Former Technical Director of Bitmain Reveals how S9 Became the Dominant Crypto Miner
Bitmain’s team in Chengdu realized the cross-generation advantage to its competitors through dynamic triggers, which is also the main reason for the success of S9 miner. The design of S9 is a miracle of bitcoin miner, which has become the most productive bitcoin miner with a life cycle of more than three years.
Xie Dan, former technical director of Bitmain joined Bitmain with a small team of two people and became the full customization department of Bitmain in October, 2014. The initial direction is the domino logic in the email.
Domino logic is a relatively mature dynamic logic structure, its main form is as follows:
Compared with the corresponding static circuit, domino logic circuit has the following advantages and disadvantages:
1. Because the signal only needs to drive N-tube, the capacitance is small and the speed is fast.
2. Since point A is a dynamic point, the final reverser M3 / M4 is necessary.
3. Point A has the problem of charge sharing.
4. M1, M2, m3, M4 are new devices, and PMOS are reduced devices.
5. Because M1 and M2 are clock-driven, their power consumption is at least twice that of other ordinary signals.
Therefore, in terms of power consumption, since M1 and M2 are clock signals, their power consumption is at least calculated according to twice the power consumption of signals, then the power consumption of M1 + M2 + m3 + M4 increases the power consumption of six MOS tubes. For general logic, there are not many standard cells with more than 12 MOS tubes. In terms of area, M2 + M4 increases the area of two NMOS tubes, while PMOS is less, but it is difficult to reflect the saved area in the layout. That is to say: although domino logic will speed up the chip speed, the area and power consumption are slightly increased.
When Xie’s team submitted the conclusion in January 2015, we basically confirmed that domino logic is not suitable for mining chips. Because of the pure parallel operation, the speed of bitcoin mining machine chip is not so important. The biggest cost of the miner lies in the electricity charge, so the power consumption is the most important. When we evaluate the rule of thumb of area and power consumption, the weight of power consumption is more than three times of area and speed.
The failure of our domino logic attempt didn’t hit our exploration, because we found a standard unit suitable for dynamic logic with many products and pipes, and this standard unit has a very obvious instance, flip-flop. So we go back to the beginning and take the dynamic trigger as the key target.
Considering that the area and power consumption of the new dynamic flip-flop are only half of the original, the new pipeline has doubled the speed under the same area and power consumption. Of their bitcoin mining machine chip, they have changed from a 32 level pipeline to a 64 level pipeline, doubling their computing power.
In 2015, Xie’s team completed the design of BM1385 chip at 28nm (Ant miner S7), and completed the design of BM1387 chip at 16nm (Ant miner S9) at the end of 2015. In terms of performance, their 28nm chip has almost the same performance as their competitors’ 16nm chip, and our 16nm chip is half the cost of our competitors. Through the dynamic trigger, they realize the cross-generation advantage to our competitors.