[PConline] in mid June, there was a processor in the Linux kernel patch, which was normally a normal operation in the past, unlike the processor code named Dhyana, which was behind the Hygon (Hai ray), and it could also use the code path of the AMD Zen (Zen) processor, from the name of the name. The code shows a great relationship between the two, because the Dhyana processor is the Chinese version of AMD's authorized China Hai GUANG company, which means that AMD authorizes the most advanced X86 processor technology to China, which has aroused hot debate both at home and abroad.
Zen architecture processor is one of the most advanced X86-64 processor technologies, and can be said to be the Pearl of the IT technology industry. Why does AMD easily sell it to Chinese companies? What impact will China have on the market after acquiring top X86 processors? Why does Intel have no objection to AMD? Today we are going to talk about the influence of AMD's X86 technology authorization on China and AMD and Intel.
The status quo of domestic processors: a variety of processor systems are flourishing, and lack of high-performance universal processors.
In the past two years, everyone has seen the emphasis on the semiconductor industry at home. The processor is the key point. The current domestic processor system can be said to be a hundred flowers, from ARM to MIPS to Alpha, Power and X86 and so on, almost all the processor architecture has been touched.
ARM processors are publicly authorized and are well done at home, such as the HUAWEI's Hester and purple light, which are shipped over hundreds of millions of ARM processors each year. In the MIPS processor, there are dragon core, Jun Zheng and other companies in China, and long core has been able to do its own research and development after the authorization of the MIPS instruction, but MIPS is not the mainstream of the market.
There is also a Alpha instruction processor in China. The Shenway processor of Jiangnan Institute uses this instruction set. Of course, it is also modified. It is mainly used in supercomputing processors, the strongest supercomputer Shenwei in China.
As for the IBM dominated Power processor, Zhongsheng Hongxin has also obtained the authorization of the Power 8 architecture, but the Power architecture is not the mainstream of the market at present, and the prospect of development is not good.
A truly high-performance and versatile processor is only X86, and Shanghai is currently working as a X86 processor in Shanghai. These companies are joint investment groups affiliated to the Shanghai SASAC and VIA weyson, and VIA is the only three existing X86 authorized company.
Megabyte's X86 processor is improved on the VIA X86 processor, the latest KX-5000 architecture is 8 core architecture, and the performance is close to Intel's Core i3-6100 processor. Considering the basis of the previous VIA processor, this progress is not easy, but it is far from the high performance. The 8 core won 2 nuclear processors is not worth being proud of. .
Therefore, as far as the current domestic processors are concerned, we have made some breakthroughs in a few fields, but we lack a high performance general processor, which can not be applied to important markets such as desktop and server. The global PC shipments are more than 350 million units a year, and the X86 server, though only ten million, is of higher value, and this city is of higher value. With the absence of the Chinese processor companies and the domestic demand for autonomous control, the coveted high performance X86 processor is easy to understand, and that's why Hai Guang is willing to spend $293 million from AMD's Zen processor authorization.
AMD is not the first time to authorize X86 to China.
When it comes to X86 technology authorization, AMD is not actually the first time to do so. As early as 2005, AMD CEO Hector Ruiz announced the AMD's X86 Geode technology to the Peking University microprocessor center when he visited China in China.
But that authorization did not change much at home. First, the x86 Geode technology authorized by AMD was not advanced, and it was also the technology acquired by AMD, mainly low power processor technology, and the authorized Peking University was an academic, educational institution, and did not produce anything in the commercial market. Benefit.
But AMD this time authorization X86 technology is different, the first AMD authorization is the latest Zen architecture processor technology, specifically the EPYC server chip, and the authorized party of Tianjin sea is the company of the middle branch dawning. There was a CAS academic background, but the business was stronger, so the two sides met, Chinese companies first obtained high performance, 64 bit X86 processor technology, and AMD received $293 million for authorization. At the beginning of 2016, the money was still important for AMD. In the year, AMD was still in a loss.
This cooperation is far more complicated than the last time that AMD and Haiguang set up a total of two joint ventures.
The two companies that are set up are also different. The AMD led CHMT company will be responsible for the customization of the Zen processor, while the Chinese controlled CHICD company will customize the other parts of the SoC and the marketing work according to the customer's needs.
It can be seen that this complex equity - based joint venture ensures AMD's control over domestic X86 processors, while the Chinese holding companies are mainly responsible for processor customization and marketing, and AMD still controls the technical direction of the home-made X86.
Why does Intel don't oppose it? It turned out that Intel also authorized X86 to Chinese companies
Many people know that the X86 processor is developed by Intel, and the X86 authorization is in the hands of Intel. Then why does Intel not jump out of opposition when AMD authorizes X86 technology to Chinese companies? After all, this is not good for Intel.
It has been two years since AMD announced licensing X86 technology to Chinese companies. Intel has not expressed its opposition. That means they will not object to this issue. There are three main reasons.
Second, AMD still controls the development of China's X86 processor through complex ownership structure design, and the core technology is controlled in its own hands, while the AMD and Intel cross licensing agreements do mention that both sides have to accept the other's consent when accepting the holding of other companies, but AMD is not being controlled by people now. There is no violation of the agreement.
Finally, Intel does not come up against authorizing X86 to Chinese companies. It is also because it is not good to offend China. It is a need for China to get the advanced X86 processor authorization. The opposition of Intel is bound to dissatisfy the government and disadvantageous to the image of Intel.
In fact, Intel himself has also done the X86 processor to authorize the Chinese company. Intel invested 1 billion 500 million dollars in the mobile market, invested 1 billion 500 million dollars in the purple light company, and authorized the advanced 14nm X86 architecture to the exhibition. The latter introduced the SC9853I processor last year to use the Intel 14nm process X86. Architecture, so on the X86 authorization, Intel company can not oppose AMD's action.
AMD two hands: EPYC sell correctly, domestic X86 attack the specific market.
AMD dared to authorize the most advanced X86 technology to Chinese companies. This is not a mere demonstration of the Chinese market. It is definitely a picture. First of all, AMD has found a new model of money through the X86 authorization. The $293 million authorization fee is not high, but it can be saved in 2016, and this is only a preliminary cost. After the subsequent processor listing, AMD can also benefit from the two joint ventures, after all, AMD still holds 51% and 30% of the shares.
There is no doubt that domestic X86 processors authorized by AMD technology have a specific target market, such as government procurement, high-performance supercomputing, military workers and so on. In these areas, it has been demanding self-control, especially in some occasions involving important economic, military and government affairs, if the other ARM and MIPS are transferred. Although they can be controlled independently, they are not common to the X86 processor, and the software support involved is too complex and can't keep up. Now it is much easier to have a domestic X86 processor.
More important is that AMD can continue to sell its own EPYC server processor in the Chinese market, which is not two selected with the domestic X86 processor. AMD's EPYC processor has now been approved by Tencent, Baidu and other companies, applied to the Tencent's cloud services, and compared to the Intel processor. The price of cloud servers is lower, which improves the competitiveness of enterprises.
AMD now has less than 1% share of the server's processor market, almost negligible, Intel's share is 99%, and revenue from the server chip chip market is up to $about 20000000000 a year, and the profit rate is far higher than the PC processor, and AMD is even more than a 10% share from Intel. With $about 2000000000 in revenue, AMD now has a revenue of $about 5000000000 a year.
Therefore, for AMD, the authorized X86 processor is equivalent to an ally in China, because China is already the largest chip market in the world, AMD and Intel have 20-25% from the Chinese market, and the Chinese market is growing rapidly, and AMD grabs the market share of the server from Intel. The key is the Chinese market, authorizing X86 to Chinese companies is a win-win cooperation, there is no reason to refuse, the only not very pleasant is Intel.
Conclusion: the prospect of domestic X86 is bright, but there are still some problems.
In the X86 authorization, AMD and China are all needed, so AMD can make money and expand the market share. China can obtain a powerful and technically advanced X86 processor, and customize the processor function according to the domestic needs to realize the independent and controllable requirements, and the complex shareholding of both sides. The structure also guarantees that AMD patents will not be easily lost, and objectively trains more processors in China, which is a win-win situation.
AMD's own EPYC chip still has advantages in architecture and technology.
It can be predicted that the future of domestic X86 processors is bright, but it still has many unsolved problems. Compared with the EPYC processors that have already been put on the market by AMD, the progress of domestic X86 processors is much behind, and at present we know very little about the details. For example, what improvements have been made in China, when the domestic X86 processors will go on the market, and another key is which manufacturing process will be used by the domestic X86 processors.
On this issue, CIC will be able to produce 14nm process by early next year and not necessarily be suitable for high performance X86 processors, so this problem will affect the quality of domestic X86 processors, and the significance of autonomous control will be discounted if it is not used in domestic production.