Is Bitcoin viable, energy wise?

[GambitBTC]

My house is powered by bitcoin

[myrkul]

To be honest, I haven't really found a principled method of discovering how close we are to that lower limit. Right now there is plenty of new demand so it is kind of hard to tell. My expectation is that it will always be some small fraction of the speculative volume.

A hopefully growing fraction. If Bitcoin is only ever about the speculating, we'll never get out of the "tulip bulb in a display case" phase.

[myrkul]

To be honest, I haven't really found a principled method of discovering how close we are to that lower limit. Right now there is plenty of new demand so it is kind of hard to tell. My expectation is that it will always be some small fraction of the speculative volume.

A hopefully growing fraction. If Bitcoin is only ever about the speculating, we'll never get out of the "tulip bulb in a display case" phase.

Tulip bulbs are still a going concern BTW.

So they are, but you won't find any in display cases.

But as far as Bicoin's speculative bubbling goes, we will see lots more of that ongoing. Just wait until it starts crossing over the inflation rates of prominent currencies in the coming decade or so, we haven't seen anything yet. Concerns about how to effectively manage the energy intensiveness of operating the network are well founded IMO.

Probably. But I can see "bitcoin heaters" being fairly profitable, or at least popular.

[mobodick]

You're not looking at total cost, for the system. You're trying to isolate one portion (the transaction portion) and pretend all the rest doesn't really actually matter...

No, i just want to keep things apart as much as possible.
Mining costs are not transaction costs.

I'm sorry, maybe I misunderstood somewhere, but doesn't mining confirm the transactions?

Yes, but the ammount of confirmed transactions is unrelated to the hashing power.

That is not entirely true.

The number and volume of transactions is a measure of velocity, which affects the prevailing exchange rate, which ultimately drives difficulty. So, it does not happen immediately but transactions do eventually translate to hashing power. The feedback loop is inherent to the design.

Difficulty follows hashing power pretty tightly as far as i can see. So i don't think that difficulty is driven by transactions.
Anyway, isn't difficulty supposed to keep the number of blocks found constant despite hashrate?

It is driven by the decision of node operators to add or subtract hashing power. Ultimately, this decision is driven by the prevailing sentiment of node operators of an expectation of profit. Price leads difficulty.

Aah, here i make a distinction between expectations about price and the actual price.
I would agree that expectations lead hashpower. But expectations are future prices, at best.

[mobodick]

You're not looking at total cost, for the system. You're trying to isolate one portion (the transaction portion) and pretend all the rest doesn't really actually matter...

No, i just want to keep things apart as much as possible.
Mining costs are not transaction costs.

Mining costs ARE transaction costs. Transactions and mining costs are inseparable. And once the system reaches maturity, mining costs will ONLY be transaction costs (since there's nothing else to "mine".

You can't say that the transaction costs is all that matters on the traditional banking side, but that the entire system costs matters on the Bitcoin side.

You have to compare apples to apples. Either compare transaction costs only, or compare complete costs (infrastructure and transactions together). Since you can't separate transactions from infrastructure (on either side, honestly), looking at 'total cost' is the only apples to apples comparison that can be made...

Comparing traditional banking as a 'transactions only' issue, to the complete infrastructure of Bitcoin just skews the discussion, and isn't an honest way of addressing it.

Does that mean that in the end the hashpower will scale lineary with the ammount of transactions?
If that is the case then it would indeed be better to just count them as one system.
If not then i'll stick with looking at them as two separated but related systems.

[Adrian-x]

Aah, here i make a distinction between expectations about price and the actual price.
I would agree that expectations lead hashpower. But expectations are future prices, at best.

Yes and unrealized expectations create either unutilized hardware (reserve hardware) as people stop mining. This is what seems to have happened after the 2011 peek. Or reserve Bitcoins as I think will be the case with asic miners.

[Adrian-x]

Tulip bulbs are still a going concern BTW.

But as far as Bicoin's speculative bubbling goes, we will see lots more of that ongoing. Just wait until it starts crossing over the inflation rates of prominent currencies in the coming decade or so, we haven't seen anything yet. Concerns about how to effectively manage the energy intensiveness of operating the network are well founded IMO.

@ chodpaba
Can you explain what is meant by crossing over... in bold above?

[myrkul]

Someone should call me out on this shit.

Didn't you just do that?

[mobodick]

You're not looking at total cost, for the system. You're trying to isolate one portion (the transaction portion) and pretend all the rest doesn't really actually matter...

No, i just want to keep things apart as much as possible.
Mining costs are not transaction costs.

I'm sorry, maybe I misunderstood somewhere, but doesn't mining confirm the transactions?

Yes, but the ammount of confirmed transactions is unrelated to the hashing power.

That is not entirely true.

The number and volume of transactions is a measure of velocity, which affects the prevailing exchange rate, which ultimately drives difficulty. So, it does not happen immediately but transactions do eventually translate to hashing power. The feedback loop is inherent to the design.

Difficulty follows hashing power pretty tightly as far as i can see. So i don't think that difficulty is driven by transactions.
Anyway, isn't difficulty supposed to keep the number of blocks found constant despite hashrate?

It is driven by the decision of node operators to add or subtract hashing power. Ultimately, this decision is driven by the prevailing sentiment of node operators of an expectation of profit. Price leads difficulty.

Aah, here i make a distinction between expectations about price and the actual price.
I would agree that expectations lead hashpower. But expectations are future prices, at best.

Take a moving average of price and compare it to difficulty.

You will find that price leads difficulty in this way by a fairly predictable ratio. It is pretty darn reliable.

I was just describing a theoretic mechanism by which it works out that way. 

I don't think you can conclude that a price component is driving hashrate. I think the data is flawed and that actual hashrate could just as well drive price.
I'm not denying that there is a correlation, tho.

Anyway, to test your claim i had to take a 100 day average to get from price to something that is shifted enough to fall on top of the difficulty data.
The difficulty (and according to the data, the hashrate) is correlated to price so closely that peaks happen even when price changes at that time suggest the outlook is bad. If i look at the 2011 peak i can see that according to the available data more hashrate was added after the crash/hack/turnaround then before the 'crash'. Even better, hashrate and difficulty only peaked when bitcoin was in freefall and already around $10. Difficulty then stayed pretty constant for some time while price collapsed to under $5.

That doesn't make a lot of sense to me. If price expectation was the actual drive then mining would have started collapsing right after the crash and would not have slavishly repeated the peak of the price.  It doesn't make a lot of sense for a node operator to add hashing power after price peaked and is on its way down.
What is much more likely is that actual hashrate was much closer in time to the price movements.
What also seems likely to me is that there is alot of high frequency content missing from the data.

But then again, the graphs fit so well  

[Peter Lambert]

Difficulty follows hashing power pretty tightly as far as i can see. So i don't think that difficulty is driven by transactions.
Anyway, isn't difficulty supposed to keep the number of blocks found constant despite hashrate?

It is driven by the decision of node operators to add or subtract hashing power. Ultimately, this decision is driven by the prevailing sentiment of node operators of an expectation of profit. Price leads difficulty.

Aah, here i make a distinction between expectations about price and the actual price.
I would agree that expectations lead hashpower. But expectations are future prices, at best.

Take a moving average of price and compare it to difficulty.

You will find that price leads difficulty in this way by a fairly predictable ratio. It is pretty darn reliable.

I was just describing a theoretic mechanism by which it works out that way.  

I don't think you can conclude that a price component is driving hashrate. I think the data is flawed and that actual hashrate could just as well drive price.
I'm not denying that there is a correlation, tho.

Anyway, to test your claim i had to take a 100 day average to get from price to something that is shifted enough to fall on top of the difficulty data.
The difficulty (and according to the data, the hashrate) is correlated to price so closely that peaks happen even when price changes at that time suggest the outlook is bad. If i look at the 2011 peak i can see that according to the available data more hashrate was added after the crash/hack/turnaround then before the 'crash'. Even better, hashrate and difficulty only peaked when bitcoin was in freefall and already around $10. Difficulty then stayed pretty constant for some time while price collapsed to under $5.

That doesn't make a lot of sense to me. If price expectation was the actual drive then mining would have started collapsing right after the crash and would not have slavishly repeated the peak of the price.  It doesn't make a lot of sense for a node operator to add hashing power after price peaked and is on its way down.
What is much more likely is that actual hashrate was much closer in time to the price movements.
What also seems likely to me is that there is alot of high frequency content missing from the data.

But then again, the graphs fit so well  

You have to remember that back in 2011 everybody was saying "Sure the price has been dropping, but it could jump up again any moment now!" What happens is that as the price rises people see difficulty is not as high as the price, making mining profitable, so they start building a mining rig. That takes time, and there will be people who do not get their mining going until after the price has peaked. But they do not want to have wasted all that energy they spent building this thing, so they start running it anyway. Also, when the price drops, people who are already mining have already done the hard work of getting the thing set up, so they keep on mining and hoping the price will go back up again; if they are still marginally profitable they will keep mining but not add any more capacity.

The other thing to consider is that technology has been improving, and so as more efficient ways of mining come to market the total hash rate will increase even if the price remains constant.

[mhps]

It costs you very little energy to carry cash or a bank card. Bitcoin requires multiple computers and a whole lot of actual computation to even do one transaction. It is much more expensive energy wise than any other form of common currency.

Uh... How do you think your credit card works? [Hint: It requires multiple computers and a whole lot of actual computation to even do one transaction.]

Credit card, and most  of the financial infrastructure we have with the exception of the FED, and bitcoin are not mutually exclusive. The argument that traditional financial industry is much less  resource-efficient than a similarly sized bitcoin-based financial industry may be true but the difference is not that big. You could have bitcoin credit cards once the bitcoin base is big enough and issuing a bitcoin credit card is profitable. Exactly because of the high cost of maintaining the blockchain for a personal node, most lay people *will* use a financial institute of somesort (e.g. an online wallet) to manage their money.  Once people'e income goes into financial institues, all actors of the ecosystem come to live.

[mobodick]

You have to remember that back in 2011 everybody was saying "Sure the price has been dropping, but it could jump up again any moment now!" What happens is that as the price rises people see difficulty is not as high as the price, making mining profitable, so they start building a mining rig. That takes time, and there will be people who do not get their mining going until after the price has peaked. But they do not want to have wasted all that energy they spent building this thing, so they start running it anyway. Also, when the price drops, people who are already mining have already done the hard work of getting the thing set up, so they keep on mining and hoping the price will go back up again; if they are still marginally profitable they will keep mining but not add any more capacity.

The other thing to consider is that technology has been improving, and so as more efficient ways of mining come to market the total hash rate will increase even if the price remains constant.

Yes, and these are the dynamics i came up with to justify the difficulty graph.

But i cannot understand the 2 month delay in peak of hashpower compared to price.
In those days i think most people mined on GPUs.
2 months is way too long to order and build a couple of pc's with GPU's.
It takes me maybe an hour to get from pc components to a ready to use computer. That includes OS and drivers and all that. Ordering components doesn't take 2 months either.
So at the very least one would expect that there would have been a strong negative effect on the difficulty just past the price peak. People investing in mining rigs could have simply canceled their orders.
I can remember that right after the price peak there were a lot of people forced to sell their rig because price expectations were not met. There was a sell off of sorts. I cannot find that back in the data. It looks as if the miners are completely oblivious to current price and always react to past price with a more or less fixed latency. Somehow the information about past price tunneled its way into the future without being affected by months of negative change just to have a singular effect on capacity in the future. How did this information of past price survive so far into the future?

The timescale of the data suggests that miners must have been ordering en masse after the price peak, while the price was going down.
For me this is a strong indication that the difficulty data is delayed somehow and the actual reaction of miners was much closer to the actual price, possibly even preceding it.
The data does not make any sense to me otherwise. Miners do not live in a universe where they only see an old price and are only capable of reacting months afterwards. Most of them are capable of reacting within a week to market changes but somehow they were incredibly blind and completely irrational when looking only at the price changes.

Another thing that bugs me is that the data is so smooth. I'm pretty sure hat hashrate in those days was much more erratic. There were tests with clusters and stuff like that. None of those show up on in the hashrate data. This suggests that the available data is already averaged (also implying a latency).
As i said above, i needed to take a 100 day running average to get from price to difficulty. 100 days is a loooong time. I see no reason for the network to have such a strong lingering memory of a sentiment from 3 months back.

So while i'm tempted to see the simple side of things there are still some details that don't make sense at all to me. I cannot find a justification for the way difficulty repeats movements with such a high latency and smoothness.

[mobodick]

Beware the Psychologist's Fallacy. You are trying to make an inference based on a narrow population which fits your estimations... The data is the data.

I would say that you need the miners to be a narrow sort of folk much more than i do.

For the peak to be delayed this consistently it would require the bulk of the people that added the hashrate to behave in a singular way.
The data tells us that most miners started adding capacity months after the rise and fall in price and follow the movements of the price of the past. This would require the bulk of the adders to only be able to react to price in 2 months from now. This canot be the reality for the bulk of the miners.

My view is more that of a biologist than a psychologist. There should have been a better distribution and effects of price should already have been visible a short time after price drop because in a natural distribution there are miners that do check the price and do cancel orders. The data suggests that these people are non-existent which i find very strange. The set of behaviours that are needed to go from price fluctuation to difficulty fluctuation as seen in the data is actually pretty narrow.
Can you explain why most miners in that period failed to react to the sudden drop in price?

Sure, the data is the data. And flawed data is flawed data.
All the data i find is some form of estimate.
The interesting question is how this data relates to actual hash rate progression.
And we need the actual hashrates to make causal relations.

[mobodick]

As to their specific motivations I can only speculate, but I would presume they are acting rationally. To me the data would suggest that miners have limited resources and are taking a long view. They are basically bootstrapping their operations, trying to operate and grow them based on funds that come from the operations themselves. The delay comes not from their efficiency in building new rigs, but from the decision making process of where future price is going, and balancing sort term vs. long term interests.

Call it what you want, i still find the graph irrational. Most of the miners supposedly decided to start mining some time after bitcoin started going down. The worse bitcoin got, the more people started mining, according to the data.
I can't see any motivation for miners to start investing exactly in that period. Future price looked pretty bad at that time but the bulk of the hashing power didn't react to price despite price being widely available. Even if you assume the bulk of these people were slow decision makers you still would have to live with the fact that none of these people noticed anything about the price collapse (or maybe they just all ignored it). Even if someone takes a long time to make a decision they will most of the time incorporate the latest information available. But what we see here is none of that. We see only that a trend is repeated seemingly without considering information that was already available.
It just does not make sense for the bulk of the people to start hashig after the price has been steadily declining for months.
Price goes up, price goes down. Where is the pressure that will in a few months realize an increase in hashing power? Why would we see a bump in hashrate after the bump that supposedly caused it is already gone?
I think your idea requires most of these miners to be a very particular decision maker that acts on past information and even despite different current price constantly acts on a past price.

It looks much more like the decision making process already took place months before it was registered in the data.

[mobodick]

The rational case is that the lower the exchange rate goes the more upside potential there is.

Its like saying a btc mined when it is cheap is more valuable then a bitcoin mined when it is expensive because it has more upside potential.

[mobodick]

The rational case is that the lower the exchange rate goes the more upside potential there is.

Its like saying a btc mined when it is cheap is more valuable then a bitcoin mined when it is expensive because it has more upside potential.

Not really, the cost to mine is related to difficulty, and mining when the difficulty is lower is cheaper. If you can not mine all the time then it makes sense to mine when it is cheaper for you to do so, particularly when you expect that it will be more expensive later on.

Sure, so with a falling price and a growing difficulty it would make no sense, according to your interpretation, to invest in mining capacity.
Yet this is exactly what the data shows.
Your interpretation does not apply if you assume the data is correct.

[mobodick]

The rational case is that the lower the exchange rate goes the more upside potential there is.

Its like saying a btc mined when it is cheap is more valuable then a bitcoin mined when it is expensive because it has more upside potential.

Not really, the cost to mine is related to difficulty, and mining when the difficulty is lower is cheaper. If you can not mine all the time then it makes sense to mine when it is cheaper for you to do so, particularly when you expect that it will be more expensive later on.

Sure, so with a falling price and a growing difficulty it would make no sense, according to your interpretation, to invest in mining capacity.
Yet this is exactly what the data shows.
Your interpretation does not apply if you assume the data is correct.

It does not apply to node operators who can not tolerate much risk. The case I was describing was for miners with deep pockets... You know, the ones with the most hashing power.

Even people with deep pockets would be fools to expand when prices are going down AND difficulty is going up fast.
So your super know-it-all node operator just got a little bit more special just to satisfy the chart.

You also have to figure that mining just got a lot more cost effective with the ASICS.

I don't think so. Their popularity makes that the minig market will become saturated.
Difficulty will compensate and everyone owning an ASIC will get a portion of the same 25 BTC per same time unit times their portion of the hashrate.
With mining hasrate is not important. What is important is your relative hashrate compared to the rest of the network. In the end the daily reward in BTC is the same (bar reward halvings).
So it is only now that owning an ASIC is a real opportunity.
Once all ordered ASICs are delivered the 100% ROI point for newly delivered devices will be very very far away, if not infinitely.
ASICs are not cost effective in the long run because everyone has access to them.

The interesting thing is that most of these devices were ordered last year, when the price didn't boom yet (the recent peak).
So i can say that the recent price peak was the result of projected hash power.
The recent bubble only started to get traction after Avalon started delivery. At that time Avalon had already sold their second batch of ASICs.
But then BFL delayed and the price fell agian. Avalon could not keep the momentum going and we had a crash.
Once (if?) BFL really starts delivering for real the price will soar.
There was a reaction to price when BFL recently sent out a few units to devs. It didn't hold once people noticed that it's only 6 units and that the rest will come later, much later.

Anyway, ASICs are a completely different game for miners wanting to expand. GPUs were widely available and anyone could get a fairly beefy mining rig within a week or two. With ASICs at this moment most of the already bought hashrate is still dormant.

In this recent bubble is was quite clear that price was driven by hashrate expectations.
And not the other way around, like you keep arguing and making up new things to justify it for.
But it's also not purely hashrate, like i was saying before. It seems to be the expectation of hashrate that drives price the most.

[mobodick]

The ASICs will get cheaper and more energy efficient. And the exchange rate will climb. Miners will still be able to optimize profitability by smartly allocating capital and operating budgets to best advantage the waxing and waning of the exchange rate. And so we will continue to see the relationship between difficulty and price.

Hashrate does not drive price, like at all. It is the other way around.

And I would not choose to make a bet against Moore's Law.

Build your own models... And actually test them. Go Bayesian with it, trust me, it's a blast.

I'm sorry but what you say makes no sense.
On the one hand you say that hashrate doesn't drive price, like at all.
On the other hand you don't want to bet against moores law, suggesting that bitcoins value is in hashrate.
Please make up your mind!

Anyway, you won't gain anything by adopting a more efficient technology (besides the small window of opportunity when you have a superior mining device while the rest doesn't yet). By buying the latest tech you only prevent a loss in BTC income.
It's only your relative hashrate that counts because in the end the same 25 BTC per time unit will be divided amongst the contributers, no matter the current hashrate.