Author: River wine
Source: Zhenghe Island
On January 13, 2022, LoongArch held an ecological innovation conference online, and the theme of the meeting was LoongArch, the latest research and development achievement of LoongArch in 2021.
As the first chip architecture with complete independent intellectual property rights in China, in July last year, LoongArch launched the first LoongArch-based chip - LoongArch 3A5000.
As soon as this chip came out, it received a lot of attention, and after layers of evaluation from official to private, a unified conclusion was finally reached:
After 20 years of precipitation, this independently developed chip based on domestic architecture has finally caught up with the pace of the times.
At the meeting, Ni Guangnan said:
"The release of loongArch's autonomous command system (LoongArch architecture) is an important achievement of China's information industry and is expected to become a strong base for the country's internet information industry in the future."
In December last year, Loongson Zhongke officially entered the science and technology innovation board, and the ecological innovation conference on January 13 and the Loongson ecological white paper released at the same time were also a big move. From the emphasis on ecological construction, it is not difficult to see the ambition of Loongson Zhongke, if Loongson Zhongke can be successfully listed, then Loongson Zhongke will become the first domestic independent CPU stock to impact the science and technology innovation board, taking a historic step in the domestic chip industry.
After all, after nearly 20 years of stubbornness in the chip architecture, Loongson Zhongke's body has carried too many Chinese expectations.
01, ten years of dormancy, thick accumulation and thin hair
For China's high-tech products, Zhai Dongsheng, a professor at Chinese Min University, once made such an analogy: it is like a flower, inserted in someone else's vase.
Since China was suppressed by the United States, China's chip industry has been facing a dilemma.
In the face of the continuous suppression of the United States, in October last year, at the Fifth Plenary Session of the 19th Central Committee of the Communist Party of China, it was clearly put forward: adhere to the core position of innovation in the overall situation of China's modernization construction, and take scientific and technological self-reliance and self-improvement as the strategic support point for national development.
Undoubtedly, under the increasingly severe international situation, China's chip development path will encounter more and more challenges in the future, relying on scientific and technological "hard power" to break the monopoly of the chip industry has become the only way out for China's chip industry.
Hu Weiwu, chairman of Longxin Zhongke, said at the ecological innovation conference:
"In the past few decades, the base of China's information industry has been in Silicon Valley, subject to two bases, namely Intel + Microsoft, and ARM + Android, China's entire information industry has been expanded on these two bases, but now we have to change the base."
The pedestal that China wants to change is LoongArch.
As the most basic link in the chip production chain, the commercial authorization of the chip architecture is in the hands of the Americans, and from the perspective of the entire industrial structure, the risk is extremely great.
It's like being held hostage, and once commercial licensing is interrupted, Chinese chips will lose everything.
In 1986, Liu Chuanzhi of Zhongguancun in Beijing made an important decision: to break off cooperation with IBM and lead Lenovo to the route of independent research and development.
It was also in this year that after repeated discussions and studies by the State Council, a document entitled "Outline of the High-Tech Research and Development Plan" was approved.
Extended from this document is the famous "863 Project".
After the promulgation of the plan, the domestic academic community has a great disagreement on the question of "whether to independently develop chips".
One side believes that since it is determined to engage in chips, it is necessary to engage in a full set of chips, from design, research and development, and then to later production, to create a complete set of chip ecology.
The other side believes that China's chip field started too late, blindly catching up is unrealistic, it is better to use the existing technology development in the United States, and the ecology can be slowly built.
After some heated debate, the scientists finally agreed on the latter's point of view.
The reason is also very simple: at that time, China was too poor to afford high research and development costs, and under the pressure of the reality, it could only choose to compromise.
But a temporary compromise does not mean giving up, everyone is waiting, waiting for a suitable time, accumulating.
Such a wait, that is more than ten years.
02, the building should be beautifully built, and the foundation should be laid well
In July 2021, Loongson Zhongke released the latest 3A5000 series chips, the main frequency is 2.3GHz-2.5GHz, 4 core design, using 12nm process.
Objectively speaking, compared with Intel, IBM and other large companies, the data of Loongson 3A5000 is far from enough to see, but it is gratifying that the performance of 3A5000 has gradually approached the mainstream of the market, and the computing power can be comparable to Intel's i5 series 6 years ago.
In order to achieve this goal, Loongson Zhongke has been walking for 20 years.
In 2001, under the knowledge innovation project of the Institute of Computing of the Chinese Academy of Sciences, the Loongson Research Group was formally established, and in only one year, the "Loongson No. 1" was born.
With the birth of "Loongson No. 1", China's chip industry once again sounded the horn of attack, in the same period, the "Shenwei" of Shanghai High-performance Integrated Circuit Design Center and the "soaring" of the National University of Defense Technology have started, and domestic chips have ushered in a thriving situation for the first time.
With the expectations of the whole industry, the Loongson research group lived up to expectations, and launched the "Loongson II" in 2004, when the performance could already match intel's Pentium III released in 1999, and the gap between domestic chips and international chips was approaching at a speed visible to the naked eye.
However, the shortcomings of domestic chips have also begun to be gradually exposed.
As Hu Weiwu said: the production process is one leg, and the command system is another leg. Although the production process has been raised, the foundation of the chip industry, that is, the most basic chip architecture, chip IP, CPU instruction set, etc., is still limited by several large companies in Silicon Valley.
It's like, although you learn a dish, although the taste is getting better and better, the recipe of this dish is ultimately someone else's.
This burden of creating a "recipe" naturally fell on the shoulders of Longxin Zhongke, and since then, Longxin Zhongke has begun a long "long march" for more than ten years.
The launch of the LoongArch architecture is the best answer of LoongArch to this decade of precipitation.
Before the launch of LoongArch, most of the domestic CPU companies relied on foreign architectures, such as HiSilicon and Feiteng using ARM licensing, Haiguang and Mega Core using X86 licensing, and the Chinese Academy of Sciences also purchased MIPS licensing.
But with the updating and iteration of technology, the old instruction set becomes bloated in the continuous superposition, and it begins to be unable to meet the needs of the market, and building a house on someone else's "foundation", even if it is beautiful, it is a castle in the air.
Compared with the mainstream X86 and ARM architectures on the market, in fact, The biggest advantage of LoongArch is not performance.
The greatest significance of loongArch's birth lies in the independence of commercial copyright.
Long before the launch of Loongson 3A5000, the LoongArch command system has passed the intellectual property assessment of China Electronics Information Industry Development Institute, an authoritative third-party institution in China.
In the evaluation, it was basically determined that the LoongArch command system meets the original standards, which is different from the popular alpha, ARM, MIPS, POWER, RISC-V, X86, and is a new command system design.
What does it mean that LoongArch is able to pass a "check"? This means that from now on, China truly has its own chip architecture, and it is one step closer to shaking off the hand of the "card neck".
In addition, the long-term insistence on independent research and development of CPU cores is also one of the characteristics of Loongson Zhongke, in the new version of the prospectus, Loongson Zhongke also revealed that Loongson independently developed hundreds of IP such as GPU cores, memory controllers and PHYs.
Designing a high-complexity CPU chip is very complex, and its interior is composed of many IP, including CPU, GPU, memory controller, IO interface controller, high-speed SRAM, high-speed interface, phase-locked loop and other core IP.
The IP core of GPU and memory interface is a very high threshold for integrated circuit practitioners, and it is basically impossible to do without the accumulation of 5-10 years of experience.
Limited by commercial licensing, domestic chip design companies have to spend high prices every year to purchase GPU cores and memory interfaces, IP cores and other architecture licenses, and the quotation for a single use is basically from one million US dollars.
In this regard, loongson Zhongke's strategy and decision-making are undoubtedly extremely forward-looking and strategic.
More than a decade ago, when all chip companies were committed to creating a chip manufacturing process, only Loongson Zhongke adhered to its heart and insisted on deep ploughing in the field of chip architecture, just like Hu Weiwu said:
"The road of building an independent information technology system and industrial ecology is not an easy road, but after it is passed, the future is bright."
Longxin Zhongke also proved with actions that they not only had this strength, but also really made this thing happen.
03, looking forward to the future, the dragon takes off
But as Ren Zhengfei said: The reason why Huawei can survive is not a sense of honor, but a sense of crisis.
In the field of high-end technology, although Loongson Has opened up a new path, there are endless challenges coexisting with opportunities.
Referring to the strategic planning of Loongson Zhongke and the future of China's chip industry, Hu Weiwu said:
"What is a solution? Like computers and mobile phones, these two big ecosystems are big solutions, so we must build our own ecosystem, otherwise the industry will always be in the hands of others. ”
While aligning with cutting-edge technology, how to create a new industry ecology based on the LoongArch architecture has become the primary problem facing LoongArch.
LoongArch's biggest advantage is that in the entire design and production process, the maximum extent to achieve "localization", all the functional modules in the chip, are all designed by Loongson Zhongke.
Including all the source code of the CPU core, on-chip interconnect bus, DDR4 memory controller, and various IO interface modules, loongson 3A5000 uses almost no third-party technology except for the basic design environment provided by the tape-out manufacturer.
In the choice of process process, Loongson Zhongke has always been more conservative, in 2021 the latest launch of the 3A5000 processor, the use of 12/14nm process node.
The previous 3A4000, which used a mature process of 28nm, has not yet seen a 7nm process node product.
Compared with the international cutting-edge technology level, Loongson Zhongke does have a certain gap, but as far as the current situation of China's chip industry is concerned, in fact, loongson Zhongke's "thoughtfulness" is not necessarily a bad thing.
As far as the domestic market is concerned, the current 14nm-28nm process can meet the needs of most information application scenarios in China.
As a chip architecture independently developed in China, the security provided by LoongArch can greatly meet the confidentiality needs of national functional departments.
Similarly, due to LoongArch's guarantee of data security and accurate entry into commercial pain points, it has successfully achieved commercialization in the fields of banking and communications industry, and has established close cooperative relations with companies such as China Mobile and China Minsheng Bank.
Up to now, nearly 100 manufacturers in China have launched hundreds of LoongArch-based ecological products, such as LoongArch desktop, server, network security, password, etc.
Since 2020, Loongson Zhongke has led the establishment of the "Loongson Ecological Adaptation Service Industry Alliance", attracting more than 70 adaptation centers and related institutions to join, forming a linkage system of adaptation services throughout the country.
In the subsequent development, LoongArch is committed to creating a Branch of LoongArch in the international community, and at the same time forming the LoongArch Alliance, gradually opening the LoongArch instruction set to the public free of charge.
In addition, Loongson Zhongke also intends to open up the IP core implementation code of some processors, so as to create the LoongArch ecological alliance with the greatest degree of openness, attract like-minded partners to join and jointly create an independent ecology based on LoongArch.
"Scientific and technological self-reliance and self-improvement" is an important strategy of the 14th Five-Year Plan of the Country, through the continuous research of technology and the creation of ecology, the independence and autonomy of China's CPU industry can be fully realized.
In the interview with Zhenghe Island, Hu Weiwu also mentioned that although the Loongson 5000 series has already made a milestone leap, Loongson Zhongke will not stop, the 6000 series will be launched this year or next year, in the new series, Loongson Zhongke will try to introduce the international cutting-edge 7nm process.
"When we get rid of the introduced performance, no one will use Intel anymore."
As a pioneer and leader in the industry, Loongson Zhongke has a long way to go in the future, and it is expected that the future Loongson Zhongke can bring us a new independent domestic CPU ecological field as soon as possible.
At the same time, it is also expected that the "Loongson" series can be put into the commercial market as soon as possible to achieve a complete closed loop of the domestic chip industry.
The future is still long, let's wait and see!