Hacker News Reader

Comment by dllthomas | Hacker News Reader

The point about efficiency increasing seems wrong. Yes, efficiency of hashing has increased, but it's an arms race. When Visa buys more efficient computers, we use less electricity to accomplish a goal. When I buy a better ASIC I still pour as much juice into it as I can afford to churn out hashes - and if the network is producing too many hashes they up the difficulty. As efficiency increases, with bitcoin we keep pouring the same amount of electricity into mining, but we're also pouring resources into building single-purpose chips that will be useless when bitcoin migrates to a different hashing algorithm.

View on HN

The total cost of mining a block will approximate the block reward plus transaction fees. To see how much of that cost is electricity, compare the cost of the hardware to the electricity it uses.

For example, on Amazon right now I see a Bitmain AntMiner U1 USB miner. It costs $70 and they claim it uses 2 watts and is a recent chip.

Let's say it has a lifetime of three years. Three years of two watts is 26 kWh, which will cost about $2.60, which means that for this chip, the electricity is only 4% of the total mining cost.

Of course the rest of the computer uses power, I'm assuming that either the computer is being used for something else as well, or a large operation has a special chassis holding lots of these chips and not wasting power on much else. Anyone who's not so efficient won't be able to compete.

The fact that bitcoin's mining cost is becoming dominated by hardware rather than energy is a very positive development, in my opinion.

Right now the ASIC vendors seem to be capital-constrained and are not producing enough ASICs to meet demand. At equilibrium we should expect that the cost of mining will equal the rate of BTC generation and a significant fraction of the cost of mining will be power.

Probably. On the other hand if chips get as optimized as they can get aside from further advance of Moore's Law, and they're produced in mass quantities, maybe people will start using them as heating elements in home water heaters.

A lot of the emerging currencies are focusing on memory-constrained hashing algorithms. RAM has price/energy ratios similar to what I mentioned for ASICs, and is already in mass production and about as optimized as major industry is able to make it.

Just a ballpark, how many bitcoins would be earned by such a machine per unit time?

The network hash rate is currently about 13,000,000 Gigahash / second.

That unit does 1.6 Gigahash / second.

The bitcoin earned in a period of time will roughly follow that proportion.

A piece that's missing is the total bitcoins earned, which depends on payout size (possibly including transaction fees) and difficulty.

Right. I figured the part where it represents something like a 1 in 8 million stake in the mined bitcoin answered most of the question.

> but we're also pouring resources into building single-purpose chips that will be useless when bitcoin migrates to a different hashing algorithm.

I don't get this argument. Why do they have to be used for something else after "you're done with them"? You've made your profit with them (since if you aren't going to make a profit with them, you aren't going to buy them in the first place), so what else would you use them for?

The ASIC's serve a purpose: mining Bitcoin and maintaining the Bitcoin network. I don't get why there has to be "something else" they need to be useful for. Bitcoin is the point.

There is the opportunity cost and the social cost. And there is no canonical, singular "point" about mining that one could refer to that would obviate those questions. But maybe the social cost is low, e.g., many miners and Bitcoin speculators would otherwise have been unemployed, and would have contributed at least as much carbon to the Earth's carbon budget had they done something else.

I see no reason to anticipate Bitcoin moving to a new hashing algo, given the investment in existing hardware, and that the hardware owners are the ones who vote with their gigahashes on how the network functions.

However, even if it were to change, there are several altcoins that use the same hashing algorithm (PeerCoin being the biggest), and/or someone will invent a new blockchain service to take advantage of the existing hardware. While I think there is great merit to ASIC-resistant hashing algorithms of ProtoShares/PrimeCoin/etc, that ASICs would ever become useless seems highly unlikely.

Attacks against cryptographic algorithms only get better. If SHA256 is weakened sufficiently, bitcoin (and altcoins that use SHA256) will need to move. Which it can, but that makes existing ASICs worthless.

Just make sure to only mine during a polar vortex. If only there were a bitcoin heatpump.

Its worth nothing that as this arms race proceeds, the historical record of Bitcoin transactions becomes harder to undo, and 51% attacks become too costly for even some nation-states.

Does anybody know how much energy, i.e., Joules or whatever per Bitcoin earned through mining?

I have been curious about this too.

Unfortunately though, my guess is that it is very very tough to estimate due to the sheer variety of hardware that is being used.

I am sure that a large number of people are still mining with GPUs which consume more power/mhash than ASICs.

I came across someone last week who was mining with a CPU, I am confident he's not the only one.

Not much more than 1 bitcoin's worth...

But possibly less, if there's some sort of arbitrage going on. Also, the energy requirement can't fluctuate as much as the price of bitcoins does.

Note I'm not trying to be argumentative. I'm actually genuinely curious. Even just order of magnitude. Folks in the media talk about the energy cost, but never with any attempt to quantify it.

For sure. It can especially be less due to short term capital constraints. If power gets cheaper, people want to produce more but it takes time to make more ASICs. To give a real thorough answer, though, takes more numbers and possibly more expertise than I have.

It looks like the best ASICs are at around 1.5 gigahash / joule. Certainly less than 2.

So 13 million gigahash per second translates into at least 6.5 megawatts. But 2 is a pretty aggressive figure, it's probably easily above 10 megawatts.

Reply on news.ycombinator.com