Bitcoin mining heater: Will it be worth it?

This is the second post in a series on setting up using a bitcoin miner to heat my Nana's house

As mentioned before, I’ve embarked on a project to heat my Nana’s house with bitcoin mining this winter. I’ve talked about why I’ve done this in the last post. What I want to try and cover in this post is my attempts at trying to work out if this will make me money or if I should instead have just paid for the heating of my Nana.

The calculations I want to cover are:

  • How much will it mine in bitcoin?

  • What will the value of that bitcoin be?

  • How much energy will it use?

  • How consistently can I keep the device on?

There are also some important questions around the ability to make the system acceptable as a system that will be living in my Nana’s house.

How much will it mine?

The first question is perhaps the most complicated to answer. It seems like something that should be simple. The device can run 100TH/s. That will always be more or less the case. The problem is that the difficulty of mining bitcoin doesn’t stay consistent over time. In general, as more people take up mining the difficulty increases. My understanding is that the difficulty increases such that the yield is proportional to the total hash rate you control.

The problem is that predicting the total hash rate is like trying to predict the stock market, there are some general trends that you can be relatively sure of but the short term fluctuations are completely unpredictable as it relies on individuals turning on or off the machine.

The recent banning of mining from china has caused the difficulty to drop below it’s trend line.

The other thing other than just how many people are mining that need to be thought about is if one of the ASIC manufacturers will release a new chipset that will be massively more efficient than the current hardware. It is the progression from one style of mining to another which really affects how the profitability of things changes with time.

For example, the original mining code ran on the CPU as all programs start by doing. Someone then felt that it was worth their time trying to optimise this code to run on graphics cards that can do lots of calculations at once. This made it harder for CPU mining to turn a profit even in the absence of electricity costs. After that came the project to convert the code to an FPGA. These are hardware that is reprogrammable. The hardware could be programmed to specifically mine bitcoin, or more specifically calculate double shar256 hashs.

The most optimal way though was always to make a custom chip architecture that could then be produced using the most up to date gate technology. This allows the cost per hash to tumble and the overheads to get as low as we know how to make them for a specific computation. This is where companies like antminer come in. They design and then have made custom chips that are designed to mine as much bitcoin as possible for as little energy as possible.

The endeavour of custom made chips is not a cheap one, however, if you can scale it up it can be a very profitable one.

The question after all this with relevance to the project of heating my Nana’s house with a bitcoin miner is; Are we close to the limit of mining efficiency per watt.

Based on the following graph which models the energy efficiency of antminers hardware over recent years it seems that we are asymptotically approaching the limit for current ASIC technology. Other technological advances are possible. One which will have far-reaching consequences is that of photonic computing. This could allow for all sorts of parallel computing to be done far more rapidly than is currently possible.

It should be possible to see if these chips will become a thing quite a way out. As photonics are quite a general-purpose technology, it would make sense that they will appear as proof of concept demos from chip designers who are pushing the forefront of this. One such company is Lightmatter. They are building out photonic chips that optimise matrix multiplications, or more specifically one calculation that is important to matric multiplications, that of the multiply-accumulate. This multiply-accumulate operation is something that there exist theoretical designs for and which Lightmatter are trying hard to make a reality.

My reasoning is that as companies like antminer are more in the industry of chip design and hardware construction than the manufacture of chips themselves I don’t see them pushing chip design in general forward and so until the technology from X is proven and available from the chip fabricators it is unlikely they will be able to make much progress.

That means the only place that can improve is using more advanced semiconductor technology, they are currently using 7nm processes so they really don’t have far to go with regards to miniaturisation.

Without a step-change

Without another step-change, in the energy efficiency of the devices, the main thing causing the difficulty to increase is simply more of the machines coming online. That will lead to the complexity trending towards what is profitable with the lowest energy cost. This is far more limiting than has existed in the past and so I don’t expect the difficulty to rise rapidly as it has as a result of the step changes in the past.

That said I have produced a very simple regression model based on the past few years and used that to project the profitability of the project over time.

What I haven’t taken into account is how the difficulty changes with the price of bitcoin.

If the difficulty will approach what is profitable with the lowest cost of electricity it will follow the price of bitcoin elastically. Since running the system I haven’t seen much upward elasticity in the mining pool. This is likely because the most efficient devices aren’t that old yet and even with the lower price swings of bitcoin are still quite profitable and even high energy prices.

I expect that over time the difficulty will move more elastically with price as different groups find their electric costs to make the device unprofitable at a given coin price.

What the value of bitcoin will be

This is an interesting question and I have no insight beyond the millions of analysts and crypto-bros that will tell you what the bitcoin price “should” be or “could” be.

For this project, the price of bitcoin is, however, an important factor for the profitability of the enterprise. One question that I’d like to try and answer, which can be answered independent of future knowledge of the bitcoin price is how investing in this system will perform over time relative to simply putting the money into bitcoin itself.

The difference here is that after whatever period it takes to pay back the device, including the heating savings if at the end of that period bitcoin is at the same price as it was at the beginning of the experiment or lower the experiment will have been a better investment.

On the other hand, if the price of bitcoin has increased at such a rate that its gains are, later on, this will benefit the mining approach as the earning of the miner are amplified by a lower difficulty from the low price of bitcoin in the period where the mining is occurring and a higher price after the fact.

Energy prices

As you may know, global gas prices have gone up substantially this winter causing European and particularly British electricity prices to rise. Thankfully my Nana is on a fixed economy 7 tariffs. 

The prospect of an economy 7 tariff is interesting because the price of electricity in those 7 hours is incredibly low.

That did give me the idea of trying to capture the full 54kwh the miner needs for the other hours on the day solely in this 7 hour, storing them in a battery overnight.

The other option might have been to try and use some solar energy to offset the more expensive daytime costs.

This also wound’t work as the house is both thatched and listed. The other option was wind which I did look into but there there are more issues with planning constraints.

How constantly can I keep the device on?

This is an interesting question and one that this project will help me to answer for projects going forward. I have generally been checking on the device every morning (remotely) to make sure that there haven’t been any changes in temperature (decoration put in the exhaust vent) or that the device has turned itself off.

I have had two downtime issues so far (Writing back in early October). One was from the heat and sound of the device getting too much for my Nana. From a sound perspective, soundproofing has since been added. From a heat perspective, I have reassured my Nana that the heat from this device is a waste product and so it is better to open windows than it is to turn it off.

The other downtime event came from an IP address change. This must-have confused either the device or the pool and resulted in the device not doing any hashwork for 19 hours. Although from an energy cost perspective there is no loss, as the device was idling it is lost time and a lower monthly hashrate than a 100% uptime would allow for. That cuts into the ability to pay down the capital cost of the device. The other issue was that the house that my Nana had got used to being up to 27 celsius was dropping and had reached 18.

The solution turned out to be the classic turn it off and back on again. I also needed to request what the updated IP address was so I could continue to log in to track the device.


Popular posts from this blog

An exploration in number systems

Structural engineering with cardboard

The twelve fold way