Canaan Creative is a world leader, but domestically produced CPUs still lag behind. What is the difference between dedicated chips and general-purpose chips?

With annual sales exceeding 1.3 billion yuan, profits of 470 million yuan, plans to go public in Hong Kong, and the successful mass production of the world's first 7nm chip, Canaan Creative, an integrated circuit chip design company, has become extremely popular.

As for domestic CPU chips, even with the country's key support, the CPUs produced by Zhaoxin, Hygon and other companies can only run office software, and they have difficulty running more complex games, not to mention the recently popular "Chicken Eating" which has high configuration requirements.

Why is it that the domestically produced CPU chips, which have been supported by the state for nearly 20 years, have remained unknown and unable to catch up with the world's leading level, while a company selling mining machines has achieved an annual profit of 470 million in just a few years and has achieved the feat of being the world's first to launch the most advanced 7nm chip?

This has to start with the "Pumpkin Zhang" of that year.

Studying FPGA, becoming "Pumpkin Zhang", and establishing Canaan Creative

Back in 2011, Zhang Nangeng was a graduate student studying computer science at the Beijing University of Aeronautics and Astronautics, majoring in integrated circuit design. Like many engineering students, his life was boring. He often used anime to kill time, watching more than 500 anime a year.

By chance, he learned about Bitcoin. After he knew the operating principle of Bitcoin, he began to think about developing machines specifically for Bitcoin mining. At this time, Bitcoin has entered the GPU mining era from the personal CPU mining era, and has begun to enter the field of field programmable gate arrays (FPGA) era. Scattered graphics card assembly arrays cannot meet the "mining" needs of miners. Computer equipment seems to have been used to the limit. At this time, it is necessary to produce "customized equipment" specifically for Bitcoin miners to mine. This happens to be the professional field of Zhang Nangeng, an engineering student studying integrated circuit design, so he began to study FPGA mining machines in his spare time, and participated in community discussions in the Bitcoin community with the ID ngzhang, so people in the community nicknamed him "Pumpkin Zhang" based on the pinyin.

After doing some research on FPGA mining machines, Zhang Nangeng began to customize FPGA mining machines for foreigners. At that time, he thought it was just a simple "work-study program".

However, this did not last long, as it was broken by a piece of foreign news. In June 2012, an American agency claimed that they were going to develop a powerful special mining machine called ASIC, which would have a computing power far beyond the current level for the algorithm used in Bitcoin. Such a computing power monopoly is likely to lead to a 51% attack, which would destroy the entire Bitcoin world.

Zhang Nangeng could no longer sit still. As a Bitcoin enthusiast, believer, and beneficiary, in order to "maintain world peace" and "work and earn tuition fees," he decided to develop an ASIC mining machine himself and sell it to miners to fight against the possible "computing power monopoly" by American institutions.

He named the ASIC mining machine he designed and developed "Avalon", which is the name of the "Island of God" in King Arthur's legend.

The first batch of ASIC Pumpkin machines cost $1,299 and went on sale in September 2012, with only 300 units sold.

The second batch of ASIC Pumpkin machines were priced at 75 BTC and went on sale in February 2013, with 600 units sold.

The third batch of ASIC Pumpkin machines were priced at 75 BTC and went on sale in March 2013, with 600 units sold.

These three batches of ASIC mining machines were sold out as soon as they were released, because within a few days of their release, the ASIC Pumpkin machine could generate hundreds of bitcoins a day, and the mining income could reach up to 200,000 yuan a day. Therefore, the mining machines he produced were quickly sold for 250,000 yuan each, and many people stayed up all night for several days before and after the opening of the reservation day to buy them.

The market was so hot that Zhang Nangeng was of course fearless and wrote down the arrogant "tyrant clause":

1. No refunds will be given under any circumstances. This is because we will use the advance payment to purchase parts and therefore cannot refund.

2. No guarantee. This includes but is not limited to: the risk of no third-party evaluation, the risk that we are scammers, the risk of not being able to ship on time due to objective and supervisory reasons, the risk that bitcoin itself has serious problems and cannot continue to exist, etc.

3. There is no sales customer service. Due to limited manpower, if you cannot understand the entire content of this announcement, please do not purchase. At the same time, QQ and Wangwang do not provide reservation consultation, and all inquiries will not be responded to.

4. Once the delivery address is confirmed, it cannot be changed.

The name "Pumpkin Zhang" spread throughout the circle in a short time. He and his partners who developed ASIC mining machines together registered and established the company Canaan Creative, and the Bitcoin mining industry officially entered the ASIC era.

Although the mining industry experienced the rise and loss of contact with Baked Cat, two mining accidents, and changes in national policies in the following years, Canaan Creative continued to grow and develop. From 2013 to 2017, Canaan Creative successfully developed and mass-produced 110nm, 55nm, 28nm, and 16nm chips. In these four years, through 10 capital increases, the company's registered capital was changed to 300 million, a 3,000-fold increase.

The "Pumpkin Zhang" of that year has now become one of the major shareholders of a company valued at billions, and the company also began seeking a listing in 2017.

On August 8 of this year, Canaan Creative officially released the "world's first 7nm mass-produced chip" at a press conference held in Hangzhou, representing the highest level in this field.

Looking back at the entire process of Canaan Creative, a graduate student studying integrated circuit design, while "working part-time to reduce study", promoted the development of an industry, and the company he founded became the second largest company in the industry, and launched the world's first 7nm chip representing the most advanced technology. It has to be said that it is a legend. As the saying goes, heroes are made by the times. In the digital currency boom, "Pumpkin Zhang" really took off and, through hard work, flew to the world's leading position.

The success of "Pumpkin Zhang" is not only the result of the young genius's personal efforts, but also accompanied by luck, persistence, faith and other factors. In the early stage of the entire industry, he became famous with his own technical genius, and continued to expand the team during the development of the industry. The technical team grew from a few people to nearly 100 people, and finally became an internationally leading technology and the second strongest company in the industry. This whole process is exactly the process of the development and prosperity of the digital currency mining industry.

The difficulties of domestically produced CPUs

During the years when Canaan Creative has been moving towards glory, the independent domestic CPU chips that have been developed for more than ten years are still far behind the world. Some time ago, the ZTE incident once again pushed "domestic chips" to the forefront of public opinion, and Liang Ning's article "A Past Story about Domestic Chips and Operating Systems" attracted widespread attention. As Liang Ning, who personally participated in the development of independent CPUs and the "network computers" that go with them, told the unknown bitter history of domestic chips in the article.

So why is it that despite so many years of strong support from the country, domestically produced CPU chips are still lagging behind internationally, while Canaan Creative, a small company, is able to lead the world?

The key point lies in the chip type, this "core" is not that "core".

Canaan Creative designs ASIC chips, which are simply customized chips for specific fields. Compared with common general-purpose chips such as CPUs and GPUs, as well as semi-customized FPGAs, the computing power and efficiency of ASIC chips are directly customized according to the needs of specific algorithms, so they can achieve advantages such as small size, low power consumption, high reliability, strong confidentiality, high computing performance, and high computing efficiency.

However, it should be noted that the advantages of ASIC chips are based on specific algorithms. Since ASIC chips are fixed for specific algorithms, once the chip is designed, the algorithm it adapts to is fixed and cannot be applied to other fields that are not suitable for this algorithm. Once the algorithm changes, it will be unusable.

To make an inappropriate analogy, ASIC chips are like kitchen knives, which can cut vegetables quickly and labor-savingly. General-purpose CPUs are like Swiss Army knives, which can cut vegetables, open bottles, tighten screws, and act as scissors, but each function is not particularly easy to use, and cutting vegetables is relatively labor-intensive.

In the field of mining, Bitmain once released an ASIC mining machine for the CryptoNight algorithm, but after Monero changed its core algorithm, this mining machine became almost completely ineffective, which led to the fork of Monero. Most Monero miners began to accept the new algorithm, so they mined XMR, while a small number of miners (100% of whom bought Bitmain's Monero mining machines) could only mine based on the original algorithm due to the limitations of the mining machines, so they mined XMC (Monero Classic).

Unlike the "specialized" ASIC chips developed by Canaan Creative, the domestically produced CPU chips that are supported by the state are "general-purpose" chips. The chips in the ZTE incident refer to "general-purpose" chips. According to statistics, chips are currently the largest imported product in my country, costing 1.58 trillion yuan each year, and most of them are general-purpose chips.

In the international market, chip design and manufacturing are separate. Both the design and manufacturing of "general-purpose" chips are several levels more difficult than special-purpose chips.

Canaan Creative mainly designs chips, while manufacturing is outsourced to TSMC, which accounts for nearly 60% of the global chip manufacturing market. In fact, several of the world's most well-known semiconductor companies, such as Qualcomm and Apple, design their own chips and then have them manufactured by TSMC.

The independent domestic CPU chip enterprises that the state supports include design and manufacturing, which are performed by different enterprises. In terms of manufacturing, SMIC can already manufacture 45/40nm chips. Although there is still a gap with TSMC's 7nm, due to the failure of Moore's Law and the slowdown of global CPU performance growth, the road for domestic chips to catch up will no longer be so difficult. (Moore's Law refers to the fact that every 18-24 months, the power of CPU transistors will double, and the performance will also double accordingly. It was proposed by Gordon Moore, the founder of Intel, in 1965)

The reason why Moore's Law fails is that the more silicon circuits are integrated into a specific space, the greater the heat, which makes the transistors of silicon materials unstable in a smaller space at the 5nm level, thus affecting performance. Therefore, 5nm is considered to be the performance limit of silicon integrated circuits. If there are no new materials to replace existing materials, then the existing integrated circuits will reach their performance limits and Moore's Law will fail. The fact that the performance difference between Intel's 10nm chips and 5nm chips is not much proves this.

Although we can catch up in CPU manufacturing outsourcing, independent design is different.

Since the independently designed CPU needs to be used in the supporting system and software market, it is meaningless to design the CPU independently without designing the supporting system and software. Even if the CPU, supporting operating system, and software have been designed, the user migration cost still has to be considered. Simply put, it is very difficult for people who are used to the combination of Intel CPU + Windows operating system + Office software / other professional office software to independently choose the combination of domestic CPU, supporting operating system and software that may not work well. Unless the combination of domestic CPU + supporting operating system and software not only catches up with the existing combination of Intel CPU + Windows operating system + Office software / other professional office software, but also surpasses Intel CPU + Windows operating system + Office software / other professional office software by one level in terms of user experience and price, it is possible to achieve real user migration and profitability, which is extremely difficult. Therefore, the road to domestic production of independent CPUs is long and arduous.

Media exaggeration

Back to Canaan Creative, after it announced the mass production of the "world's first 7nm mass-produced chip", some media used all kinds of exaggerated headlines that ignored facts and misled the public in order to attract attention.

Image from Chip Intelligence

First of all, Canaan Creative produces ASIC chips, which cannot be used in "supercomputers" at all. It is an obvious mistake to call it a supercomputing chip.

Secondly, Canaan Creative is mainly responsible for design, not mass production.

Third, comparing the technical parameters of ASIC chips with those of general-purpose chips from Apple, Qualcomm, and Nvidia shows a complete lack of knowledge.

Fourth, completely breaking through the blockade is simply ridiculous. What kind of historical moment is this? A dedicated chip company with an annual profit of hundreds of millions is completely powerless to fill the country's overall 1.5 trillion gap in general-purpose chip imports, right?

I won’t give more examples.

In addition, Canaan's 7nm chip is not the world's first 7nm mass-produced chip. The real first should be Apple, but it just disdains to announce it. In January this year, TSMC General Manager Wei Zhejia said that TSMC has more than 10 customers taping out 7nm process technology. In the first quarter of this year, TSMC's 7nm process has been mass-produced.

As for "mass production", I was also surprised to see it, because for TSMC, 7nm is the most advanced process, and its production capacity is very limited. Such limited production capacity should be given priority to major customers, such as Apple, Huawei, Qualcomm, etc. I wonder where Canaan Creative ranks. Of course, there is no clear definition for mass production. In some fields, at least dozens of products can be considered mass production.

Industry benchmark

Of course, Canaan Creative’s hard work in research and development also has a very positive significance for the entire chip industry. The first launch of 7nm has made it a new industry benchmark.

From a technical point of view, the 7nm chip greatly improves mining efficiency and reduces energy consumption by 40%. This is good news for the Bitcoin mining industry, which has been criticized for wasting energy.

In addition, a Taiwanese semiconductor expert said: "The first mass production of advanced processes originally needs to be led by dedicated chips with simple structures. In the past, it was memory chips, then FPGAs, and now mining chips and TPUs. Little did we know that many technologies of TPUs still need to be learned from mining chips. The significance of the first mass production is to prove that the yield rate (D0) of logic transistors has reached the mass production level, which is a very important indicator for wafer fabs. As for the analog circuits required for subsequent communication chips, they are generally not made with the most advanced processes."

For Canaan Creative, which is about to go public, the leading core technology will surely bring about an increase in profitability and will also add weight to its IPO approval.