I keep some notes on the topic http://www.gwern.net/Notes#a-bitcoinbittorrent-driven-econom... but so far I haven't seen any proposals which might work.
When a block of transactions is successfully hashed, the miner is awarded new bitcoins, as an incentive to do that necessary work of putting those transactions in order.
I was worried about energy consumption, until I worked out how much the numbers. Starting with an extreme assumption that the bitcoin money supply reaches parity with the dollar by 2021 or so, total power consumption used directly for mining is only a couple gigawatts, about a tenth of a percent of the world's total. I worked that out like this:
First assume the cost of mining equals the value returned. If mining is very profitable, more people will mine and difficulty goes up. If mining loses money, people drop out and difficulty goes down. Overall mining will always tend to be in the neighborhood of breakeven, with just barely enough profit to keep people from dropping out (though that's the equilibrium, and in periods of rapid growth we could be far from equilibrium at times).
Pick a date, figure out how many reward halvings we'll have between now and then, and you get how many coins we award every ten minutes. Multiply by coin price to get value awarded each time. If you assume electricity is half the mining cost, divide that value in half. Then divide by the cost per kWh to get kWh per ten minutes. Multiple by six to get kWh per hour, which is the same as just kilowatts.
And now you know how much power production you need to support the bitcoin economy given your assumptions.
The problem is that 'useful' calculations aren't suited for extremely fast verification.
If you didn't require a difficult computation, anyone could compute a different order of transactions and it would look just as valid. Easier hashing would not accomplish the goal of making a universally-agreed-upon transaction order.
That's only true for GPU/PC based miners.
66GH/s @ 600Watts, Power is 1-2% of mining cost at current Difficulty / $Exchange / power cost ($0.12). Here is a link to info about the "free nights" plan. https://www.txu.com/residential/promotions/mass/free-nights....
The ASIC based system is much more power efficient than GPU/PC systems. It's even better than FPGA based systems which are (~10-20x) more power efficient than GPU/PC systems.
If Bitcoin mining economics are able to sustain ASIC production, it will probably render FPGA based systems uneconomical.
Even with just two viable ASIC production runs, I suspect that GPU/PC mining is dead.
Thanks for the link, that's pretty cool...wish my utility did that.
Parameters that I changed:
Difficulty 6,968,775 (I don't really think it's likely to double)
Hash Rate 50,000 (since 100% uptime is unicorns)
Electricity rate (USD/kWh) 0.12 (my current rate)
Power consumption (W) 600
Results:
Coins per 24h at these conditions 3.6083 BTC (~3.6*20)
Power cost per 24h 1.73 USD
Revenue per day 71.70 USD (Gross)
Revenue Less power costs 69.97 USD (net)
It still looks very good to me, am I mistaken?
PS, I wish I could take advantage of the free nights, unfortunately, I expect that if this night time load-shedding problem persists for any length of time (a few years), someone will step in and find a way to arb it out, and any big investment in doing that on my part may be lost.
If these types of hardware end up finding use in a new problem space (like bitcoin mining), then in theory that could make them more popular, and thereby cheaper (as well as increasing the acceleration of the technology, leading to better, faster, hardware for everyone who uses them).
So in theory, even if Bitcoin mining doesn't have any "direct good" that comes from it, it could still be indirectly helpful to the larger scientific computing community.
Some issues with "useful" computational work is that it needs to be computationally easy to verify the result (a hash function is but other work may not be), also it may not be crptographically strong so someone could come up with some method to solve the problems much more efficiently than others and make the network insecure.
If you think the BitCoin project is entirely useless, even as a proof-of-concept, then yes, these ASICs are useless, but that's an entirely different discussion.
> ... toy problem, instead of addressing a real scientific problem that needs compute.
What, like generating Mersenne primes? A lot of these other so-called “real” “scientific” problems have virtually no scientific relevance. I think research into crypto-currencies is just as “real” as finding more optimal Golomb rulers or searching for alien lifesigns in radio noise.