World Bitcoin mining average ASIC efficiency
Bitcoin ASIC mining average is ≈ 24.7 J/Ths
|
This spread sheet calculates -
|
Mining equipment performance improvements-
|
The green column on the left is the out of the box advertised efficiency measured in Joules per TeraHash JTH, most efficient at the top. Miners can then alter the software and or use emersion mining methods to improve efficiency or increase Hash rate. The right hand green column shows observed improvements. Many miners will just run them standard, others miners with newer latest tech may over clock them (red) in an attempt to get a return on investment ASAP. The middle green column represent an average of standard vs Tweaked.
Key factors
- ASIC Efficiency
- Power consumption
- ASIC Power Mode (High Medium Low)
- After market software such as Braiins can increase efficiency by up to 30%
- Immersion mining can increase efficiency by up to 25% or allow to be over clocked
- Quantity produced - Mostly educated guess but quantities highlighted in yellow have solid evidence (links below)
- Functional percentage - As mines get older they wear out.
- UP time - 10,000 brand new mines in an industrial mining unit would have much more up time say 98% than an amateur using a 5 year old ASIC for heating say 30%
Quantities & up time etc are then tweaked to match the World hash rate. The world Average J/TH can then be calculated & thus Bitcoins overall power consumption.
Calculate J/THash - Method B
Some mining companies straight up tell you
Marathon-25.2 JTH September 2023
TeraWulf 25.7 J/TH July 2023
, others announce their hash rate & power consumption. A quick calculation gives us their average Joules per THash, this is likely to be inaccurate because they may like to brag "we have 100MW of infrastructure" but may only be using 60MW, or "we have 5 ExaHash" but only have 3 switched on. This gives clues to the big mining companies efficiency. Some companies execute Demand Response obviously they curtail their most inefficient machines first and maybe dialling down the newer machines. Both these would result in greater overall efficiency.
Marathon-25.2 JTH September 2023
TeraWulf 25.7 J/TH July 2023
, others announce their hash rate & power consumption. A quick calculation gives us their average Joules per THash, this is likely to be inaccurate because they may like to brag "we have 100MW of infrastructure" but may only be using 60MW, or "we have 5 ExaHash" but only have 3 switched on. This gives clues to the big mining companies efficiency. Some companies execute Demand Response obviously they curtail their most inefficient machines first and maybe dialling down the newer machines. Both these would result in greater overall efficiency.
Disproving other wild energy claims
United Nations University Study
United Nations University claim -
"In the 2020-2021 period, the global BTC mining network devoured 173 terawatt-hours (TWh) of electricity, “
"In the 2020-2021 period, the global BTC mining network devoured 173 terawatt-hours (TWh) of electricity, “
|
Claimed Power consumption = 173 TWh/year
Hours a year = 8766 Watts = 19.7 GW World Hash (2020-2021)= 125,000,000TH/s |
173,000GWh / 8766 = 19.7 GW
19,700,000,000W / 125,000,000TH = 158 W/TH = ❌ Incorect |
As we can see from the best on market graph, during 2020-2021 the best on market to would be 32 W/TH and the only just economically viable would be around 96 W/TH, so the average would be somewhere in between.
The deceptive report constantly compares compares apples to bananas, and in a staggeringly unscientific way neglects to make it clear that "2020 - 2021 period" actually means a 2 full years energy and emissions. A bit like saying my car does 123 miles per 2 gallons 🙄
The deceptive report constantly compares compares apples to bananas, and in a staggeringly unscientific way neglects to make it clear that "2020 - 2021 period" actually means a 2 full years energy and emissions. A bit like saying my car does 123 miles per 2 gallons 🙄
Cambridge
Working backwards, higher bitcoin mining power estimates seem excessive. Remember this is an AVERAGE J/TH so it must include vast numbers of antiquated junk to drag the average down.
The Cambridge estimates are a bit of a shot gun approach ranging from 57 - 120 - 196 TWh per year which straight away tells me they don’t really know & rather confusingly talks about Bitmain Antminer S7 (2015 out of date mining hardware). However we shall do the maths for the middle estimate.
The Cambridge estimates are a bit of a shot gun approach ranging from 57 - 120 - 196 TWh per year which straight away tells me they don’t really know & rather confusingly talks about Bitmain Antminer S7 (2015 out of date mining hardware). However we shall do the maths for the middle estimate.
|
Claimed Power consumption = 120 TWh/year
Hours a year = 8766 Watts = 13.7 GW World Hash = 300,000,000TH/s |
120,000GWh / 8766 = 13.7 GW
13,700,000,000W / 300,000,000TH = 45 J/TH = ❌ Incorect |
If we refer back to the spreadsheet’s efficiency column (green shades) we can see this level of ASIC operating efficiency is at best so so, with the vast majority of hash rate way above this. In fact 45 J/TH is now so far down the green column it's now in the brown 💩area
(these calculations are accurate for the claimed values at the time of Cambridge's publication - February 2023 )
(these calculations are accurate for the claimed values at the time of Cambridge's publication - February 2023 )
Digiconomist
In 2020 the widely quoted & practically the only thing you are allowed to find on the internet Digiconomist claimed
205 TWh/year (or 205,000 GWh)
205 TWh/year (or 205,000 GWh)
|
Claimed Power consumption = 205 TWh/year
Hours a year = 8766 Watts = 23.4 GW World hash (then) = 120 ExaHash |
205,000 GWh / 8766 = 23.4 GW
23,400,000,000W / 120,000,000TH = 194J/Th = 🤡 Utterly ridiculous ! |
|
In 2020 the typical ASIC efficiency coming on to the market was around 29 -38 J/TH (green column) ,such as the Bitmain T19 84TH - 37.5 j/TH
Also currently claims 51,000 tonnes of e-waste every year ?!! More like 250 tonnes/year. See my calculation for Bitcoin e-Waste. Spurious claims from an account run by an employee of a Dutch Central Bank (no conflict of interest there then 🤔). |
|
|
Let's assume for a moment that the ASICs never become unprofitable, never wear out & are never turned off. If we add power consumption for every miner ever all on full we get 174TWh with an average efficiency of 31.8 J/TH (top of spreadsheet area in pink) but in reality equipment does wear out, become unprofitable, and does get turned off, so anything with a higher fugues is proven to be inaccurate.
|
Then there's this doozy from the World Economic Forum .They didn't even bother to take it down 🤣
|
Conclusion
- No serious miner is going to mine at a loss with antiquated junk when there is equipment that is 3-4 times more efficient on the market
- Newer more efficient ASICs 16 - 24 J/TH will dominate the bulk of the hash rate.
- During price rises older mining gear may be turned back on but there was only so much made / functional.
- With the price really low ($21k) at the moment probably anything less than 38 J/TH (see green/brown column) will be turned off.
- Hash rate has increased mostly due to more efficient machines - NOT necessarily more power consumption.
Sources
|
Finding how many of which model have been sold / are still in operation is tricky. In an ever changing picture occasionally we get glimpses of sales in the 10-100s of thousands of units (Yellow highlight).
Immersion cooling allows over clocking of the mining rigs adding even more confusion, enabling the rigs to run 20-30% faster than the manufacturer specifications. Under clocking can allow ASICs to run more efficiently than the advertised manufactures rate, improving efficiency by 25%. |
|
|
Links to websites
|
