Topic: The efficiency of the Bitcoin concept (Read 707 times)

Maybe in the future an alt-coin that solves real world problems can rise but for now we have to be content with burning some fossils to support our cryptographic currency desires

I didn't see that. Those numbers seem quite irrational 650 watts per gigahash I can see that with gpu's but there are a lot of FPGA's and Asics out there. 15 cents per kilawatt may be slightly high, I know mine's about 7.8 cents but I'm on the low end. Hardware cost is not anywhere near 1000 per gigahash and is bandwidth really a factor?

I'm not certain, but I suspect that it has a lot to do with the recent drop in exchange rate.  The exchange rate has dropped over 20% in the last 10 days.  Some miners may not yet be aware that they are operating at a loss, others may be holding out hope for a quick rebound in exchange rate.  If the exchange rate stabilizes where it is now, we'll probably see some miners cutting back on their mining activities.

Also, the numbers you are looking at are making certain assumptions about mining costs:

650 Watts per gigahash
15 cent per kilowatt hour
hardware costs $1000 per gigahash every 2 years
bandwidth $1 per gigahash per year

If any of those estimates are incorrect, then the estimated mining profits will be incorrect as well.

The chart assumes the network requires 650W per GH/s.  While that may be roughly true for GPUs it vastly overstates the cost of ASICs and thus during this transition that chart is next to useless.   

Modeling the mining cost of GPU is pretty easy.
Modeling the mining cost of ASICs is pretty easy.
The key factor is what % of the network do ASICs make.

In equilibrium yes.  That is what I attempted to estimate in updated post above. 

However right now the network is obviously out of equilibrium for a couple of reasons
a) many miners scaled back on GPUs and FPGAs in anticipation of explosive difficulty growth due to ASICs
b) ASIC supply is constrained.

Simple version is ASICs are highly disruptive and until their rollout is "complete" the network won't be in equilibrium.  ROI% on ASICs deployed today are off the charts obviously if possible people would deploy significantly more hashing hardware however deployed hardware is constrained not be economics but lack of availability.

By "complete" I mean a point where there is such availability of ASICs that the only reason more hardware isn't deployed is due to economics (i.e. miners stop buying more ASICs because it doesn't make economical sense to do so). 

~3.5PH/s is a ballpark guesstimate when the network is in equilibrium given:
* current exchange rate (~$100 USD : 1 BTC)
* miners deploy new hardware up to a 100% annual ROI%
* high availability of ASICs at ~$25,000 per TH/s hardware costs and $9,000 per TH/s annual electrical costs.
* there is no development of significantly superior ASICs (say magnitude better in terms of MH/$ and MH/W).

3,500 TH/s @ $9,000 per TH/s in annual electrical cost would put annual energy requirements at ~$30M (very loose guesstimate).

How is it that they estimate such negative mining profits? Or am I just reading that wrong?

Very interesting conversation, power usage has been making me hold back on running more GPU's especially now the difficulty has raised somewhat 

I suppose if you look at the exchange rate and the total subsidy plus fees earned, you can get a pretty good estimate.  Theoretically the total mining cost should be slightly lower than the total mining revenue.  That should provide the information needed to calculate the missing variable, right?

What is difficult to factor right now is how much of the network is ASICs.  It is likely the most important factor because ASICs are so much more efficient (<10W per GH/s) compared to GPUs that the overall network efficiency varies significantly depending on how much ASICs contribute to total network hashing power.

For example if we assume the average GPU is 650W per GH/s and the average ASIC is 10W per GH/s.  Lets convert that to 650 KW per TH/s and 10KW per TH/s respectively.  The network is roughly 170TH/s.  I am going to assume 10 cent per KWh instead of 15 cents as marginal miners are pushed out first, which should make the average energy cost of the network lower than average global energy cost. $0.10 per kWh = $100,000 per GWh

Is the network made up of 100TH/s of GPUs and 70 TH/s of ASICs?
100 * 650KW + 70*10 = 65,700 KW = 575 GWh annually = ~$58 mill USD annual electrical cost

Is the network made up of 150TH/s of GPUs and 20 TH/s of ASICs?
150 * 650KW + 20*10 = 97,700 KW = 855 GWh annually = $86 mill USD annual electrical cost

Is the network made up of GPUs and 120 TH/s of ASICs?
50 * 650KW + 120*10 = 33,700 KW = 295 GWh annually = $30 mill USD annual electrical cost

One thing is certain at current difficulty if ASICs replaced all GPU the annual electrical cost would be ~$2M USD (10 kW per TH/s * 170TH/s * 24 * 365 *$0.10 per kWH) and at $25,000 per TH/s, the deployed capital would only be a mere $5M.   Both are much smaller than the value of total mining reward.  If ASICs completely replaced GPU and the network hash power remained 170TH/s the annual return on capital deployed is in excess of 2,400% [ ($120M - $2M ) / $5M ].  It doesn't take a rocket scientist to say that reward greatly exceeds risk and as such there will be demand to deploy more hashing power.  This means difficulty is going up ... a lot.  The only reason difficulty isn't 20x higher is the slow rate of ASIC deployments.  People want more hashing power they just can't deploy it fast enough.

Lets assume eventually this bottleneck breaks and miners continue to add hashing capacity until they (collectively) feel it is no longer worth the risk.   That will be expressed as a return on capital.  How much return?  Hard to say but it depends on what return miners are willing to accept.   Lets assume 100% annual ROI% and exchange rate remains the same.

Global Mining Revenue - Electrical Costs = Return on Capital

Given:
Annual Gross Mining Revenue = $120M
Annual hardware cost = $25,000 per TH/s
Annual electrical cost  = $8,760 per TH/s

Then:
$120,000,000 - $8,760 * x = $25,000 * x
$120,000,000 = $33,760 x
x = 3,554 TH/s  Yes that is 3.5 PH/s or difficulty ~20x higher than today (difficulty 400M).


On edit: fixed error ($87,600 vs $8,760 annual electrical cost per TH/s).

POW is preferable to POS for the purposes of introducing a currency since in the POS system you are reward for holding on to your coins, creating a disincentive to spend them. In the short run POW is more appealing, but in the long run integrating a POS with POW can make for a more stable currency. Additionally, rather than using POS for a "currency" it can more appropriately be applied to something like a stock certificate, since you would want to hold onto that coin if you think the company it represents is going to grow and do well in the future. The rewards system in POS is somewhat comparable to the distribution of dividends.

This page:
http://blockchain.info/stats

Has some estimates based on assumptions of power consumption of 650 Watts per gigahash and electricity price of 15 cent per kilowatt hour.

Feel free to adjust the numbers if you think that it too conservative or aggressive of an estimate.

The network won't become more efficient due to more efficient hardware.  While an ASIC may have 100x the efficiency (in terms of work per unit of energy) of a GPU, economic theory says that eventually difficulty will rise by a factor of 100x or more.  The effects of supply and demand.  While miners are 100x more efficient in terms of energy per unit of work they are doing 100x as much work and thus the network will use the same amount of energy.  What caps difficulty is the risk vs reward.  Collectively miners have a point where the risk of deploying new hardware isn't worth the potential reward.  Above that point you will see a net reduction in hashing power and difficulty declines, below that point you will see a net rise in hashing power and difficulty.

What will make the network more efficient (in terms of energy to secure it)
1) More transactions.  While total network energy cost remains the same, more transactions means the cost per transaction declines.
2) Higher energy costs.  Higher energy costs mean miners costs will go up and thus everything else being equal difficulty will decline.
3) Lower "speculation" premium in the exchange rate.  Miners costs are in USD (or other fiat) and thus "difficulty follows price".  Today the price is partially supported by speculation (the belief that the exchange rate will rise rapidly in the near future).  This inflates the global rewards to miners and thus hashing power and thus difficulty.  As Bitcoin becomes more mature the growth rate will slow and the speculative premium will decline.  This makes global miner revenue more in line with actual economic activity.

That's makes sense. I hadn't thought about it being self limiting in that way. Has anybody ever done any power calculations for the entire network? Like estimating the current global hash rate and then using an average of power consumption per hash? I think it might be interesting to know that and see the efficiency of the network overtime as mining technology improves.

While there is an technical limit on difficulty, a practical limit exists due to the desire for profitability.  Right now we are witnessing a huge increase in difficulty, because ASICs are so much more efficient then GPUs. We saw the same thing in the move form CPU to GPU.  However eventually difficulty will be capped by profitability.  Would you mine for a guaranteed loss (not the risk of losing capital but an absolute guaranteed loss)?  I would imagine not and most rational people wouldn't either.  How high difficulty goes, depends on how low of a ROI% (margin) miners are willing to accept.  Logically miners are (collectively) unwilling to accept less than 0% annual return so that becomes the upper cap.  

It is difficult to estimate an exact "max" but you can get a ballpark by taking current exchange rate and BTC generation rate to find the global mining revenue.  Now estimate the lifespan of the average ASIC (will it be 1 year before it is economically obsolete? 3 years? 10 years?  Calculate the hardware cost plus lifetime energy cost in USD, convert that to BTC.  Now factor in what return on capital (ROI%) is the minimum you estimate miners will accept.  At this point you can estimate how high difficulty will rise before miners will balk at the idea of deploying more hardware. 


Bitcoin seems very energy intensive because it is for most people the first time they ever participated in any global network.  Bitcoin today is a rounding error on the annual global energy consumption.  Hell youtube probably consumes more energy so people can watch LOL cats and be tricked into thinking clips have boobs.

And yet fails to gain any traction in the marketplace.  Why do you suppose that is?  Is there something about proof-of-work that makes it preferable?

proof of stake uses a more energy efficient model for transaction confirmation. look into ppc

Perfectly agree. Also, when judging something, one always should consider the alternatives. The fiat currency system is no better when it comes to power consumption. Even though I don't have any reliable data I'm quite sure that to keep up the fiat payment system the power consumption is higher than for BTC (even if you take comparable transaction volumes). No modern currency (i.e. one that can be transferred electronically) will have zero power consumption.

I have also read some posts that in the future more people might use online wallets as they become more secure and trustworthy and some BTC user might not even run their locally installed wallet. That measn that there will be a few super-miners and a number of reliable and trustworthy wallet providers which in total should be quite energy efficient compared to many miners with rather inferior hardware.

The process uses the free market to be self limiting.

If the cost of the amount of power required exceeds the revenue generated from the block subsidy and transaction fees, then any reasonably intelligent and fiscally responsible miner will shut down their mining rig.  As more miners shut down their rigs less power is used and the difficulty decreases.

site your miner in iceland and make use of their abundant (and cheap) green power from geo-thermal energy. I am sure you can find a hosting option than makes sense for you.

As I sit here and think about Bitcoin I cant help but wonder how "Green" the Bitcoin system is. The system is designed to get harder as more miners go online with no upper limit, and the difficulty rises and more miners go online the more power is used. The extra power is consumed with no benefit other than a more encryption. Is there not a better way? Even though the technology is improving (eg more efficienct ASIC's) Network wide thats still a huge amount of power. I like the idea of and open p2p money system but can't help but feel there might be a more energy conscious way to go about it.