufacturer’s cable kit or matched aftermarket equivalents.

Sizing examples for common units

A Bitmain S21 at 3500 W rated wall-draw with 1.20 headroom needs a 4200 W PSU. The shipped APW12 PSU handles this on 240V input.

A Bitmain S21 Pro at 3500 W rated, same sizing — 4200 W PSU at 1.20 headroom.

A MicroBT M60S at roughly 3360 W rated draws 4032 W at 1.20 headroom, which the shipped P29 PSU covers on 220V.

A Canaan A1628 at roughly 3145 W rated draws 3775 W at 1.20 headroom, sized against the shipped AUC PSU.

A Goldshell AE Max II altcoin unit at lower wattage uses an internal PSU sized by the manufacturer.

For sourcing current-generation hardware where PSU and hashboard ship matched, the Coin Web Mining catalog covers the in-stock lineup.

When the PSU is the bottleneck

If a setup is power-limited by PSU rather than circuit capacity, upgrade options exist. Aftermarket PSUs from Vnish, Braiins firmware tooling, and third-party suppliers can replace stock PSUs with higher-output equivalents. The risk: voiding warranty, mismatched cable harnesses, and unsupported firmware integration. Stock PSUs are sized correctly for stock firmware; swap only when there is a specific reason and a clear understanding of the warranty and reliability consequences.

One last sanity check operators skip too often: the PSU’s nameplate efficiency rating (Platinum, Titanium) applies at a specific load fraction, typically 50–80 percent. A 3,200 W ASIC on a 4,000 W PSU runs close to the efficiency sweet spot; the same ASIC on a 3,600 W PSU runs hotter, draws marginally more from the wall, and ages the PSU faster. The right PSU is almost always one size up from the rated load — not because the ASIC needs more power, but because the PSU lasts longer and the wall draw is slightly lower over the unit’s life.

Two final operational notes worth pricing into a PSU decision. First, the cable run from PSU to ASIC drops voltage proportionally to its length and gauge — runs over two meters at the higher 12 V output current should use thicker wire than the stock connectors typically include. Second, dual-PSU configurations for high-power units like the S21 family must be wired against a balanced load or the weaker unit will fail first; check the manufacturer’s harness diagram rather than guessing.

References

• Bitmain APW12 and APW171 PSU specifications — Bitmain
• MicroBT P21G and P29 power supply documentation — MicroBT
• Canaan Avalon AUC PSU and miner specifications — Canaan
• ASIC Miner Value power and efficiency database — ASIC Miner Value

Operators sizing a multi-unit deployment can browse the Coin Web Mining catalog for matched PSU-and-miner combinations, or request a bulk quote for orders of five units or more.