Flickr / zcopleyBitcoins are a bit like the Internet. Or, rather, the Internet as it was in the mid ’90s: something strange, coming out of geekdom into mainstream perception, greeted by puzzlement over how it works, why it works and why anyone would think it’s useful.
A common analogy for Bitcoins is gold: like gold, they have value only because people want them, the supply is limited, more Bitcoins are created only by ‘mining’ for them and the difficulty in mining grows as they are mined. But rather than being stored in underground vaults Bitcoins are simply entries in a notional ledger held across many computers around the world.
The actual mining of Bitcoins is by a purely mathematical process. A useful analogy is with the search for prime numbers: it used to be fairly easy to find the small ones (Eratothenes in Ancient Greece produced the first algorithm for finding them). But as they were found it got harder to find the larger ones. Nowadays researchers use advanced high-performance computers to find them and their achievements are noted by the mathematical community (for example, the University of Tennessee maintains a list of the highest 5000).
For Bitcoins the search is not actually for prime numbers but to find a sequence of data (called a ‘block’) that produces a particular pattern when the Bitcoin ‘hash’ algorithm is applied to the data. When a match occurs the miner obtains a bounty of Bitcoins (and also a fee if that block was used to certify a transaction). The size of the bounty reduces as Bitcoins around the world are mined.
The difficulty of the search is also increased so that it becomes computationally more difficult to find a match. These two effects combine to reduce over time the rate at which Bitcoins are produced and mimic the production rate of a commodity like gold. At some point new Bitcoins will not be produced and the only incentive for miners will be transaction fees.
The raw performance of a Bitcoin mine is measured in hashes per second (i.e. the number of tries per second to find a block). With the difficulty and bounty settings it becomes possible to calculate the expected rate of Bitcoin production. An ordinary computer can do this work running software and typical high-end PC (using an Intel Core i7) can perform about 6.7MH/s (6.7 million hashes per second). This would have been quite successful at mining a couple of years ago but today it would have an expected rate of 0.0005BTC per day (this is a actuarial measure: a miner finds a block or doesn’t, and in this case it would likely take decades to find one).
Fortunately, there are more powerful computers available quite cheaply: your graphics card used for gaming is actually a massively parallel computer (quite often containing more than a thousand CPUs). My PC’s GPU (Graphics Processing Unit) contains a Radeon 6870 chip that will do 300MH/s. That’s enough for 0.02BTC per day (it would take on average more than three years to find a block). But there’s the cost of the electricity to run the PC (and the cost of the PC) to take into account. The PC uses about 250W of electricity (the graphics card alone uses 150W).
At current difficulty and electricity price ($0.15/kWh) it would cost $44 in electricity for each Bitcoin mined. But crucially, the low probability of finding a block means that the economics are likely to have shifted before one is found. The only viable way to mine Bitcoins with a GPU is to have lots of fully-amortised cards in a datacenter running right now. In short, if you didn’t get into Bitcoin mining ages ago with your PC, you’re too late now.
There is another way to calculate hashes faster: using FPGAs (Field Programmable Gate Arrays). In essence this is custom programmable electronics. Rather than using general purpose processors running software, the hardware is directly performing the calculations. FPGA programming is quite specialist engineering used in applications like HFT (High Frequency Trading) and ULLDMA (Ultra Low Latency Direct Market Access). An FPGA mining rig can achieve a hash rate as high as 800MH/s, but at some expense (the cost can run to $1000 per module).
A high-end rig of 8 modules could expect to generate 0.5BTC per day on average (allowing the hardware costs to be amortized over 130 days at current BTC prices and difficulty levels). However, anyone who invested in FPGAs recently will likely have lost their money: a new generation of mining hardware has hit the market rendering all previous mining rigs obsolete.
ASICs (Application Specific Integrated Circuits) are the kind of chips found in mobile phones and other electronics. They are expensive and time-consuming to design (taking months or even years) and hard to get the production right (requiring large silicon wafers, people in cleanroom bunny suits, etc.) but once they work properly can be manufactured cheaply in quantity.
The first dedicated mining rigs built from ASICs shipped a few weeks ago: Avalon claims their first product achieves 65GH/s, uses 650W of electricity and is priced at ฿72 (yes, it’s priced in BTC). Butterfly Labs say they will be shipping imminently kit that will do 50GH/s (currently priced at $2499). This should at current difficulties find a block about every 6 days, giving a BTC rate of 3.75BTC per day. At current BTC/USD prices that would allow the hardware cost to be amortized over just 5 days.
ASIC-based Bitcoin mining has created a step-change in Bitcoin mining economics. In keeping with the analogy with gold, there is now a rush to be amongst the first with the new mining technology and scoop the current set of Bitcoin nuggets that are easily mined before someone else does – the difficulty will step up as ASIC-based rigs become widespread. And just as in California in 1849, the suppliers of mining equipment are likely to be the ones who make the most money. In the end the limiting factor will be the amortization costs of ASIC rigs (given the economics of silicon this need not be high) and electricity (perhaps the future of Iceland’s economy is hosting air-cooled Bitcoin mining rigs powered by volcanoes?)
As to whether all of this is a fool’s pursuit and Bitcoins are really modern-day tulips (albeit tulips that weigh nothing, don’t decay and have no storage costs) it is difficult to say. In the end a form of money is a human construct that has utility when it is widely agreed that it has utility. Of course, it could be a huge geeky joke. Like the Internet used to be.