Bitcoin “mining”: an energy consumption black hole!

People who hope to get a piece of the Bitcoin boom are flocking to mining factories to join the craze of mining. At the same time, the astonishing power consumption of Bitcoin mining has attracted attention.

According to a study by PowerCompare.co.uk, an energy price comparison service, more than 29.05 terawatt-hours (1 terawatt-hour is 100 million kilowatt-hours) of electricity was used for Bitcoin mining in 2017, exceeding the annual average electricity consumption of 159 countries in the world. The entire country of Ireland (with a population of 4.77 million) only consumes 25 terawatt-hours of electricity a year.

Recently, there were reports that some residents in a community in Kunming, Yunnan were mining Bitcoin, which caused a large-scale power outage in the community, the transformer was burned, and more than 400 meters of cables were damaged.

Why does mining Bitcoin consume so much electricity?

Integrated circuit power consumption is the root cause

"A Bitcoin mining machine is essentially a special computer," said Cao Junwei, deputy director of the Institute of Information Technology at Tsinghua University, in an interview with Science and Technology Daily.

How does this special computing device mine Bitcoin?

In the world of Bitcoin, a data block is recorded approximately every 10 minutes. All mining computers are trying to package this data block while completing the submission task, and those who successfully generate this data block can get a Bitcoin reward. The stronger the computing power of the mining machine, the faster the "mining" speed, and the higher the probability of mining Bitcoin.

Now Bitcoin "mining" is moving from "miners" fighting alone to "team mining", that is, "mining" through "mining pools". If your mining machine's computing power accounts for one-fifth of the entire "mining pool"'s computing power, if the "mining pool" can mine 5 bitcoins a day, then you can get 1 bitcoin.

Behind the race against time for mining, electricity is an important cost that cannot be ignored. A "miner" did some calculations: if the mining machine runs 24 hours a day, 0.0018 bitcoins can be mined, and one bitcoin can be mined in 556 days. His mining machine has a power of 1350 watts, and a mining machine with a power of 1000 watts consumes about one kilowatt-hour of electricity per hour. Two mining machines consume 32.4 kilowatt-hours of electricity per day, and the daily electricity bill is 16.8 yuan. In other words, it costs 9367 yuan to mine one bitcoin.

"Fundamentally speaking, this is a problem of power consumption of integrated circuits." Cao Junwei told Science and Technology Daily that from a hardware perspective, the circuit itself consumes electrical energy, and Bitcoin mining machines have relatively high requirements for computing power. When large-scale "mining" is carried out, the mining machines are relatively concentrated, so the power consumption is naturally large when added together.

Mining machines aren’t the only ones sucking power

In fact, in addition to Bitcoin mining, news about the astonishing power consumption of computing equipment is very common.

For example, supercomputers. According to reports, the annual electricity bill of the Tianhe-2 supercomputer is close to 100 million yuan, which is equivalent to the annual electricity consumption of a small or medium-sized city. The annual electricity consumption of the Sunway TaihuLight supercomputer is about 15 megawatt-hours, which is equivalent to the electricity consumption of three Tsinghua Universities. The power consumption is still amazing.

Another example is data centers. Lv Tianwen, secretary general of the China Data Center Energy Conservation Technical Committee, said in an interview with Science and Technology Daily that in 2016, the total power consumption of China's data centers exceeded 110 billion kWh, exceeding the total power generation of the Three Gorges Dam in 2016, and increased by 11.7% compared with 2015.

"The power consumption of a data center includes four parts: IT equipment, power supply and distribution system, fresh air cooling system and others." Lu Tianwen told Science and Technology Daily that IT equipment power consumption accounts for about 50% of the total power consumption, the fresh air cooling system accounts for about 35%-40%, and the power supply and distribution system accounts for about 15%-18%.

Lv Tianwen introduced that PUE (Power Usage Effectiveness) is an important indicator for measuring the energy efficiency of data centers. It is the value obtained by dividing the total energy consumption of data center equipment by the energy consumption of IT equipment. At present, the average PUE level of domestic data centers is above 2, which means that the power consumption of non-IT equipment exceeds the power consumption of IT equipment itself.

"In data centers, a large amount of energy consumption occurs in non-IT equipment, and a lot of electricity is wasted." Lv Tianwen told Science and Technology Daily that the power supply and distribution system and the refrigeration system have great potential for energy saving. For example, the fresh air refrigeration system accounts for 35%-40% of the total energy consumption. After optimization and energy-saving transformation, 10%-20% of energy consumption can be saved. From a national perspective, 10 billion to 20 billion yuan in electricity bills can be saved.

Breaking the limits of Moore's Law is the key

However, Lu Tianwen admitted that in data centers, energy saving of IT equipment is the most fundamental, because IT equipment accounts for the majority of energy consumption, and the power supply and distribution system and fresh air cooling system all serve IT equipment.

Lv Tianwen did some calculations for the Science and Technology Daily reporter. If the power of IT equipment is 1,000 watts, then it needs to be equipped with a 1,000-watt power supply and distribution system and a 1,000-watt fresh air cooling system, with a total power consumption of 3,000 watt-hours per hour. If the IT equipment reduces its energy consumption to 500 watts, it needs to be equipped with a 500-watt power supply and distribution system and a 500-watt fresh air cooling system, with a total power consumption of 1,500 watt-hours per hour.

This means that once the energy consumption of IT equipment is reduced, the energy consumption of the entire data center will also drop significantly.

"But nowadays, most of the energy-saving renovation of data centers focuses on non-IT equipment such as power supply and distribution systems and refrigeration systems, which is actually a paradox." Lv Tianwen told Science and Technology Daily that once a data center is put into use, it is difficult to renovate IT equipment because IT equipment downtime will cause huge losses. Therefore, it is usually encouraged that data centers use better-performing and more energy-efficient IT equipment at the beginning of design and construction.

So how does the energy consumption of IT equipment come about? Lv Tianwen explained that the major energy consumers of IT equipment in data centers are network storage devices and servers, of which network storage accounts for about 60%. In the past, mechanical hard disks consumed a lot of energy, but now data centers are increasingly using network storage technologies such as SSD solid-state drives, which can significantly reduce energy consumption in this area.

"Whether it is Bitcoin 'mining', a data center, or a supercomputer, the power consumption problem is essentially related to the bottom-level integrated circuit." In Cao Junwei's view, with the development of chip technology, there are more and more devices on a unit area of ​​integrated circuits, and the functions are becoming more and more complex. At the same time, the power consumption of integrated circuits per unit area is also getting higher and higher.

Cao Junwei introduced that in addition, from the software level, its power consumption is related to the design of the algorithm and the efficiency of the software. From the system level, it depends on what network structure the cluster built by many IT equipment uses and whether it is efficient. From the environmental factor point of view, it involves heat dissipation, refrigeration and other systems.

As for how to fundamentally make these computing devices no longer so power-consuming, in the long run, it may be necessary to use quantum computing and new semiconductor materials to break through the limitations of Moore's Law and make computing devices more efficient and energy-efficient.