Hacker News Reader

Comment by ajross | Hacker News Reader

> Even still, there is no way to know

That seems unconvincing. In fact how much energy is required per hash is something that's really quite easy to measure accurately. At any given time there are no more than a few dozen hardware platforms in use (they get priced out rapidly, everyone is running the newest stuff in the BTC world). Just plug them in and measure them.

All that's left from there is to come up with second order corrections about stuff like grid efficiency. But we absolutely know to quite high precision how many joules were spent to produce the coins we observe to have been created.

View on HN

> everyone is running the newest stuff in the BTC world

That's not true. The profitability of a miner is not measured with their hardware generation but with the energy cost needed to operate them. If you have access to very cheap energy, they can still be very profitable. There is a reason the market price of old mining rigs is not 0.

> There is a reason the market price of old mining rigs is not 0.

Electrical costs vary by a factor of two or three around the world. Maybe five at the extreme outliers. Modern mining hardware is tens of thousands of times more efficient than the stuff used less than a decade ago. Also due to growth in the market, the rate of production has been increasing rapidly so the total number of old rigs is negligible anyway.[1]

To rather accurate first approximation, all BTC mining is done with hardware less than two years old or so.

[1] Also also, the recent crash has had the effect of depressing mining revenue and accelerating the retirement of the old junk.

Others have mentioned it, but I'll say it too:

what you know is an estimate of total network hashrate. that's it.

but! total network hashrate moves around all the time, so it is really just an estimate. it has gotten better over time, but these machines don't operate perfectly 24/7 and older hardware is still in use. power outages happen too. stuff is just coming on and offline all the time.

you don't know efficiency (watts to hash) of the equipment on a global scale. you can extrapolate, but it is impossible to know for sure.

there are other losses, like stale and invalid shares. also impossible to know. you don't know how much is 'wasted' either... every single power supply on the planet is a snowflake. this would actually increase power usage estimates since it is effectively truly wasted energy.

disclaimer: I've PoW mined since 2014. Even I, with many many megawatts of power, have a hard time calculating exactly how much power we use and that's having access to the raw data!

If you have access to the raw data, then you should be able to quite easily provide a good estimate for what the article is trying to assess--what's the actual operational cost of bitcoin mining. Estimating the energy consumption is only a vehicle to doing this because those of us who don't have any raw data to do this.

What you should be able to do is:

a) Open up your electricity bill and look at the total energy consumption for a given period. Admittedly, I'm not familiar with how electricity is invoiced, but I'd be surprised if there wasn't something that was at least coarsely indicating total energy consumption for a billing period. (Or you can skip several middlemen steps and just take the total cost of the electric bill.)

b) Tot up all the bitcoin mining rewards you received in the same time period.

c) Count how many bitcoin blocks were mined in the same time period.

d) Divide b by c, now you know what fraction you are of the total bitcoin mining network.

e) Divide a by d, now you have an estimate of total worldwide energy consumption of bitcoin.

The only reason I can see that might make this hard is if you mine more than just bitcoin, in which case, replace "bitcoin" in the above steps with "sum over all of the cryptocurrencies you mine"--which is closer to what people want to know anyways. Even then, you can use crude weighting metrics to estimate what the consumption of just bitcoin is, which would at least let you know if the estimate is in the right ballpark or not. The exact number doesn't matter; it instead matters if the real number is closer to 14 TWh/yr, 140 TWh/yr or 1400 TWh/yr.

I will repeat it again:

you don't know efficiency (watts to hash) of the equipment on a global scale

In other words, one person might use 1watt to produce 1hash. Someone else might use 2watts -> 1hash.

That would be literally half as efficient and muck up any quantifiable numbers on a global scale.

Never mind all of the other things I mentioned above.

The reason why this is always such a highly debated topic is literally because it is something impossible to quantify.

But we really don't care about that factor of 2. It literally doesn't matter, whether Bitcoin wastes half a Luxemburg or a full Luxembourg.

And, iirc, then they estimated total consumption from well-funded and very active mining pools. I.e. exactly the operators that you would expect to run the most efficient mining equipment. Meaning all you do is being overly optimistic, as in "we claim mining uses less energy than it actually does"

Again, though, that's just not true. The hardware is well-understood and easily measured. Sure, there is slop in the measurement (how much of the fleet is using platform A vs. platform B). But those are comparatively small numbers. Sure, like you say, we might have a factor of two lurking in there (seems high, but I'll grant it).

But a factor of two wouldn't change any of the analysis! If the article said "60 TWh" instead, would that be any less horrifying? No, it wouldn't.

We don't know what hardware is on the network and we don't know how all of the power is generated. If it is 60TWh of clean energy that isn't being used for anything else, who cares?

The statistic you're poo-pooing puts error bars that are a factor of 2× in either direction. Tolerating that degree of error bar isn't all that difficult, and if you bothered to continue to read the article, you'd discover that the author themself discusses possible error bars, and deflates the estimated operational cost of bitcoin miners by half anyways.

And, FWIW, the kind of methodology that was used to produce the Bitcoin energy consumption is the same kind of methodology that is used to produce statistics like GDP, inflation, employment--basically every macroeconomic indicator. If you're going to complain that you can't quantify Bitcoin's energy consumption, then to be intellectually honest, you need to complain that the true inflation rate or the true size of any country's economy is impossible to quantify as well.

And the only way to hypothesize something different is if there's a ASIC or something that can mine much better than existing equipment (in which case the group that had it would suddenly shoot up in the blocks mined lists) or there's a weakness in the cryptographic protocol somewhere that we don't know about.

I must note that this in fact happened in the past (AsicBoost). Currently, I don't think there is AsicBoost-like secret information, but you can't be sure.

https://arxiv.org/abs/1604.00575

Let’s assume you have AsicBoost 2.0 - what would be the most strategic way to deploy it? I assume it’d look like a hashpower increase to the rest of the world, which may or may not go noticed; but perhaps have them run “normal” and then fling it wide open just before? after? a havening?

I’d guess the optimal strategy is “get that thing turned on as fast as possible and keep as much of the cluster running 24/7 as you possibly can before you get leapfrogged.”

Reply on news.ycombinator.com