Definition
Efficiency (J/TH) measures how much electrical energy an ASIC miner consumes to perform a fixed amount of work, expressed in Joules per Terahash. Because one Watt equals one Joule per second, the figure is identical whether written as J/TH or W/TH, and lower numbers mean a more efficient machine.
Also known as: W/TH, power efficiency, energy efficiency.
Efficiency is the single most important specification for long-term mining economics. Two machines can post the same hashrate, but the one that produces it for fewer Joules pays less for every hash and stays profitable deeper into rising network difficulty. To calculate it, divide a miner’s wall power by its hashrate: a unit drawing 3,500 W at 200 TH/s runs at 17.5 J/TH (3500 ÷ 200). That ratio, multiplied by your electricity rate, is the dominant line item in any honest mining profitability model.
How efficiency works on real ASIC hardware
Efficiency is set by three things: the silicon process node of the hashing ASIC, the voltage and frequency the chips run at, and how much overhead the control board, fans, and PSU add. SHA-256 hashing power scales roughly with the square of voltage, so the relationship between speed and power is not linear — pushing frequency higher to chase more hashrate raises J/TH faster than it raises terahash. This is why the most efficient operating point is almost never the maximum-hashrate operating point.
The hardware history makes the trend concrete. The 2016-era Antminer S9 (BM1387 chip) runs near 84 J/TH at stock settings. By 2020, the S19 and S19 Pro (both BM1398, at 76 and 114 chips per board respectively) reached roughly 29.5 J/TH at nameplate. The 2023 Antminer S21 (BM1368) brought that down to about 17.5 J/TH, and the 2024 S21 Pro and S21 XP (BM1370 chip) push further still, into the mid-teens. Each generation is a step change driven by a newer chip on a smaller node — not by running old hardware harder.
- S9 (BM1387): ~84 J/TH stock — roughly 1,180 W for 14 TH/s.
- S19 / S19 Pro (BM1398): ~29.5 J/TH at nameplate, ~3,245 W.
- S21 (BM1368): ~17.5 J/TH, ~3,500 W for 200 TH/s.
- Bitaxe-class single-chip (BM1366 / BM1368): tens of Watts for a few TH/s — convenient for home learning, not built to compete with farm-grade J/TH.
Tuning, firmware, and the efficiency trade-off
You do not have to accept a miner’s stock efficiency. Underclocking and undervolting — lowering frequency and the per-domain core voltage together — almost always improves J/TH because you give up a little hashrate to give up proportionally more power. On a tunable S19j Pro, for example, dialing back from a stock-style 30.5 J/TH point toward a lower-power profile can reach the low-to-mid 20s J/TH, trading some terahash for meaningfully cheaper hashing. Custom firmware exposes this control through autotuners that calculate per-chip frequency targets at runtime to hold a power or efficiency goal, rather than relying on fixed factory presets.
A critical hardware caveat: efficiency tuning that lowers voltage requires runtime voltage control. Many modern S19/S21-class “NoPic” hashboards lack the PIC microcontroller that older boards used to adjust voltage, which limits how far some machines can be undervolted in the field. If you are evaluating a fleet, knowing which boards support runtime voltage adjustment is as important as the headline J/TH number. Our firmware overview at firmware comparison covers how different toolchains expose tuning, and the ASIC troubleshooting guides explain how to read power and hashrate telemetry to spot a board that is drifting off its efficiency curve.
Efficiency also degrades in the real world: dust, failing fans, weak hashboards, and high ambient temperature all push J/TH upward as chips throttle or draw more power for the same work. In cold climates, that “waste” heat can be recaptured — a point that matters for Canadian operators who treat the heat as a product rather than a loss.
When you are choosing hardware, J/TH is the number to anchor on, because it compounds across every kilowatt-hour for the life of the machine. Compare current-generation efficiency before you buy, and weigh it against your power cost and cooling setup. You can browse efficiency specs across available models at our miner catalog, and our team can help match a unit’s J/TH profile to your electricity rate. For broader context, see the related entries on electricity cost, difficulty, and the ASIC hardware that turns those Joules into hashes.
In Simple Terms
Energy efficiency of a miner in Joules per Terahash. Lower is better as it means less electricity per unit of hashrate.
Efficiency (J/TH) measures how much electrical energy an ASIC miner consumes to perform a fixed amount of work, expressed in Joules per Terahash. Because one Watt equals one Joule per second, the figure is identical whether written as J/TH or W/TH, and lower numbers mean a more efficient machine.
Also known as: W/TH, power efficiency, energy efficiency.
Efficiency is the single most important specification for long-term mining economics. Two machines can post the same hashrate, but the one that produces it for fewer Joules pays less for every hash and stays profitable deeper into rising network difficulty. To calculate it, divide a miner's wall power by its hashrate: a unit drawing 3,500 W at 200 TH/s runs at 17.5 J/TH (3500 ÷ 200). That ratio, multiplied by your electricity rate, is the dominant line item in any honest mining profitability model.
How efficiency works on real ASIC hardware
Efficiency is set by three things: the silicon process node of the hashing ASIC, the voltage and frequency the chips run at, and how much overhead the control board, fans, and PSU add. SHA-256 hashing power scales roughly with the square of voltage, so the relationship between speed and power is not linear — pushing frequency higher to chase more hashrate raises J/TH faster than it raises terahash. This is why the most efficient operating point is almost never the maximum-hashrate operating point.
The hardware history makes the trend concrete. The 2016-era Antminer S9 (BM1387 chip) runs near 84 J/TH at stock settings. By 2020, the S19 and S19 Pro (both BM1398, at 76 and 114 chips per board respectively) reached roughly 29.5 J/TH at nameplate. The 2023 Antminer S21 (BM1368) brought that down to about 17.5 J/TH, and the 2024 S21 Pro and S21 XP (BM1370 chip) push further still, into the mid-teens. Each generation is a step change driven by a newer chip on a smaller node — not by running old hardware harder.
- S9 (BM1387): ~84 J/TH stock — roughly 1,180 W for 14 TH/s.
- S19 / S19 Pro (BM1398): ~29.5 J/TH at nameplate, ~3,245 W.
- S21 (BM1368): ~17.5 J/TH, ~3,500 W for 200 TH/s.
- Bitaxe-class single-chip (BM1366 / BM1368): tens of Watts for a few TH/s — convenient for home learning, not built to compete with farm-grade J/TH.
Tuning, firmware, and the efficiency trade-off
You do not have to accept a miner's stock efficiency. Underclocking and undervolting — lowering frequency and the per-domain core voltage together — almost always improves J/TH because you give up a little hashrate to give up proportionally more power. On a tunable S19j Pro, for example, dialing back from a stock-style 30.5 J/TH point toward a lower-power profile can reach the low-to-mid 20s J/TH, trading some terahash for meaningfully cheaper hashing. Custom firmware exposes this control through autotuners that calculate per-chip frequency targets at runtime to hold a power or efficiency goal, rather than relying on fixed factory presets.
A critical hardware caveat: efficiency tuning that lowers voltage requires runtime voltage control. Many modern S19/S21-class "NoPic" hashboards lack the PIC microcontroller that older boards used to adjust voltage, which limits how far some machines can be undervolted in the field. If you are evaluating a fleet, knowing which boards support runtime voltage adjustment is as important as the headline J/TH number. Our firmware overview at firmware comparison covers how different toolchains expose tuning, and the ASIC troubleshooting guides explain how to read power and hashrate telemetry to spot a board that is drifting off its efficiency curve.
Efficiency also degrades in the real world: dust, failing fans, weak hashboards, and high ambient temperature all push J/TH upward as chips throttle or draw more power for the same work. In cold climates, that "waste" heat can be recaptured — a point that matters for Canadian operators who treat the heat as a product rather than a loss.
When you are choosing hardware, J/TH is the number to anchor on, because it compounds across every kilowatt-hour for the life of the machine. Compare current-generation efficiency before you buy, and weigh it against your power cost and cooling setup. You can browse efficiency specs across available models at our miner catalog, and our team can help match a unit's J/TH profile to your electricity rate. For broader context, see the related entries on electricity cost, difficulty, and the ASIC hardware that turns those Joules into hashes.