Which excavator is the best? A comparison of mainstream mining machines, with Antminer leading the pack

Bitcoin continued to grow against the trend by 125% in 2016, and on June 6, 2017, it even exceeded RMB 20,000. Along with a large influx of capital and consumer-grade graphics card panic, it and other derivative currencies such as Litecoin and Ethereum have once again returned to the public's attention. The mining machines responsible for "producing/mining" these digital currencies also approached or even exceeded the threshold of 10TH/s in 2016 with the advanced 14/16nm process and stronger architecture.

What is the concept of 10TH/s computing power? Take the most powerful gaming card GTX 1080 Ti as an example. Its public version has a computing power of 1.88TH/s and a full-load power consumption of about 300W. The GTX 1060, which has the highest cost-effectiveness and the widest audience, has a public version computing power of "only" 0.67TH/s and a power consumption of about 130W. The former is a flagship product with a launch price of 5,700 yuan, and the normal price of the latter also starts at 1,500 yuan. Mining machines, a specially customized hardware, not only have violent computing power, but they are also monsters with a wall power of thousands of watts.

This time we got the more representative Bitmain Antminer S9 (13.5T computing power), Ebit E9+ (9T computing power) and AvalonMiner 741 (7.3T computing power) in the market for comparison. Now let's take a look at the strength of these "monsters".

Product parameters and information

• The developer of Antminer S9 is Bitmain, which is one of the earliest mining machine R&D companies and is also the world's largest "mining" company. Its businesses include cloud mining platform Hashnest, the world's largest mining pool "Antpool" (yes, it's bigger than Guochi and Yuchi), and other businesses.
  The Antminer S9 is the first mining machine to use a 16nm process chip. It was released in June 2016 and is equipped with the BM1387 chip independently developed by Bitmain. The single-chip computing power reaches 74GH/s. According to the difference in chip frequency, its S9 series includes products ranging from 9.5T to 14T, mainly in terms of core frequency and power consumption. It has been the mining machine with the highest return rate since its release, and it is also the mining machine with the largest computing power.

• Ebit E9+ is a product of Zhejiang Ebang Company. Its predecessor E9 was released in December 2016. It was the first to be equipped with 96 self-developed Bitcoin mining chips WD1227 (14nm process), with a total computing power of 6.3TH/S and an energy consumption ratio of 140W/T. In addition to increasing the number of chips and increasing the computing power to 9TH/s, this year's E9+ has added a dual-fan design and improved the appearance.

• The AvalonMiner741, which has the lowest power and computing power, is a product of the old mining machine manufacturer Avalon. Contrary to everyone's previous impression, in order to reduce product costs, Avalon is one of the few manufacturers that does not equip its top-level products with control circuits. Avalon 741 has the highest basic threshold among the three mining machines. It requires an additional controller or Raspberry Pi, and can only work with an AUC3 converter and AUC3 data cable.

Appearance and workmanship comparison: you get what you pay for

The Antminer S9 adopts a dual-fan three-computing board mining machine design. There are guide rails designed for stacking all around the body. It provides peripheral configurations such as a Mirco-SD card slot, IP address feedback button, RJ45 network cable interface, reset button and status indicator light. The structure is more complicated than the other two machines.

Bitmain equipped the S9 chip with a heat sink, which is a rare design in mining machines. The layout of the components, workmanship and production difficulty of the hashboard are the highest among the three machines. Each hashboard has 63 BM1387 chips developed by Bitmain, and a single chip has almost the same hashing power as the previous generation S7.

The opening design of the Ebit E9+ is more regular than that of the previous generation E9. The computing board has independent fixing screws, but both the shape and workmanship are very "mining machine", and the fan still retains the logo of the foundry. The computing board power supply layout and circuit board wiring are very regular and unified, but the heat sink and assembly are quite rough. There is also hand-torn glue pad paper between the heat sink and the circuit board. A lot of silicone grease is squeezed out on the heat sink, which is easy to accumulate dust. Not only is it difficult to take care of, but it will also affect the heat dissipation efficiency in serious cases.

The design of Avalon 741 is significantly different from common mining machines, because it uses an additional independent control circuit board, and there is no exposed wire on the outside. It is the only single-fan mining machine among the three. The two hash boards are separated on both sides of the machine body, and in the middle is a very high and large-angle heat sink, which is powered by 6 6-pin PCI-E interfaces.

Ebit E9+, Ant S9, Avalon 741

The three machines are all based on 12-14cm fans. The single-fan design of the Avalon 741 is short, but the width is still the largest. The Antminer S9 and the Wingbit E9+, which have the same power level, are very similar in appearance and size, except that the Antminer S9 is slightly smaller. The best workmanship is the Antminer S9, which has a violent stacking of materials for heat dissipation and a guide rail designed for stacking. The Avalon 741 has a simple design, while the Wingbit E9+ is relatively ordinary.

System and usage: each has its own characteristics and shortcomings

Although mining machines sound high-end, after several years of development, they are much easier to use than before. Except for Avalon, which requires the operation of Raspberry Pi as an intermediary, Ebit E9+ and Antminer S9 do not require additional control accessories. We only need to connect the 6-pin PCI-E interface on the power supply to the mining machine one by one. The three mining machines here, each computing board is powered by three 6-pin interfaces, while Ebit E9+ and Antminer S9 need to connect power to the control circuit board.

All three machines have DHCP (automatic IP address assignment) function and a mechanism that will fully operate only after being connected to the network cable. We power on and connect the network cable, search for the IP address of the mining machine in the router background, and directly open the corresponding address through the browser to enter the mining machine background.

Antminer S9 is manufactured by Bitmain, so its software maintenance is the most formal and complete. The official website and forums have a lot of instructions and videos, and the firmware is also complete. It is also the only machine among the three that retains the manual frequency control function. But the interface has not been translated into Chinese, which is a little regrettable.

The backend homepage of Ebit E9+ is very intuitive, with thoughtful designs such as preset mining pool addresses. The frequency and fan speed on E9+ are not adjustable, and the fan speed is as high as 6000 rpm, which directly leads to the noise problem mentioned below.

After Avalon manually flashes the firmware to the controller/Raspberry Pi, it still needs to be directly connected to the computer to enable the DHCP function, which is a high threshold. The home page displays comprehensive information, the UI is relatively beautiful, and the expert mode has a lot of detailed settings, but it is still not translated into Chinese.

Heat and noise

The test environment was kept consistent, using the same account to mine AntPool, using PPS+ mode, and the test location was a conference room with an area of ​​about 20 square meters, with a room temperature of 29-30 degrees. Bitmain's APW3+ power supply (1600W, with a maximum conversion efficiency of 93.5%) was used for power supply. It is characterized by solid materials, and each 6-pin interface can output 12V 13.3A, which can directly meet the current violent mining machines of 1600W and below, without considering the cross-interface problem of dual power supply. Temperature and noise detection was performed after the mining machine ran for 3 hours.

At 10cm and 100cm away from the air outlet of Antminer S9, the former is around 82-83 decibels, but at 1 meter away, the noise drops to around 71.9 decibels. The actual volume attenuation is faster than that of Yibit E9+, and the listening experience is not that exaggerated. There will not be too obvious sound after closing the door, and it is very pleasingly stuck in the volume range acceptable to civilian use.

The highest temperature of the Antminer S9 body appears at the power board connection, which is about 48.4 degrees; the air inlet is only 35 degrees, and the air outlet is 59.7 degrees. The heat diffusion range of the air outlet is larger than that of the Ebit E9+. Considering that their power is around 1600W, it seems that the fan design with a speed difference between the front and rear does have a certain effect.

At 10cm, 50cm and 100cm away from the air outlet of Yibit E9+, the noise reached 83.4, 77.8 and 75.4 decibels respectively. The power of the dual 6000 rpm fans is indeed not to be underestimated. The sound is similar to that of a vacuum cleaner, and the volume is very loud. It is the noisiest among the three machines. Although the decibel value seems to be not much different from that of the Antminer S9, the human ear's hearing experience is a whole level higher. People who see it for the first time will definitely be shocked by the volume. It is no longer suitable for use in residential or office settings.

The hottest point of the Ebit E9+ is below the fan interface, at about 43.7 degrees; the hottest point of the top circuit board is 42.7 degrees; the highest temperature of the air inlet is 41.2 degrees, and the air outlet is 57.5 degrees. Except that the side of the body is a little hotter than the Antminer S9, the heat dissipation performance is good.

The noise levels were 76.5, 68.1 and 62.2 decibels at 10, 50 and 100 cm from the air outlet of Avalon 741. The fan speed was around 4500 rpm in working state, and the noise was well controlled.

The hottest point of the Avalon 741 also appears below the fan interface, which is about 39.2 degrees, and the rest of the side of the body is below 37 degrees; the highest temperature at the air inlet is 48.1 degrees, which is higher than the other two mining machines; the air outlet is about 50.6 degrees. After the power is reduced, the overall heat control pressure is much less.

Comparison of computing power and power consumption: Antminer S9 is the best

Antminer S9 computing power curve every 10 minutes

Antminer S9 hourly computing power curve

The official hashrate of the Antminer S9-13.5T version is 13.5TH/s. In the actual 10-hour test, the first 4 hours were set to the lowest frequency of 550M, with a theoretical hashrate of 11.85TH/s and an actual average of around 12.1TH/s. The following 6 hours were set to the default frequency of 627M, with a theoretical hashrate of 13.5TH/s, but the actual maximum value exceeded 14.11TH/s, and the average was above 13.8TH/s, both of which were higher than the nominal value.

The standby power consumption of Antminer S9 when not connected to the Internet is around 90W, the peak power reaches 1625W, and the power is around 1583W when operating smoothly.

Yibit E9+ computing power curve every 10 minutes

Yibit E9+ hourly computing power curve

The frequency of Ebit E9+ cannot be set. The computing power fluctuates in a 40-minute cycle. After 1 hour, it begins to stabilize at around 9TH/s. The average computing power in 8 hours is around 8.9TH/s, which is slightly lower than the official value of 9TH/s, but the deviation is not large.

The standby power consumption of Ebit E9+ without Internet connection is about 170W, the peak power reaches 1651W, and the power is about 1606W after stable operation, with little fluctuation. What is surprising is that its official power is lower than that of Antminer S9, but the actual power consumption is about 20W higher.

Avalon 741 hashrate curve every 10 minutes

Avalon 741 hourly computing power curve

The actual 10-hour average hashrate of Avalon 741 is around 6.8 TH/s, which is quite different from the official 7.3 TH/s. There are obvious fluctuations in minute statistics, and the overall performance is a bit weak compared to the previous two machines. The actual standby power consumption is around 100W, the peak power reaches 1298W, and the power is around 1291W after stable operation.

The actual and nominal hashrates of the Antminer S9 and Ebit E9+ are close, while the performance of the Avalon 741 is not so good. The difference in hashrate is very obvious. The hashrate of the Antminer S9 is 1.55 times that of the Ebit E9+ and 2 times that of the Avalon 741, which is almost twice as much as the other two. This difference in hashrate performance will lead to a huge difference in revenue when the price of Bitcoin is high.

Although the price of Bitcoin has fallen before the deadline, it is still as high as 17,890 yuan per Bitcoin. According to the 103.41% profit ratio of Bitmain Antmine's PPS+ model, the daily profit of 1TH/s computing power on July 1 was 0.00036689 BTC. The profit of the three mining machines is very exaggerated after being converted into RMB.

Summarize

In terms of appearance and maintenance, the Antminer S9 is a very authentic enterprise-level product, with solid and durable materials and workmanship; the Avalon 741 is relatively simple and more like an electronic consumer product; although the circuit board of the Ebit E9+ is well made, its assembly style is very "mining", and it still remains at the level of the previous generation of mining machines;

In terms of usage and system, Antminer S9 is the simplest and has the most complete information. Although the backend is not translated into Chinese, the official forum has complete maintenance information and the firmware is updated in a timely manner. Ebit E9+ has a good backend homepage, which is very convenient for observing the status of the machine, but the official maintenance information is lacking. The main problem of Avalon 741 is that it requires other accessories to work, the threshold for use is high, and the cost savings are not reflected in the price.

Calculated based on the price of 0.6 yuan/kWh*

In terms of the core computing power and power consumption, only the Antminer S9 can slightly exceed its nominal value, while the Ebit E9+ and Avalon 741 are both lower than the nominal value, and the latter is still a certain distance away from the nominal value. Surprisingly, the Ebit with a lower nominal value is not far behind the Antminer S9 in terms of total power. The computing power efficiency of the Antminer S9 is 8.72G per watt, which is far higher than the Ebit E9+ and Avalon 741, both of which failed to cross the threshold of about 5.4G/W.

The computing power advantage of Antminer S9 is really too great. With a performance gap of 1.55 and 2 times, it can easily make up for the investment in computing power costs. Even after a small drop in the Bitcoin quotation, it can pay back more than 2 months earlier than the other two mining machines. It is still the most efficient and fastest-returning mining machine, without a doubt.

The power and output per watt of the Ebit E9+ and Avalon 741 decrease in turn. The main problem of the Ebit E9+ is the brutal noise, while the core computing power of the Avalon 741 is relatively weak. Energy efficiency and payback speed are the same as workmanship. You really get what you pay for.