In the eyes of Academician Ni Guangnan, RISC-V is the most promising "game breaker" to break the monopoly of X86 and ARM in the next few decades, and it is also the only way for the mainland to realize the independent control of the CPU architecture. However, in the twelve years since risc-V was born, there has still not been a single "handle". People always say how potential RISC-V is, but if this ecology has not been able to "create gods", then it is just a bunch of stragglers. Bao Yungang already knew this, and he formed a team to create a new RISC-V force "Xiangshan", with only one goal: to create "Linux" in the CPU field.
Two big mountains
In 2013, Alibaba's last IBM minicomputer rolled off the production line. A successful example, like a reassuring agent, set off a tide of "going to IOE" from the private sector to the official. Subsequently, the domestic server manufacturer "Wave" began to gleefully take over the domestic market share of overseas companies such as IBM, transforming into a domestic "brother", with more than 30% of the domestic market.
But the problem is far from solved. Inspur and other domestic servers have a fatal weakness - no chips. Selling domestic servers, but the core is a foreign core, this is a "half of the defeat". Subsequently, we tried to design our own chips, such as "Loongson", "Kunpeng" and so on.
You think this problem is solved? It's still far away. We fought all the way from the "server country" to the capital "Chip City", only to find that this "Chip City" is a huge "national China", and the walls he built are thick enough to make every changsheng general afraid. To break through this majestic city, you must start with the architecture of the roots – bypassing the X86 and ARM.
Taking Huawei as an example, the basis of its chip design is the purchased "ARM architecture", which cannot be expanded and upgraded. The next generation of AMR architecture comes out, and Huawei has to spend a huge amount of money to buy it. Once ARM is not sold, the result is understood. And the "Dragon Core", like the lonely general "Zhao Zilong", although the silver gun is in hand (purchased a permanently developable MIPS architecture). But the enemy is outnumbered (the ecology is too poor, the mainstream operating system, application software do not take him to play), can only keep the less master (Loongson is used in the area of the most need for security).
Spread rumors with a mouth, refute rumors and run off legs
The domestic bitter "CPU" has been for a long time, and X86 and ARM are like two big mountains, enveloping all downstream industries in China. Especially after the X86 and ARM were cut off from Russia, our last bit of "illusions that should not exist" also disappeared. As a result, many people began to look for new fantasies to fill the unacceptable sense of loss. At this time, the first generation of "Xiangshan" processor "Yanqi Lake" developed by the Chinese Academy of Sciences team ignited their emotions and led them to a climax.
So, all kinds of bragging came. For example, take the "Yanqi Lake" benchmark Intel's i9 10900k, ARM's A73 and other top processors; in technology, it completely bypasses the United States. How long was "Yanqi Lake" established? How long have Intel and ARM been around? Could it be that the chip industry also has a story of "being born in Rome"? Moreover, chip technology has developed for many years, and the technological contribution of the United States in it cannot be denied. These obviously unrealistic boasts are still believed.
Even more outrageous, "Yanqi Lake" was developed, and it has not yet taken the EDA process, nor has it been a tape-out, and the stability and power consumption data are unknown. The Internet has begun to decipher its data, nonsense about "first-class level", "comparable to Intel" cloud nonsense. In the end, they also gave the conclusion that "Fragrant Mountain" will be our turnaround point.
These false reports have bothered the research team, and they can only choose to publish articles to refute the rumors. But it turns out that it is obviously unrealistic to rely on the strength of a few people to fight against the entire self-media. A child who has just been born, still learning to speak in his teeth and teeth, and a toddler has become a giant who turns the tide in the eyes of people with hearts. The reason why this naked phenomenon of "catching ducks on the shelves" appears shows that netizens' desire for domestic chip breakthroughs exceeds rationality, and the chip crisis makes ordinary people feel anxious.
Xiangshan is still in the "qi training" period
But the real "Fragrant Mountain" is actually like this. Its first version is "Yanqi Lake" and the second version is "South Lake". The RTL code for "Yanqi Lake" was completed in April 2021 and uses a 28nm process tape-out at a frequency of 1.3GHz; "South Lake" does not currently have a tape-out, but is expected to use SMIC's 14nm process with a target frequency of 2GHz. From the data point of view, there is still a big gap between the performance of "Xiangshan" and the top processors of the same kind.
The first is not to get stuck in the neck
Just started, the performance of "chicken ribs", not "the chosen son", "Fragrant Mountain" is really a bubble? Don't be lost, if you can see the other side of "Fragrant Mountain", you will feel that the willows are dark and the flowers are bright and another village.
First of all, "Fragrant Hills", as a member of the RISC-V family, has a popular advantage - it will not be stuck in the neck. Regarding this matter, many people still have doubts: RISC-V's hometown is in the United States, and when the order is issued, can we still use it? In fact, don't worry, the essence of RISC-V is an "open instruction set" that does not get stuck. We can understand it this way, RISC-V is a statement announced by American scientists to the world, and if the United States wants to prohibit everyone from creating on the basis of this statement, it may be "unconstitutional", and no one will take care of it.
The former everyone must understand, the latter is actually very simple, take the X86 and ARM to make a comparison to know. The old and strong X86 and ARM have undergone the baptism of time, expanding, iterating and upgrading many times, forming a thick patent barrier on the basis of the original. Latecomers simply can't get around it, and can only spend money to buy licenses from ARM (Intel generally does not authorize). But RISC-V is different, it is a very basic thing, there is no patent barriers, everyone develops the architecture on the basis of RSIC-V. Moreover, the United States does not have a "product" (commercial IP) that everyone relies on in the RISC-V field, and the card does not get a card.
This advantage of Xiangshan is very general, because everyone is engaged in it. Like The domestic Ali, Huawei, ZTE, etc., foreign countries like Qualcomm, Samsung, Apple, etc., even X86 king of Intel also have to engage, which shows its popularity. In fact, the top companies in the IT industry will not miss this opportunity, this is the best time for them to escape the clamps of Intel and ARM and protect the security of supply. For Intel and ARM, they also need to go deep into the enemy camp and know each other to ensure that they are still evergreen.
The Heart of Open Source
In the same end, the ultimate purpose of these enterprises is actually the same, to extract or even monopolize the commercial value of RISC-V. They will go all out to develop excellent RISC-V processors, and then license them for a fee; then they will continue to upgrade on this basis, maintain a leading edge, and make money by taking the X86 and ARM models.
Fragrant Hills stands out from the crowd. His goal is to create an "open source thread" like Linux, a platform on which all developers can develop and innovate. Liux, as a representative of open source software, can be accessed and modified by anyone to make it better serve consumers. In this way, Linux can brainstorm and grow.
For example, the unlicensed X86 is like a commercial housing developer, only selling houses (products), the price is particularly expensive; and the ARM that is willing to authorize is like a hotel, you want to go in and live, but you have to give money, and if you don't give money, you will be driven out; and what Xiangshan has to do is to provide free design drawings, everyone can learn, and then build their own house. Of course, now this drawing is more "spicy chicken", and it can't build a good house. But everyone can get together, put forward their own opinions and modification plans, make this drawing better and better, and eventually build no less than the house and hotel of the first two.
Fragrant Hills may be the future
"Open source processor" can be said to be the true essence of Xiangshan. To move in this direction, the research team is likely to make little money, and they will end up with more academic acclaim. The author believes that although Xiangshan undoubtedly "dominates the world", it represents the real future of domestic processors.
Risc-V's biggest problem now is that it is too weak. If it wants to grow, it needs at least rapid development and sufficient ecological support. It's like the little master brother who has just entered the Shaolin Temple, if he wants to defeat the master brother, he must grow faster than the master brother's martial arts, and then in the process, he has to have a full meal, otherwise he doesn't have the strength to practice kung fu.
But what are domestic companies like now? Most of them do their own research, trying to overtake in curves. This is unrealistic, no matter how strong your research and development strength is, can it be compared to Samsung, Qualcomm, and Intel? Not necessarily. Well, we take a step back, even if you have strong research and development strength, but RISC-V is still in the early stage of development, you must have to invest a lot of money in the early stage, your pants belt with Samsung, Qualcomm has a strong? Not necessarily.
Therefore, in the face-to-head competition, we are indeed the same running line, but if we really run, we will most likely not be able to win. What Xiangshan does is to let these companies not fight alone, and everyone is working together to develop the RISC-V architecture efficiently so that we can outperform. Moreover, Xiangshan is open source, everyone can use, those overseas companies plagued by X86 and ARM, as well as engineers interested in open source processors around the world, will be willing to participate in and jointly promote the "Xiangshan" project.
To put it bluntly, what Xiangshan did was to turn private rice into a big pot of rice. It is capable of fundamentally solving the monopoly problems of X86 and ARM. From this point of view, to solve the crisis of domestic enterprises being stuck in the neck is to "go along the way".
There is still a long way to go
Having said so many good things about Xiangshan, are you optimistic again? Don't worry, we still have to pour cold water in the end.
Although the Xiangshan project has set a lofty goal like "the unity of the world", it is a daydream to talk about ideals without reality. Unfortunately, "Fragrant Mountain" was really poor and white before. We also saw that although there are bright spots in the product, the overall performance gap is large; in terms of development resources, there are not many people, and there are not enough resources, in the end, there is still no one to support.
However, recently there is good news that the Xiangshan project has a new home: Kaixin Institute (Beijing Open Source Chip Research Institute), a private non-enterprise that has only been established for more than four months. A website describes the Kaixin Institute as an innovation consortium jointly initiated by a group of leading enterprises in the industry and top domestic scientific research units. It can be seen that the top players in China have joined forces to strengthen the RISC-V ecosystem, and the concept of Xiangshan meets their needs, which will inevitably become the key project of kaixin academy.
This is a good sign, at least in terms of resources, the Xiangshan project should be no worries. Next, the Xiangshan project should consider how to accelerate technological development and ecological construction. As for whether Xiangshan will succeed in the end, we have not yet known. But to borrow the words of Bao Yungang, the head of Xiangshan, it is: first live like Liux for 30 years. (Bluegrass)
Deeply pickpocket the "Xiangshan Project" of the Chinese Academy of Sciences, don't be deceived by "some people"
Don't be deceived by the self-media! The Chinese Academy of Sciences "Xiangshan" is not so strong in the "qi training" period