A Raspberry Pi only uses 2W at most: 400ma @ 5V. This has to come from a power supply that has some conversion efficiency of 50-80% or so. Let's be conservative and say we're using a 50% efficient wall plug transformer. Then each Raspberry is using 4W of wall power to generate the 2W it actually needs to run. Say electricity is $0.15 per kWh. Then in a year of 365*24 = 8760 hours, a single Raspberry Pi is going to use 8760 * 4 = 35,040 watt-hours or 35 kWh per year costing 35 * 0.15 = $5.25 of electricity per year.
Thus the power to run a Pi is thus real but pretty negligible in the overall cost picture. You can run 25 Pis nonstop for a solid year for the same amount of juice required to leave a 100W incandescent bulb on day and night for a year.
As a reasonable estimate, a single Pi costs around $120 as capital outlay to set it up for a year of service as an NXT node: $35 for the Pi itself; $30 more for the memory card, cables, cluster support gear, etc; $ 5 for a year's worth of power; and say $50 per year (around $4 / mo) to park it in a server farm. A network of 300 Pis thus costs $36,00 to set up and run for a year (of which $15,000 per year or $1250 per month is "office rental" and "human tech support" and "internet bandwidth").
If there are truly 5000 NXT per day generated in transaction fees as assumed above, that's 365*5000 = 1,825,000 NXT per year. If you are using 300 Pis at $36,000 per year to capture ALL those transaction fees, you are spending $36,000 / 1,825,000 NXT = $0.02 per NXT or around 2 pennies per NXT to capture the NXT using Raspberry Pis. At 0.00005 BTC/NXT and $800/BCT, a NXT is currently worth $0.04 or 4 pennies each.
Rough ballpark estimate: Capturing NXT with Raspberry Pis could double your money in a year at current NXT prices. The Raspberry cost is fixed; the return-on-investment ROI gets better and better the higher and higher NXT goes.
http://coen.boisestate.edu/ece/files/2013/05/Creating.a.Raspberry.Pi-Based.Beowulf.Cluster_v2.pdfI seem to recall that people are setting up VPSs for around $20 per month or $240 per year, which is double the cost of setting up a Raspberry Pi for a year as outlined above. Thus while NXT is at its current $0.04 each, capturing NXT with VPSs is a break-even proposition.
These are actually great numbers. The return-on-investment or ROI for forging NXT is positive NOW and will be even more profitable when NXT coins become more valuable. Thus even with NXT as a Proof-of-Stake coin without mining, there is STILL financial motivation for people to run the nodes required to keep the NXT system going. For a pre-mined coin, that is a remarkable statement.
Oh and the daily cost of the NXT support web running on 300 Raspberry Pi nodes is $36,000 / 365 = $98.63 per day, call it $100 per day. Do it on 300 VPSs instead, and the daily NXT support web cost is around $200 per day.
I have been adding some text to the NXT whitepaper at the wiki and my angle for the section I'm writing is Bitcoin Problem / NXT Solution. Here's some interesting text I've researched, which considering my earlier Raspberry Pi calculations above, convinces me that Bitcoin is a dinosaur with a meteor in the sky coming in, while NXT is a cute furry mammal waiting patiently in its burrow for its time on the world stage:
PoW's Negative Ecological Consequences
Confirming transactions for existing Bitcoins, and creating new Bitcoins to go into circulation, requires enormous background computing power that must operate continuously. This computing power is provided by so-called "mining rigs" operated by "miners". Bitcoin miners compete among themselves to add the next transaction block to the overall Bitcoin blockchain. This is done by "hashing" - bundling all Bitcoin transactions occurring over the past ten minutes and trying to encrypt them into a block of data that also coincidentally has a certain number of consecutive zeros in it. Most trial blocks generated by a miner's hashing effort don't have this target number of zeros, so they make a slight change and try again. A billion attempts to find this "winning" block is called a "gigahash", with a mining rig being rated by how many gigahashes it can perform in a second, denoted by "GH/sec". A winning miner who is first to generate the next needle-in-a-haystack, cryptographically-correct Bitcoin block receives a reward of 50 newly-mined Bitcoins - a reward worth, at the time of this writing, around $50,000. This competition among miners, with its hefty reward, repeats itself over and over and over every ten minutes or so - by early 2014 generating rewards of over 7000 bitcoins per day worth around $7 million dollars per day.
With so much money at stake, miners have supported a blistering arms race in mining rig technology to better their odds of winning. Originally Bitcoins were mined using the central processing unit (CPU) of a typical desktop computer. Then the specialized graphics processing unit (GPU) chips in high-end video cards were used to increase speeds. Field programmable gate array (FPGA) chips were pressed into service next, followed by mining rigs specialized application specific integrated circuits (ASIC) chips. ASIC technology is the top of the line for Bitcoin miners, but the arms race continues with various generations of ASIC chips now coming into service. The current generation of ASIC chips are the so-called 65nm units, based on the size of their microscopic transistors in nanometers. These are due to be replaced by 28nm ASICs in early-2014 and 20nm units by mid-2014. An example of an upcoming state-of-the-art mining rig would be a Butterfly Labs "Monarch" 28nm ASIC card, which is to provide 600GH/sec for an electricity consumption of 350 watts and a price of $2100. On the horizon is a card from Hashblaster slated to have three 20nm ASIC chips providing 3300 GH/sec for 1800 watts of power consumption. Most operational mining rigs will probably be upgraded to this standard of performance and efficiency by mid-2014.
The mining rig infrastructure currently in place to support ongoing Bitcoin operations is astounding. Bitcoin ASICs are idiot savants - they are able to do only the Bitcoin block calculation and nothing more, but they can do that one calculation at supercomputer speeds. In November 2013, Forbes magazine ran an article entitled, "Global Bitcoin Computing Power Now 256 Times Faster Than Top 500 Supercomputers, Combined!". In mid January 2014, statistics maintained at blockchain.info showed that ongoing support of Bitcoin operations required a continuous hash rate of around 18 million GH/sec. During an day of 86,400 seconds, this means around 1.5 trillion trial blocks were generated and rejected by Bitcoin miners looking for the magic 144 blocks that would net them $7 million. Thus around 99.99999999 % of all Bitcoin computation go not to curing cancer by modeling DNA or searching for radio signals from E.T - instead, they are totally wasted computations.
The power and cost involved in this wasteful background miner support of Bitcoin is enormous. If all Bitcoin mining rigs had "Monarch" levels of capability as described above - which they will not, until they are upgraded - they would represent a pool of 30,000 machines costing over $63 million and consuming over 10 megawatts of continuous power while running up an electricity bill of over $3.5 million per day. The real numbers are significantly higher for the current, less-efficient mining rig pool of machines actually supporting Bitcoin today. And these numbers are currently headed upward in an exponential growth curve as Bitcoin marches from its current one transaction per second to its current maximum of seven transactions per second.
