[H-GEN] A modern mining rig

[Russell Stuart]

On Fri, 2014-03-28 at 08:51 +0000, Benjamin Fowler wrote:
> So imagine then, that BK-like payment systems take off, and some
> developer gets some constant wrong somewhere when designing the
> protocol... And mining then ends up taking up a small but nevertheless
> sizeable percentage of world's energy output. That could be the
> dumbest manmade environmental catastrophe ever...

There is no "constant".  The difficulty of mining is adjusted by the
protocol so that on average one block is solved every 10 minutes.

In the end, the rate of mining of the network is determined by the
amount of energy the miners are collectively prepared to purchase.  As
it happens this proposition isn't hard to check.

The miners get their money from bitcoin mining rewards and fees.  The
mining reward is currently 25 bitcoin; the value of those bitcoins has
been fluctuating wildly, driven mainly by the Chinese government
generating rumours and laws about what their citizens can do with it.
Currently it's USD$500/bitcoin.  Looking at the Wikipedia entry on
electricity [0] I am not sure what a reasonable price would be.  I
choose a median US price of USD$0.12 / KWHr (AUD$0.13).  

So with that hand waving behind us we can calculate how much electricity
mining can pay for: 

    6 [rewards/hour] * 25 [BTC/reward] * 500 [$/BTC] / 0.12 [$ / (K Watt * Hour)]
  = 625,000 [K Watt]
  = 625e6 [Watt]

An ASIC mining chip [1] achieves around 1 Giga Hashes / Joule ([2] is
where you can buy such a thing).  So if the cost of mining was
electricity alone, I would expect the mining rate to be:

    625e6 [Joule/Second] * 1e9 [Hashes/Joule]  # Watt = Joule / Second
 = 625e15 [Hashes/Second]

The mining rate is currently around 42e15 [3] Hashes per Second, so an
order of magnitude lower.  This is to be expected as miners have other
costs besides electricity, like purchasing the mining rig itself.

To me, that is pretty conclusive evidence the rate of mining is
determined by keeping two things in balance:

  1.  The rewards and fees the miners are paid.
  2.  The cost of doing mining.

Since we have [2] and thus know the cost of a real mining rig, we can
have a little fun and determine what buying one of these might return us
in profits (or loss!).

Lets say they cost $80 each.  (You would have to buy enough of them to
make freight suitably small.)

With a mining rate of 100e15 Hashes per second, and assuming you get 4e9
Hashes per second out of this thing, your earnings per year would be:

     6 [reward/hour] * 8760 [hour/year] * 25 [BTC/reward] * 500 [$/BTC] * 4e9 [Hash/Second] / 100e15 [Hash/Second]
  =  $26/year.       [4]

Remember, you paid $80 for the thing.  However in reality $80 is very
expensive.  A tray of the mining chips costs around $5 per chip [5], and the
board is open source.  Some assembly is then required, of course.




[0] http://en.wikipedia.org/wiki/Electricity_pricing

[1] https://en.bitcoin.it/wiki/Mining_hardware_comparison#ASIC

[2] http://www.ebay.com/itm/111305230463

[3] https://blockchain.info/stats

[4] I said the current hash rate of the network is 42e15 hashes per
    second, but I used 100e15 hashes per second here.  The reason is
    mining networks hash rate is currently increasing exponentially:
      https://blockchain.info/charts/hash-rate?scale=1
    In case it isn't obvious, that curve means when investing in mining
    timing is everything.  And besides, as Australia's electricity
    costs are double elsewhere we aren't an ideal place to do it.

[5] http://www.bitfurystrikesback.com/product/bitfury-55nm-rev-2-chip-tray/