Lately, at one of our Bitcoin Startup Berlin meetups, I did a short talk about Proof of WTF: Proof of Work, Proof of Stake, Proof of Burn, Proof of Resource are a core ingredient of crypto currency technologies. And as I’m not a software architect or programmer or anything close, I will not bore you with disseminating too much semi-knowledge. Let’s just leave it like this: if you want to validate a transaction (and for the purpose of seigniorage), Bitcoin-alike crypto currencies need a network of computers agreeing upon a Proof of Something. If you want to know more, try the Proof of-links above (or ask me for the presentation, which is unfortunately riddled with too many copyrighted images to share freely anywhere).
But back to work, literally. Bitcoin, being somewhat the genetic father of practically all p2p crypto currencies on the planet, is made possible by the proof of work: a bunch of computers solving a software riddle of increasing complexity. Nakamoto’s solution was brilliant, as a proof of concept. Unfortunately, it does have some side effects. In its current state, Bitcoin has to rely on a huge network of professional miners to validate transactions, which spin out new coins as well.
At the Berlin Inside Bitcoin conference, somebody from the audience asked two of those miners on the stage, how a Bitcoin born in 2014 would look like. Would it still use a Proof of Work mechanism? Or would another solution like Proof of Stake be preferable? Now guess the answer of the guys making a living from running a network of PoW-machines. Wrong. Their statement was pretty clear: Proof of Work works, but it comes at a price. Proof of Stake or other concepts would be more future-proof.
How comes?
In physics, power is the rate of doing work. So you might rightly infer that Bitcoin’s Proof of Work based system, with its increasing complexities, might need more and more power. You’re right.
If you follow the developments of Bitcoin mining, you will have noticed some strange developments.
- In the beginning, you could mine some coins with your desktop hardware, on your CPU (while browsing the web and fiddling around with a spreadsheet). Those days are long gone.
- Next step: try running the mining program on your GPU, the graphical subsystem of your PC. This worked nicely for quite a while. But even high powered multi boards GPU monsters couldn’t keep up with the demanding network.
- So, since quite some time, we’re into ASIC mining: single purpose computing units, which are just able to handle a single task: Bitcoin mining. And, maybe, heat your apartment. Literally.
As there are too many myths and not enough facts around, I tried some very simple calculations. To be honest, the results are something between totally scary and kind of reassuring.
What I did was the following:
Blockchain.info is publishing some stats on a regular base: http://blockchain.info/stats. They are all nice and interesting, but the only thing we really need here is the last number on the page, the hash rate.
Currently, as of 2014-04-09, the Bitcoin network is running at 46,848,838.03 GH/s
Now, let’s have a look at mining hardware. Here we go, the mining hardware comparison table. https://en.bitcoin.it/wiki/Mining_hardware_comparison
A brave spreadsheet soldier would normalize all those number. I’m not, so I just did a sloppy calculation, coming to the following:
- CPU mining is running on average at approximately 6.67 KW/GH
- GPU mining needs round about 0.5 KW/GH
- totally new ASICs hum smoothly at 1 W/GH
Now, according to the EIA, a typical US nuclear power plant generates 11,800,000,000 MW/h per year.
This translates into the following:
- running the current Bitcoin network on CPU mining hardware would need 83 fully loaded nuclear power plants
- with GPU mining, we still would need six extra plants
- 2014 ASICs would reduce the load to 0.01
So, basically, we’re currently running at 0.13 nuclear power plants to power up the Bitcoin network.
https://docs.google.com/spreadsheets/d/1_b-H1iA3LSbetHdiDEt46WKEhr-rnW1Jrq1SgZeW4Ao/edit?usp=sharing
Could be better, could be worse. But what are the implications? One line of thinking goes like this: the Bitcoin network is like totally useful. Spending some energy on something totally useful should not be a big deal. Eat that, tree hugger! (OK, skip the last part.)
But my besides hurting my bleeding eco-friendly heart, the power hunger of Proof of Work has some other side effects as well. The armament race for faster, less power-hungry mining hardware definitely leads to a centralization of a core part of the network. Which, in the long run, makes the network less secure: it becomes more centralized, 51% attacks become more likely.
At the same time, the price for running the network is currently mostly paid by seigniorage, the mining of new coins. But there will be a certain point in the future, where the networks has to rely more upon transaction fees. The more power-hungry the network, the higher the cost of transactions. Because, even if you think climate change is made up by black helicopter flying aliens from a socialist parallel universe, you still have to somehow pay your utility bills.
So, I guess: sooner or later Bitcoin will have to switch the base technology. Because Proof of Work works well. But Proof of Stake seems like a more sustainable solution.
gert.is 