Does an increase in mining difficulty lead to higher power consumption?

[Felicity_Tide]

I have been curious about why Bitcoin mining consumes a lot of power(energy). Although I haven't come across a mining device physically, but there are some articles and videos I have seen that talks about this consumption.

I have been curious to know, and there are certain things that have come to my mind in the last few days.

From what I have learnt, the difficulty of mining is said to readjust after every 2016 blocks, which is around 2 weeks, while the average time to mine a block is 10minutes. But I am not sure if the computational power of mining devices like ASIC for example, are built to increase relatively with an increase in difficulty level of mining. So I came up with a conclusion that: in order for these machines(mining devices) to match up with difficulty increase, it might result in more energy consumption.

Though, I also think, putting ON mining machines for long can also consume power, but there should be some additional task that tend to make the machine work more, thereby consuming more energy.

I am not certain on the accuracy of my conclusions, which is why i would like to get a more detailed answer to:

• Does an increase in mining difficulty lead to higher power consumption?.
• And also, is it possible to mine more or less than 2016 blocks for a certain difficulty level?.

I am 100% open to correction as I still see myself as a learner. Pardon any of my error and share your personal opinion. You might want to also DYOR after reading this.

[ranochigo]

Does an increase in mining difficulty lead to higher power consumption?

This depends, but you might've understood it wrongly. Difficulty increases because the network solves the block faster, which is generally due to an increase to the hashrate. This increase on hashrate can be several factors: less efficient machines being replaced, more machines coming online but it is usually the latter. Having more ASICs directly means a higher energy consumption.

Rather than saying that mining difficulty leading to higher power consumption, the more ASICs machines would have a higher power consumption and hashrates, for which the higher hashrates means that the network solves more blocks in a given 2016 window and thereby experiencing an increase in difficulty.

ASIC chips will likely stay at the same chip frequency throughout their entire life. They're usually not specifically over or underclocked periodically, but done once at the start to ensure that it is stable.

And also, is it possible to mine more or less than 2016 blocks for a certain difficulty level?

No, difficulty readjustment is triggered when the network has mined 2016 blocks since the last. It is not a function of time. It is meant to maintain the block intervals at 10 minutes.

[HeRetiK]

Power consumption and network hashrate (and increased difficulty as a result of this increase in hashrate and power consumption, rather than the other way round) are only related in the short term -- i.e. in cases where an increase of hashrate is caused by additional machines going online, rather than less efficient ones being replaced as explained by ranochigo.

Mid- to long-term power consumption depends on the budget miners have available, that is: How many bitcoins miners can expect to receive (block subsidy, transaction fees) multiplied by the exchange rate for which they can expect to sell (since electricity is still billed in fiat terms). So power consumption is more closely related to Bitcoin's exchange rate, rather than its hashrate and difficulty.

[odolvlobo]

So I came up with a conclusion that: in order for these machines(mining devices) to match up with difficulty increase, it might result in more energy consumption.

A mining device uses a fixed amount of energy, so the amount of energy the device consumes does not vary with the difficulty.

It is interesting to me that, in theory, the power consumption of mining is directly dependent on the price of a bitcoin and the price of energy, and it is independent of the efficiency of the mining equipment. This is because, due to the economics of mining, the cost of mining a block approaches the revenue from mining the block, which is currently 3.125 times the value of a bitcoin.

That will change over time as the subsidy approaches 0 and the revenue from mining a block is increasingly determined by the value of the transaction fees. Eventually, the cost of mining will approach the economic value of transacting in Bitcoin, which is about as close as you can get to ideal.

[alani123]

Bitcoin mining is something you can example with the ceteris paribus principle when it comes to the macro scale.

If the total difficulty increases but we are to assume that nothing else changed then yes, it probably means that more mining machines were switched on and hence more power has to be consumed for them to stay on.

However the reality isn't as easy. It's likely that many miners replace their old machines with more efficient ones for the same or even more hash. So this could mean that consumption stays the same of even gets lower in theory. But in reality an increase in difficulty must also mean an increase in power consumption because the increase in efficiency isn't as fast and chip sizes are already very small. Which hinders further advances in efficiency.

It's just that an increase in difficulty doesn't mean also a 1:1 increase in power consumption. This means that while there's correlation, the covariance isn't 100%. Maybe though an increase in difficulty means a more than 1:1 increase in power consumption these days because also newer mining farms require ever increasing in size cooling and monitoring facilities.

[d5000]

We can always ask the question: how much incidence in difficulty increase has energy consumption, and how much has hardware efficiency progress? As already others wrote these two are the main "reasons" for a hashrate (and thus difficulty) increase.

We can set up a small calculation how much incidence has each of both factors, if we know two variables: the hardware price per hash and the difficulty. This is still very simplified but should give us a rough result for the order of magnitude we're talking about.

The price per hash would be a rough measure for "hardware efficiency" and is governed by the same rules than Moore's Law. Unfortunately most "hash price charts" instead of offering a calculation based on hardware costs use the formula "miner income / hashrate", and while this is a valid approach for a rough approximation, it does also take into account energy consumption so it doesn't help to answer this question.

Instead, like in a discussion we had in another thread, we can use the available Antminer hardware generations as a base for our calculations. According to these calculations, the hashrate per USD has increased 8 times since 2018 (and thus the price per hash decreased to 0.125 times), i.e. in the last 6 years. If we break down this to the value per year, then we get a 40% yearly increase.

Thus, if difficulty increased 40% per year, probably all of the increase is driven by hardware efficiency progress. If it increased less, then the energy consumption is going back, and if it increases more than 40%, the energy consumption is going up.

How much increased difficulty in the last years? Let's see (source: Coinwarz):

- 2018: 1.93 to 5.58T (189%)
- 2019: 5.58 to 13.48T (148%)
- 2020: 13.48 to 18.59T (38%)
- 2021: 18.59 to 24.30T (31%)
- 2022: 24.30 to 34.24T (41%)
- 2023: 34.24 to 72.39T (111%)

We see thus clearly that in some years it's extremely likely the energy consumption went up because difficulty went up for much more than 40% (2018, 2019, 2023), while in other years probably miners retired equipment (2021, the reason was the China mining ban) and thus energy consumption went down. In 2020 and 2022 the increase was close to the 40% I estimated for hardware efficiency alone.

[mikeywith]

Does an increase in mining difficulty lead to higher power consumption?

do you mean in theory or practice? in practice, it's a simple yes, there are no ifs and maybes, we have 14 years' worth of data that proves power consumption and difficulty go hand in hand, it will probably be the case for many years to come, in fact, it would only get worse since efficiency improvement is only going to go down.

Thus, if difficulty increased 40% per year, probably all of the increase is driven by hardware efficiency progress. If it increased less, then the energy consumption is going back, and if it increases more than 40%, the energy consumption is going up.

It's more like 20%, refer to the code below. Also, price per th is irrelevant here, it has nothing to do with the power consumption unless I am missing a point.

How much-increased difficulty in the last years? Let's see (source: Coinwarz):
- 2018: 1.93 to 5.58T (189%)
- 2019: 5.58 to 13.48T (148%)
- 2020: 13.48 to 18.59T (38%)
- 2021: 18.59 to 24.30T (31%)
- 2022: 24.30 to 34.24T (41%)
- 2023: 34.24 to 72.39T (111%)

Glad you made the effort to compile it, we can use it with the average yearly gear efficiency improvement which is 21%

Code:

S9  90w/th using 16nm chips
2017 - S11 76w/th  15.5% improvement 
2018 - S15 57w/th  25% improvement
2019 - S17 45w/th  21% improvement
2020 - S19 36w/th  20% improvement
2020 - S19 Pro 29.5w/th  18% improvement
2022 - S19 XP 21.5w/th   27% improvement using 5nm chips
2024 - S21 17.5w/th 18% improvement 

The above isn't exactly yearly, but it's almost that, so it's pretty solid.

So the best year we had a 27% efficiency improvement, while the lowest difficulty increase we had was 31%, so certainly, power consumption increased.

Obviously, the missing factor here is how much of the new gear is more efficient vs not, but it's safe to assume that a lot of it is not the most efficient.

so in 2018, if we did have a 189% gear efficiency improvement it doesn't mean the power increased by 0%, because that would only be possible if the entire new hashrate came from S15 and not people adding old gears.


Anyway, now that the majority of mining is done in the U.S under publically listed companies, you can dig the sec fillings and other available sources to see, Marathon Digital, Bitfarms, Riot, CleanSpark, Iris Energy, Core Scientific and etc, all of them have been increasing their power consumption, despite increasing their overall gear efficiency.

[tech30338]

this means that the reward decrease as the difficulty increase longer time for a miner to mine, the power consumption for a month will be almost the same, but for example the reward is $100 from the previous month today you might get $80, and for example you electric bill is $100 from the last month for sure this moth is still the same.

[Felicity_Tide]

Does an increase in mining difficulty lead to higher power consumption?

do you mean in theory or practice?

 in practice, it's a simple yes, there are no ifs and maybes, we have 14 years' worth of data that proves power consumption and difficulty go hand in hand, it will probably be the case for many years to come, in fact, it would only get worse since efficiency improvement is only going to go down.

In practice, as in real life situations.
I got the explanation though.

[d5000]

Also, price per th is irrelevant here, it has nothing to do with the power consumption unless I am missing a point.

It would be a serviceable approximation if we assumed that mining hardware had always the same value for the formula "energy consumed / price", i.e. a similarly "dimensioned" miner would consume the same amount of energy.

But of course your approach is much better with the watts per hash being available. Judging from your numbers it seems that thus mining hardware with a comparable price tend to consume more energy over time, as the price per hash reduction is almost doubling the energy consumption per hash reduction. I've recently had a discussion in another thread which seems to confirm that (power supplies tend to be increasingly more powerful). So thanks for recompiling these values