CryptoSupers Smash Moore’s Law
A Hundred Exahashes
Hashes in crypto mining are the metric that parallels flop in the supercomputing world.
Crypto miners compete via Proof of Work consensus algorithms in order to win a block reward and commit a group of transactions to the blockchain. Cryptocurrency mining via Proof of Work continues to represent the most effective class of consensus algorithm to maximize security in a decentralized manner and to allow coins to accrue significant value.
In our CryptoSuper 500 race, we only look at the top mined coins that use Proof of Work, since these are the only supercomputer class workloads in the cryptocurrency world. Other consensus algorithms are much less costly, but as the marketplace has consistently demonstrated, impart much less value and security to a cryptocurrency. This is a trade-off between store of value and utility attributes.
Our third CryptoSuper 500 list is scheduled for release on November 20, 2019, at OrionX.net.
Crypto mining entered the Exascale era in 2016; three years ago, the global hash rate for bitcoin was already exceeding an ExaHash per second. As of this writing, Bitcoin’s total computational power is nearing 100 Exahashes per second; that is a billion billion hashes each second. (As of today the seven-day average is 95 Exahashes/sec.) Crypto hashes are very simple calculations, using SHA(256) in the case of Bitcoin hashing, with many repeated trials required until a winning result adhering to the pre-defined problem difficulty is achieved.
The problem difficulty is regularly adjusted by the consensus algorithm as the collective hashrate increases or decreases. Bitcoin has had 300 such adjustments to data, on an approximate two-week schedule (each 2016 blocks).
Table 1 shows the Bitcoin hashrate history at one-half block year intervals, starting from when Bitcoin was two blocks years of age. Each block year is 52,500 blocks of approximately 10 minutes’ duration, and block years are running slightly shorter than regular calendar years, recently by about two weeks. We are now in the second half of the 12th block year since Bitcoin’s blockchain launched in January 2009. Data was obtained from blockchain.com using 7-day smoothing. The hashrate is shown in Terahashes per second, a thousand Terahashes is a Petahash, and a million Terahashes is one Exahash.
In the table, the block height (number of blocks) is shown in the first column, then the block reward is shown in the following column. Note that it decreases by a factor of two each four block years; these are the key Halving events that drive inflation down toward zero. The calendar date is shown, then in the next column the total hashrate for all miners around the globe. Then we have the base 10 logarithm for the hashrate and the log of the block years elapsed. In the last column of the table, one sees the 3-year prior interval slope for a log — log regression (power law relationship) between hashrate and block years.
We examined both an exponential relationship: HR = c * exp (a* Byr) where HR is the hash rate, Byr is the elapsed number of block years, and a and c are parameters, and a power law relationship: log HR = c + a * log Byr, where a is the power law index.
The exponential relationship flattens out in recent years, but the power law holds up well across over eight orders of magnitude during the past nine years. Hashrate has been growing extremely rapidly, roughly as the 12th power of elapsed block time! For a while in the 2014–15 period hashrate was growing with a power law > 15 as a rapid switch from GPUs to ASICs was underway. Now it has settled down somewhat to only being much, much faster than Moore’s law.
The price of Bitcoin, on the other hand, has been growing roughly as the fifth power of elapsed block time, driven by enhanced security and scarcity as Bitcoin’s inflation rate drops continually, especially at each Halving. Hashrate follows price as miners are incentivized to increase their capital investment in terms of the number of systems devoted to mining and by obtaining the latest and greatest hardware. The result is that the hashrate is increasing at greater than the square of the Bitcoin price.
Doubling almost Twice a Year
In the last three calendar years, as indicated in Table 1, hashrate increased 37 times for Bitcoin, for a compound (calendar year) annual growth rate of 233%. This implies a doubling each seven months, or over three times more rapidly than the Moore’s law rate.
Hashrate follows price, price reflects security (that depends on length of blockchain and hashrate as well) and scarcity (stock-to-flow, or inverse of the inflation rate) which is a simple function of blockchain length and more than doubles with each Halving at four block year intervals.
Mining inputs include capital equipment expense, facilities expense, personnel, electricity, cooling, as well as the expected price of bitcoin or other cryptocurrencies, over the capital equipment life; various studies suggest a typical 50/50 capex/opex split.
Since the major equipment cycle is around two years, the miners must look out a couple of years ahead and guesstimate costs and future bitcoin prices to justify new capital investment. If price drops they can shut down their least efficient (e.g. older) equipment and only mine with their newer equipment that has the best hashrate per kilowatt characteristics. There may also be seasonal effects, especially since a lot of the miners are located in regions with cheap hydroelectric power that costs less when more water is flowing. If some miners shut in some equipment, the more efficient miners will increase their shares of bitcoin rewards, as the mining difficulty will decrease. Somebody gets the rewards.
Capex is sunk cost, so miners will continue to mine in the face of dropping prices as long as they can have positive cash flow after operating expenses (they may have to defer equipment upgrades, though, if cash is tight). They also can hedge their expected future production through forward sales with Bitcoin options or futures.
This rapidly growing global hashrate is a result of strong advances in the specialized rack-optimized crypto mining ASIC-based computer designs, in performance, packaging, and efficiency, and also in the optimization efforts of mining pools as they seek out the least costly locations for electrical power and cooling and optimize their facilities design. It is of course additionally a result of the substantial long-term increase of Bitcoin price, some four orders of magnitude since early 2011.
Cryptocurrency topics: orionx.net/blog