J/TH, explained properly: the number that decides everything
Efficiency is the one specification that keeps mattering after the purchase. What joules per terahash actually measures, how to convert it into a monthly bill, and why the most efficient machine is not always the right buy.
Hashrate sells machines. Efficiency decides whether you keep running them. If you only learn to read one line on a spec sheet, make it the one that says J/TH.
What the unit means
Joules per terahash is energy divided by work. A machine rated at 13.5 J/TH consumes 13.5 joules of electrical energy for every terahash of computation it performs. Lower is better: less energy for the same output.
The figure is derived, not measured independently. Take the power draw in watts, divide by the hashrate in TH/s, and you have it:
3,645 W ÷ 270 TH/s = 13.5 J/TH
That relationship is worth internalising, because it lets you sanity-check any listing in about four seconds. If the arithmetic on a spec sheet does not close, one of the three numbers is wrong or optimistic. It also means you can compute the figure yourself for units sold with power and hashrate but no stated efficiency.
Turning it into money
Efficiency only becomes meaningful when you convert it into your own electricity bill. The chain is short.
A machine’s monthly consumption in kilowatt-hours is its power draw in kilowatts multiplied by 720 hours. A 3.5 kW unit running continuously uses about 2,520 kWh a month. At €0.06/kWh that is roughly €151. At €0.12/kWh it is €302. The hardware is identical; the operating cost has doubled.
Now compare two machines. Unit A does 270 TH/s at 13.5 J/TH. Unit B does 200 TH/s at 18.8 J/TH. Per terahash per month, at €0.08/kWh:
- Unit A: 13.5 J/TH → about €0.78 per TH/s per month
- Unit B: 18.8 J/TH → about €1.08 per TH/s per month
Unit B costs 39% more to run for every unit of work it does. Over a machine’s service life that gap is usually larger than the difference in purchase price — which is precisely why efficiency is the specification that keeps mattering long after the invoice is paid.
The trap: efficiency is not the only axis
Here is where a lot of first-time buyers go wrong. They read the above, conclude that the most efficient machine always wins, and buy the flagship.
It depends entirely on your power price.
Efficient hardware is expensive hardware. You are paying upfront to reduce a recurring cost. The higher your electricity price, the faster that trade pays back — and the lower your price, the longer you are carrying capital that is not doing much for you.
At around €0.03/kWh, a cheap, thirsty machine can out-earn an efficient one on invested capital, because the power it wastes barely costs anything and you bought three of them for the price of one. At €0.15/kWh, the same machine may never cover its own consumption, and the flagship is the only thing that makes sense.
There is no universal answer. There is only your tariff. Anyone who tells you otherwise is selling something specific.
What the specification does not include
Three costs sit outside the J/TH figure, and all three are real.
The power supply is already counted — usually. Reputable manufacturers quote wall power, including PSU losses. Some listings quote at the hashboard, which flatters the number by 5–8%. If a figure looks unusually good for its generation, check which side of the PSU it was measured on.
Cooling is not counted. Every watt a miner draws becomes heat that something has to move. In an air-cooled shed with good airflow the overhead is small. In a space that needs mechanical cooling it can add 10–30% to your real consumption. Hydro units shift this cost into pumps and dry coolers rather than eliminating it.
Derating is not counted. Machines are rated at a defined intake temperature. Run them hotter and firmware will throttle to protect the silicon. The label stays the same; your effective J/TH gets worse.
Reading the figure sceptically
A few practical habits.
Check the arithmetic, as above. Ask at what ambient temperature the figure was measured. Ask whether it is wall power. Treat any efficiency claim attached to third-party firmware as a claim about that firmware, not about the machine — and remember what it does to the warranty.
Finally, compare within a generation. An 11 J/TH unit and an 18 J/TH unit are not competitors; they are different arguments about capital. The 18 J/TH machine is not a bad product. It is a product for someone whose power is nearly free.
The short version
Efficiency is watts divided by hashrate, and it converts directly into a monthly bill once you know your tariff. Do that conversion before you compare prices — not after. The right machine is the one whose efficiency matches what your electricity actually costs, and that answer is different for a site in Norway and a site in the Gulf.
We publish power and hashrate for every unit in the catalogue precisely so the division is yours to do.