2013-04-14 Bitcoin Miners Are Racking Up $150,000 A Day In Power Consumption...

[gweedo]

http://techcrunch.com/2013/04/13/the-cost-of-a-bitcoin

There’s a gold rush going on these days, or a Bitcoin rush, at least. Driven by the recent swings in the value of a Bitcoin, more and more people are learning about and becoming interested in the currency. While they could just buy Bitcoins at the current market rate, others are looking to try their luck at mining Bitcoins. And like prospectors who traveled west during the Gold Rush of the 19th century, many Bitcoin miners will find that they spend more on chasing the Bitcoin dream than they’ll ever hope to win back.

[1713510235]

1713510235

[flug]

Here's my calculation.

At $100/BTC, mining reward = 25 BTC per block * 6 blocks per hour * $100/BTC = $1500 per hour

Assumption: 50% of this will be spent on electricity, and the remaining 50% on paying off mining hardware, profit, etc.

=> $750/hour spent on electricity

= 5000 kWh/h, @ 15 cents per kWh

i.e. power @ $100/BTC = 5MW

Hence
power @ $1000/BTC = 50MW
power @ $10,000/BTC = 500MW (a nuclear reactor)

[Stephen Gornick]

Assumption: 50% of this will be spent on electricity, and the remaining 50% on paying off mining hardware, profit, etc.

Instead of pulling numbers out of thin air lets use upper and lower bounds.

GPUs are the least efficient.  ASICs are the most efficient.

If all mining was ATI 5970s (a relatively decent GPU), then there would need to be about 81,000 of them to make up today's 61 Thash/s.   Each consumes about 350W, (e.g., you can put three in a box with a 1200W power supply and end up drawing about 1,050W).  So that is 28.35kW, running a full day gives 680,400 kWh.  At $0.15 per kWh, thats about $102,060 per day in electricity.  

So that's the upper bound.  I would guess that in reality less than 1/3rd of hashing today is done on GPUs.   So that's $33K USD today from GPUs, plus a fraction of that for FPGAs and and even smaller fraction for a third of that 61T hashing coming from ASICs.

If instead all mining was done on ASICs, then let's use for the math the Avalon ASIC which does 60 Ghashs and draws 600W.  So if all mining was on those there would be about 1,016 of them needed.  That is 609 kW, running a full day gives 14,630 kWh.  At $0.15 per kWh that's $2,194 per day in electricity.  

And that's the low end.  

So an educated guess would say today that total cost to power all bitcoin mining today should be a number less than $50K, and will probably end up in the range of $15K to $25K per day one ASICs have shot difficulty up to about 500 Thash/s.   (And BFL's ASICs are more efficient on power consumption than Avalons so even at that level the amount for power could be less.)

[StarfishPrime]

Basic thermodynamic facts seem to get lost in these (often ridiculous) arguments about bitcoin mining power consumption:

The thermal output of a 1kw bitcoin mining setup is exactly equal to the thermal output of a 1kw electric heater

In other words: In a climate/season where mining heat output can contribute to heat your flat, house, etc. and your normal heating is electric, your net mining power cost should be basically zero.

In any event, the electricity consumed by the fraction of banking infrastructure soon to be displaced by bitcoin is probably orders of magnitude higher than mining operations ever will be.

[zby]

http://techcrunch.com/2013/04/13/the-cost-of-a-bitcoin

There’s a gold rush going on these days, or a Bitcoin rush, at least. Driven by the recent swings in the value of a Bitcoin, more and more people are learning about and becoming interested in the currency. While they could just buy Bitcoins at the current market rate, others are looking to try their luck at mining Bitcoins. And like prospectors who traveled west during the Gold Rush of the 19th century, many Bitcoin miners will find that they spend more on chasing the Bitcoin dream than they’ll ever hope to win back.

Thats why PPCoin was invented - with proof of stake you don't need this crazy power consumption for maintaining the network.

[flug]

GPUs are the least efficient.  ASICs are the most efficient.

There will be a tendency for the same electrical power to be used whatever the technology. ASICS just go faster for the same power.

[Impaler]

This numbers on ASIC power consumption support my theory that the bubble was partly due to a change in the amount of daily produced BTCs that miners had to sell to cover operating costs.  Back in February at value of $13 miners basically had to sell every coin made ends meet because the ~20 TH at that time cost 33K per day and the coins were worth only around 47K, so you couldn't mine to hoard coins, you could only mine to sell coins to cover costs.

Once ASICs arrive the electric costs are so small you can cover costs with just a fraction of mined coins.  And this caused a drying up of coins flowing into the exchanges.  Naturally the existing stock of coins is large compared to the daily inflow but their are also large outflows as well so with a lower inflow rate prices had to rise.  And that caused a self reinforcing cycle of both more people buying coins speculatively and more miners hoarding rather then selling their production.  By the height of the bubble I suspect virtually no newly mined coins were being sold and you could only move a coin by getting another hoarder to finally give up and sell.

The July 2011 bubble and crash probably had the same underlying cause but with the transition from GPU to CPU mining being the trigger.

[BubbleBoy]

That's an interesting theory Impaler, but I don't believe it can be more than a butterfly effect, the initial trigger of the speculative mania. When bubbles hit Bitcoin the trading volume increases 10x relative to the mining revenue. Even if miners dump all production and stock at that time, it's too late to stop the greed and mania, the demand is simply too large.

[Impaler]

As I said its the trigger for the mania, the bubble becomes self-sustaining very quickly.  Normally newly mined coins are producing a constant downward pressure on price, but leaps in the energy-efficiency of mining removes the downward pressure allowing price to build upward momentum, once the momentum is created it is clearly self sustaining and the miners themselves can become speculators by 'stealth buying' coins by just eating their own electric costs.  During most of the bubble this would have been the equivalent of buying coins at a fraction of their cost on the exchange, it would seem like a fabulous deal.

The rapid increase in trading volume is not an increase in the pool of coins being fought over, it is almost certainly just high trading velocity and it's perfectly possible for that pool to actually be shrinking as prices rise, in fact that's what we would expect.  I don't believe that miners ever dumped their coins en-mass, the bubble likely popped due to exhausting the supply of new adopters at which point the short term speculators were unable to find new buyers and drove the price down.

[kjlimo]

http://techcrunch.com/2013/04/13/the-cost-of-a-bitcoin

There’s a gold rush going on these days, or a Bitcoin rush, at least. Driven by the recent swings in the value of a Bitcoin, more and more people are learning about and becoming interested in the currency. While they could just buy Bitcoins at the current market rate, others are looking to try their luck at mining Bitcoins. And like prospectors who traveled west during the Gold Rush of the 19th century, many Bitcoin miners will find that they spend more on chasing the Bitcoin dream than they’ll ever hope to win back.

Thats why PPCoin was invented - with proof of stake you don't need this crazy power consumption for maintaining the network.

Has the proof of stake concept been implemented?  Or is it before time for that to start doing it's thing yet?

As far as power goes... my calculations last night got me less than $5 of electricity per BTC for month of March.

I'm guessing the GPU was in the range of less than $20, so not time to stop mining with my GPU yet.

[Stephen Gornick]

There will be a tendency for the same electrical power to be used whatever the technology. ASICS just go faster for the same power.

I doubt that.

It might cost $400 for a GPU that burns over $400 per year in electricity.  So given a 2 year useful life of the GPU, about a third of the mining costs are incurred for the hardware cost and two-thirds are attributed to the electricity.

With a $5K Avalon ASIC, that chews $780 per year in electricity.   So given the same 2 year useful life, about three-quarters of the mining costs are incurred for the hardware cost and just one quarter is attributed to the electricity.

So over time the tendency will be for mining to reach break-even, then at that point two to three times more electricity will be consumed when mining is on GPUs versus when on ASICs [Edit: at the same BTC/USD revenue level].

[flug]

There will be a tendency for the same electrical power to be used whatever the technology. ASICS just go faster for the same power.

I doubt that.

It might cost $400 for a GPU that burns over $400 per year in electricity.  So given a 2 year useful life of the GPU, about a third of the mining costs are incurred for the hardware cost and two-thirds are attributed to the electricity.

With a $5K Avalon ASIC, that chews $780 per year in electricity.   So given the same 2 year useful life, about three-quarters of the mining costs are incurred for the hardware cost and just one quarter is attributed to the electricity.

So over time the tendency will be for mining to reach break-even, then at that point two to three times more electricity will be consumed when mining is on GPUs versus when on ASICs [Edit: at the same BTC/USD revenue level].

Yes fair enough. Because the initial capital investment was larger, so for ASIC mining a greater proportion of the mining profits will be spent on capital repayments and less on electricity.

But still I think a fundamental equation holds that the market will balance when:

rate of mining income = rate of capital repayment + rate of profit + electrical power cost

Can we agree on that?

And as ASIC mining matures (a) capital repayment will decrease (as ASICS get cheaper, and the overall capital will get increasingly paid off) and (b) profits will tend towards zero (like with the mature GPU mining market)

For me, the only point of contention is how to quantify the capital repayment. But however it's quantified, the rate of electricity consumed will remain broadly proportional to USD/BTC. i.e. even if ASICS turn out to be 3 times more efficient over their lifetime, that won't compensate for the price of BTC increasing 1000 times.

[Stephen Gornick]

But still I think a fundamental equation holds that the market will balance when:

rate of mining income = rate of capital repayment + rate of profit + electrical power cost

Can we agree on that?

Over the long run, yes.

the rate of electricity consumed will remain broadly proportional to USD/BTC. i.e. even if ASICS turn out to be 3 times more efficient over their lifetime, that won't compensate for the price of BTC increasing 1000 times.

Absolutely, if there is a doubling in the exchange rate, there will be a proportional increase in the amount of electricity consumed, unless further efficiency from hardware design can (and likely will) come.   I had expected the cost of hardware to drop as higher quantities are produced but with demand far outstripping supply we're not there yet.  But regardless of the proportion, you are correct.  A higher exchange rate will cause a greater amount of electricity to be consumed.

There was handwaving about this in the media as early as 2010 when most mining was on CPUs, ... with projections that if bitcoin caught on it would consumer a quarter of all the world's electricity demand (or some whack number like that).  Even using the power draw if all mining were done on GPUs today, daily power consumption is a fraction of the daily energy consumed at even one gold mine (my assertion, without actually having calculated it but I've literally stood under the trucks that mine ore and can say with confidence that moving dirt around consumes one hell of a lot of energy.)

[aigeezer]

Here's another perspective.... If the entire electricity cost of all BTC miners is $150,000 per day, then the entire BTC mining effort costs a little less in a year than Goldman Sachs Chairman Lloyd Blankfein earned was paid in 2007.

Source: Wikipedia http://en.wikipedia.org/wiki/Lloyd_Blankfein

[BubbleBoy]

Allow me to quote myself from another thread on the same topic:


If the United States dispensed of dollars entirely and replaced all it's ~ 3 trillion $ M1 stock with bitcoins, the current bitcoin daily mint rate of 0.03% would cost today's equivalent of one billion dollars per day in wasted resources. At 5c/KWh that comes down to 20 billion KWh or 20 TWh / day. That's more than the entire electricity production of the United States which hoovers around 15TWh/day.

Even if we allow 75% of the mining cost to go towards capital expenses (asics, rigs, datacenters etc.) we are still talking about a full third of the national electricity production dedicated to bitcoin mining. And producing 750 million dolars worth of mining infrastructure per day is bound to have some ecological impact. You could literally gift a computer to every child in the developing world for a few months  of running the Bitcoin economy.

Bitcoin is the currency of Mordor.



Ok, let's drop the immediate obsolescence of the dollar as implausible, and say it would take Bitcoin 15 years to replace the US dollar. The block reward would halve 3 times to 3.125 BTC and the total mined by then would be around 20 million, bringing daily expansion to 0.0023%. Even in this scenario mining of bitcoins would suck a full 5 to 15% percent of the national electricity production, depending on the capital ratio.

I don't want to split hairs, but if the sheer order of magnitude of these numbers don't concern you, then you might suffer from the modern day version of the gold fever. There's just about no chance in hell a democratic society will allow such a behemoth to live (only to be subjected to the feudal whims of the early adopters that happened to be in the right forum 20 years ago, I might add).

The concept of cryptocurrency and it's privacy and liberty implications are revolutionary. The specific way bitcoins are produced in this chain is bad and can't possibly survive in the long run. Bitcoins will be replaced by better designed cryptocurrencies, there's no doubt about that.

[BubbleBoy]

Sure, but we need to compare apples with apples to understand the true ecologic impact; people keep comparing the puny total consumption of the puny bitcoin network to the whole financial sector.

[bb113]

Sure, but we need to compare apples with apples to understand the true ecologic impact; people keep comparing the puny total consumption of the puny bitcoin network to the whole financial sector.

I've been thinking a lot about that post. Bitcoins are like energy credits. The scenario in which all energy consumption goes towards bitcoin mining is absurd, it will not happen, instead the value of the mining reward will approach the amount of energy the society is willing to devote to processing transactions (say 0.1 % of available energy). Beyond this point, electricity costs will increase moreso than the value of the mining reward in terms of goods and services, since the price of these will also increase due to the increase in electricity cost.

Further, imagine a time when all 21 million coins have been mined and miners are rewarded by only transaction fees. In this future, everyone agrees to pay every single satoshi as a transaction fee to whoever mines the next block. The value of this block can not possibly be greater than the value of the total amount of energy available to that society. Therefore the "market cap" of all bitcoins in existence has an upper bound, that is the total amount of energy available at any given moment.

Maybe bitcoins should be valued in Watt-hours, not USD.

[flug]

Maybe bitcoins should be valued in Watt-hours, not USD.

Bitcoins will just be valued in whatever the market wants them to be valued.

Generally, I think there are various ways the energy problem can play out in the future, but first of all we do need to come to a consensus on whether it's a problem that should be taken seriously, and I don't see a lot of discussion of this problem around the forum. If the price of bitcoin does reach $1M at some point, we do need to calculate the implications of that.

It might be that a quantum computer comes along that's so expensive that most of the mining/transaction rewards are spent on capital repayments and not electricity. It might be that by the time it gets to $1M, the mining reward has dropped to the point where there isn't much to spend on electricity anyhow. It might be that Bitcoin is fine for a few years to come, but then when the problems arise another currency is ready to take over. We an't just ignore the problem tho.

[BubbleBoy]

the value of the mining reward will approach the amount of energy the society is willing to devote to processing transactions (say 0.1 % of available energy). Beyond this point, electricity costs will increase moreso than the value of the mining reward in terms of goods and services, since the price of these will also increase due to the increase in electricity cost.

Clearly. It's not my intention to suggest that bitcoin mining could ever suck all the electricity produced. That would send the energy price skyrocketing. But since bitcoin is money in itself, and not just another energy-derived commodity that the market can find substitutes for, there's no way for society to agree on such a low fraction as you suggest. The way bitcoins are produced demands resource destruction, and the resource destruction is proportional to the real-life value of a bitcoin. So the resources available for consumption will drop (significantly) and the their price would rise to reflect an equilibrium, making non owners of bitcoins poorer in real terms.

The main point here is that, given it's current size, the bitcoin economy is horribly inefficient and this a reductio ad absurdum proof.

Further, imagine a time when all 21 million coins have been mined and miners are rewarded by only transaction fees.

Sure, if mining covers only security then less resources are burnt and the comparison with gold becomes less valid. But as you've seen, even 15 years from today the waste problem is still major, so we would have to wait decades to finally gain a decent system. But why not create an alternate system today where say each citizen receives one coin, or some group of charities own them all and spend them into existence ?

It's also not clear if mining is really necessary for achieving security. There are various proof of stake proposals floated in the forums and the Ripple network is using a hash-free darknet consensus algorithm. We'll see where these schemes go. With mining not necessary for either coin disbursement or network security, I really can't see how Bitcoin will survive on the long run.

[d'aniel]

Sorry, not sure what thread this is being discussed in now, but here is my response to BubbleBoy's (cross) post above:

@BubbleBoy,

With the US consuming around 20% of the world's electricity production, then in the wildly successful hypothetical scenario you described, bitcoin mining consumes 1-3% of the world's electricity (using your numbers).  With the natural incentive to make optimal use of the waste heat, a likely large majority of this energy would've been consumed for heat anyway, so bitcoin mining would really only be contributing an additional few tenths of a percent to world electricity consumption, and halving every four years.  Depending on how mining is distributed amongst nations, these numbers are likely irrelevant to most in terms of their national energy policies.  And of course this does not factor in the cost savings from the monetary systems Bitcoin has supplanted.