Cryptocurrency has been on everyones agenda since gaining mainstream interest due to its highly speculative token trading market and massive winter price surges. This blog post should not be used as guidance or to follow for investment advice purposes at all, full stop, period. I am a technical developer exploring blockchain technologies and the engineering tasks involved. This is not a place to look for investment strategy but is instead research & development efforts conducted to ascertain use and growth of blockchain technologies. I built a custom designed GPU mining rig as part of those development efforts. The choice between GPU designs over ASIC platforms is that they can be switched onto alternative algorithms easily and serve multiple uses. GPU and performance components can also be easily resold on to other markets. Ethereum network will be switching to a proof-of-stake system instead of a proof-of-work consensus in the near future and as such this design is future-proofed to be applied to alternative currencies that use GPU as well (Monero, Zcash, Dash and Siacoin.). I designed the system to be ecological to run and also produce significant hashrate on multiple algorithms at low wattage. Here is a picture of the 12-GPU system I built and the technical analysis of one month operation.

12 GPU Mining Node Parts List

The following price list is the price of devices at time of purchase and may fluctuate from those shown at any moment based on demand for GPU's or specific parts.

• 12 GPU Stackable Frame - £115
• Intel Celeron G3900 2.8Ghz 2MB Smart Cache Box LGA1151 - £44
• ASUS B250 Mining Expert - £160
• CORSAIR DDR 2x8GB (16gb) 3200Mhz RAM (VENGEANCE LPX) - £200
• 12 x EVGA GEFORCE GTX 1070 Ti 8GB DDR5 - £5265
• 2 x Super Flower 8 Pack 2000W LEADEX PSU - £839.98
• 12 x PCI-E 1x to 16x Power USB3.0 GPU riser extenders - £60
• 10 x 120mm FANs - £30
• 250gb Samsung SSD - £40
• 11" LCD monitor panel - £90
• EthOS license - £28
  Total Hardware Costs £6871.98

Design Choices

The GPU's individually cost around £365 each before VAT and shipping, they were the most expensive purchase by far and a 6 - 8 GPU design is more commonly used. The expensive power supply units could also be saved on with cheaper less powerful units, the 2000W units used are overkill and the system uses about 800-900W from each PSU. However as the load put on the cables in mining systems can be high, I prefer to use PSU's that are much higher rated than needed to minimise environmental risks, power surges, lightening etc. The same applies to your power distribution units and wiring and cooling systems. I also frequently find that my cryptocurrency experiments require additional power from endpoints and will continue to use these PSU's beyond the life of this one node. Good PSU's have saved me in the past from inferior designs and poorly engineered products. You can also run miners completely headless but it'll make life easier fixing faults and troubleshooting your design with an 11inch screen to use ;-)

"Building computers is alot like putting Lego together!"

Build & Machine Maintenance

Building the machine took a few hours, the frame assembled like Ikea furniture and computer components only tend to fit one particular way. If you have never built a system before I do not suggest this as your first attempt. You don't need to use EthOS but it certainly makes configuration and installation a breeze.

The above graph shows the mining hash rate during its first month, this was during a period of frequent reboots while tuning of the GPU settings was performed. At the start of the graph you can see 6 GPU's hashing at around 180MH/s on Ethereum main net, then a doubling of power as 6 more GPU's are added. It can hash at around 380-370MH/s on Ethereum main network, however this comes with stability issues and requires tweaking of per-card configuration. 12 GPU's means that each one must be tweaked, I did 6 at a time then 3 and 2 until I found the problematic cards. Running in stock clock is stable at around 320-330MH/s but you can overclock them for higher rates. This all comes at the price of stability and can vary depending on your environment, every card added to the system increases this risk of failure when under-volting and overclocking memory on GPU's. I found you can overclock the memory as much as 650MHz on the GPU cards I obtained, however one or two of the codes prefer a lower clock for stability and do not run well at the highest clock speeds.

Controlling the power switch of the motherboard with a single-board computer helps for remote reboots and EthOS allows for remote configuration over a web server. This allows you to tweak and remotely reboot the rigs when encountering difficulties or GPU cards begin to fail.

On the subject of heating and cooling, the miner added a barely audible humming sound that was enough to disrupt phone calls and microphones. However the heat and temperature of the system is stable and the room increased in temperature by only a few degrees. It is colder this time of year and environmental differences can impact performance but it did not replace my winter heating as I had hoped. This is not the case with ASIC designs which output large amounts of heat and insufferable noise levels making them more useful as replacement heaters than actual blockchain work horses.

You can see here that the cards once set correctly would generate cryptocurrencies for days at a time, however these dips and slides are often not accounted for in many business models as they take time and costs when being solved.

Speculative Markets - Reality

This market analysis conducted from roughly one month of system miner data on 22nd Janaury 2018. The above chart shows the system powered on for one month and the Ethereum tokens returned from the main network as PoW rewards. I took a median measurement from the market past 30 days as Ethereum was around £780.42 at market rates. After a month the total mined Ethereum value £1002.69 minus the electricity actual costs of £180. The electricity costs were measured at the wall. This makes a monthly return of £822.69. As cryptocurrency token markets are highly speculative this snapshot is as close an estimate as "predictions / calculators" can get, difficulty changes aside. As such after a year you would be looking at an approximated annual return of £9872.28, that is not including any additional "dual-mined" currencies. As a startup building on the blockchain it is useful to know that such an investment could potentially make a return in a year, minus maintenance/costs. Here is a summary and build photos!

Remember tokens are nothing but rewards from a proof-of-work system conducted on a peer2peer transaction, they will only ever be worth what someone is willing to pay for them.

Motherboard & components!

Here you can see the motherboard and an alternative PSU used for testing.

GPU's installation

The GPU's installed and ready for math.

6 GPU's installed & running

Testing systems is always fun, I cast a spell, say a prayer or incantation before first boot.

GPU racked sideways

This shot shows the side profile of the system.

12 GPU's waiting for work

Cypherpunk!

Blockchains are a boom to business

Whilst many are investing in the speculative token trade, well respected leaders of Cryptocurrency projects are calling for more applicational uses of blockchains. Kodak and other big brands are pivoting onto this new technology, social media markets are opening up and being explored through Facebook. Speculative token markets are part of the operational costs of blockchain businesses but should not be the focus. Instead the technical merits and applications of a globally distributed state machine that can run tasks and transfer trust is extraordinary. The potential for distributed apps, commerce in peer2peer, streaming video services, financial and government are huge and few currencies consolidate as much potential as Ethereum. I hope this post shares some lessons with others who embark on the path of building blockchains for purpose. Gold is still gold and will always be gold but blockchain technologies have the disruption potential to build "Web 3.0" and they just might do that.

"A blockchain is for building on, not investing in!"