Tag Archives: cryptocurrency mining

Crypto Supercomputers: First Aggregated Ranking

Working with OrionX, we have just published the first aggregated list of cryptocurrency supercomputer mining pools, ranked by the economic value generated.

I have recorded a podcast about this list with Rich Brueckner, President, InsideHPC. You can listen here: https://insidehpc.com/2018/11/announcing-new-cryptosuper500-list/

A related slide presentation with a complete set of tables is available here: https://www.slideshare.net/mobile/insideHPC/announcing-the-new-cryptosuper500-list

The list is inspired by the Top500 supercomputer list that is released twice a year at the major supercomputer trade shows and conferences held each June in Germany (ISC) and each November in the US (SC).

That list is based on the performance of Linpack, a floating point intensive benchmark that solves a very large system of linear equations.

Supercomputers are based in a single location. They are very large clusters of general purpose CPU-based nodes, often augmented with GPUs, and frequently employing specialized interconnects.

Cryptocurrency mining is embarrassingly parallel. Many nodes can be racing simultaneously to solve the same cryptographic puzzle for the block reward. Mining pools may be centralized, but more likely they are decentralized to various degrees. Mining pools often have many contributors located in many countries, so even the concept of a host nation associated with the pool is fuzzy.

And the hardware employed is typically specialized ASICs or FPGAs, as well as the GPUs frequently found in traditional supercomputing simulation of science and engineering problems.

With mined cryptocurrencies, we must take a different approach and look at economic value.

For this initial list we looked at the top dozen cryptocurrencies by money supply, which is usually called market cap, and that is simply the number of coins created by a certain date, and the coin price on that date.

Of the top dozen, just half of those or 6 coins, are mined: Bitcoin, Ethereum, Litecoin, Bitcoin Cash, Monero, and Dash. Other coins are generated by premining, airdrops, or consensus algorithms that avoid mining. As a result they are centralized to varying degrees and presumably less secure.

We chose October 30, 2018 to gather prices, supply, block production, and other statistics. This was prior to the Bitcoin Cash fork into two coins, so only the initial coin is considered for the first list.

Among mined coins, a range of mining consensus algorithms are used. Differing cryptographic hashing protocols may be used. Time windows and block rewards vary. Hashing rates have a tremendous range across the set of coins, from MHash/s with Monero to ExaHash/s with Bitcoin.

Thus we cannot compare across coins based on hashing rates and block rewards per se. Instead we look at economic value. For a given coin, one can rank order by blocks produced.

We ask what is the daily value of a certain coin produced by a given mining pool? How many coins at what price? We took daily averages for the prior week, and where we had better data, for the higher value coins, we used the prior month average daily rate instead. We then extrapolated the annualized value based on the average daily rate.

We compiled statistics for the 30 largest pools on a per coin basis. We also aggregated results for pool operators that produced more than one type of coin.

The first table is a table of average daily and estimated annualized production in millions of USD for the top coins. (With the very recent price slump following the Bitcoin Cash fork, the numbers would now be lower by about 1/4 if prices do not recover for a while). About $4 billion of Bitcoin is mined (minted) per year, and around $1 billion of Ethereum. Litecoin, Bitcoin Cash, and Monero collectively contribute around  $400 million (Dash did not make the cut).

Table 1: Top 5 Mined Coins

Coin

# Top Pools

Daily M$

Annualized M$

Bitcoin

17

11.31

4,129

Ethereum

5

2.77

1,010

Litecoin

5

0.64

234

Bitcoin Cash

2

0.38

140

Monero

1

0.10

37

Totals

30

15.21

5,550

Next is a table of the top half dozen pool operators, combining different coin types if they are mining more than one of the top coins. Three are in China, one in Hong Kong, and two in the U.S.

Table 2: Top Pool Operators (aggregated across top coins)

Top 6 Operators (across coins)

# Top Pools

Daily M$

Annualized M$

Country

BTC.com

1

1.901 694

China

Antpool

2

1.747

638

Hong Kong

F2Pool

3

1.585

579

China

ViaBTC

2

1.329 485

USA

BTC.Top

2

1.222

446

China

Slushpool

1

1.215

444

USA

Total

11

9.00

3,285

Bitcoin has its own decentralized, open source, version of a central bank and a clearing house system embedded in the Nakamoto consensus. Bitcoin is presently an emerging economy with over $1 trillion in annual transactions (GDP, gross decentralized product), supported by a very economical and efficient seigniorage of about $4 billion in mining block rewards, or less than 0.4%.

The indicated inflation rate at present is about 4% in supply, but in about 18 months the block reward will have its third halving. This will decrease the block reward to 6.25 Bitcoin from its current 12.5 coins. The inflation rate will drop below 2%.

This is not like your Federal Reserve that issues forecasts and goals. Recently the Fed has been pushing to increase inflation to 2%, and happy that they achieved the increase.

With Bitcoin this decrease in inflation will definitely happen, come hell or high water; it’s math, it’s baked in to the Nakamoto consensus. Relative to the US dollar and fiat currencies in general, Bitcoin will be disinflationary going forward.

The next list will be announced in June, 2019, and we can begin tracking developments in the cryptocurrency space over time.

Will Bitcoin Consume All Electricity?

The idea that bitcoin will consume an enormous fraction of the world’s electricity is hysteria.

In a recent article in the Communications of the Association for Computing Machinery, June 2018 issue, Nicholas Weaver (a lecturer in computer science at UC Berkeley) raised this issue, in what was otherwise a good article on the security issues around bitcoin.

Weaver quotes a statistic that cryptomining consumes more electricity than Ireland. This may be based on digiconomist.net, which runs toward the high end. Other estimates are only half as large.

He states “If there is profit in mining, the miners will keep using more and more power until there is no more excess profit available”.

This is true, but he overstates things. He evidences a lack of basic understanding of economics and how businesses operate, ignoring all the complexities that go into cryptomining.

Mining costs are a combination of fixed, and variable costs. The variable cost is primarily the electricity consumed. The fixed costs consist of facilities costs, equipment costs, and people and administrative costs. Equipment costs can run over 1/4 of the total.

Total global Hash rate over the past 12 months has grown from 5 to 38 Exahashes, a factor of 7.5.

Difficulty in the Nakamoto consensus protocol has grown by a factor of 7.

Revenue per Terahash per day grew from $1 to $3+ at the peak half a year ago and with the price collapse is down to $0.30. That is before electricity.

According to cryptocompare.com, with the current BTC price of $6500 and at $.10 per kWh for electricity the profit is just $0.06 per Terahash-day currently, but that is before any of the fixed costs are recovered.

If you are not covering your fixed costs plus variable costs you will not stay in business to consume electricity.

Here’s where Weaver really gets it wrong. He states “a 10x reduction in power consumption per hash for Bitcoin mining would have little real effect on Bitcoin’s power consumption. Instead, there would just be 10x as many hash computations needed to produce a block.”

Difficulty rates depend on the total cost burden.

His statement above completely ignores fixed costs. Whether it is an individual mining rig or a huge mining farm, the fixed costs of location, equipment and labor will generally be of order half the total cost.

Do-it-yourself miners in Mom’s basement or my friend Dan might ignore their location costs and equipment burden on their cooling and they might give away their labor for free. But their rigs aren’t as efficiently operated and their electric costs may be higher. They still have to amortize their equipment costs, at least for added ASICs and GPUs.

Suppose the gross revenue is $0.30 per THash-day and the fixed costs can be held to $0.10 and the electricity cost is $0.2. This is a breakeven business example with a large electricity burden.

Now reduce the power consumption per hash by 10x, in which case the total costs drop from $0.30 to $0.12. There would be incentive to increase total hash power by up to 2.5x not 10x. A factor of 4 overestimate.

In practice, it takes time to ramp up hash power. Supplies of equipment are tight. Data center spaces are limited. System administrators are not always available. There are both practical and regulatory restrictions on power available to mining farms. 

Furthermore, ASICs and GPUs for Bitcoin and cryptocurrency mining are in particularly tight supply. As demand goes up, there is a bidding war with equipment going for premium prices. This drives up the fixed cost component of Bitcoin mining.

Doubling capacity takes many months, and is subject to financial planning scenarios about future crypto prices, equipment delivery time lags, and electricity prices and availability.

According to digiconomist.net on July 5th, Bitcoin is just 1/3 of 1% of global energy usage (1 part in 300). Global GDP is some $80 Trillion and annual transaction flows of Bitcoin are over $1 trillion. So for over 1% of the proportional GDP the related energy requirement is proportionally 3 times lower.

According to an article in ZeroHedge, gold mining is much more energy costly. Per $ of value produced bitcoin and gold are roughly comparable, but there is a lot more gold mined.

https://www.zerohedge.com/news/2017-11-04/bitcoin-vs-gold-which-ones-bubble-how-much-energy-do-they-really-consume

award bars blur business

Photo by Michael Steinberg on Pexels.com

They state that per bitcoin the energy consumption is 6.6 million barrels of oil equivalent per year while the consumption for gold mining is 123 million barrels per year.

There are about 88 million ounces of gold produced per year, with a value of around $109 billion, versus 2/3 of a million Bitcoins, value around $4.3 billion. That’s a factor of 25 in value since bitcoin is 5 times more valuable comparing one coin to one ounce.

It seems that the total energy consumed in gold mining globally is around 19 times that of Bitcoin mining. And the number of bitcoins produced per year is dropping due to the halving every 4 years coded into the Nakamoto consensus.

The whole concept is designed to shift miners’ revenue toward transaction fees as the economy develops over time.

If you want to save the environment, focus on gold mining energy efficiency. Improve it by 5% and you can cover the entire Bitcoin mining energy budget.

For a variety of reasons, other cryptocurrencies are less energy intensive than bitcoin. They are also less secure, less battle hardened.

Bitcoin is a digital gold alternative that has the advantages of very low cost portability, and lower costs to secure and store.

It is a valid alternative to gold ownership as a store of value, and is a greener solution. There is a great deal of work (pun intended) on alternatives to Proof of Work mining, including Proof of Stake protocols and delegated Byzantine Fault Tolerant protocols. Also the growth of second layer solutions such as Lightning will support a larger economy and shift miners’ revenue more toward transaction fees.