Yesterday (May 22) afternoon, ARM Chinas direct public relations call Tiger Sniff said that the new product (AI related) conference originally scheduled to be held in Shenzhen in early June was notified of cancellation and there was no specific plan to postpone it.
After hanging up the phone, I actually had the worst idea in mind.
Sure enough, just after 7 oclock last night, the BBC got ARMs internal notification document directly: because of the technology from the United States, ARM (UK) has asked employees to stop all contracts, support and pending agreements that are in force with Huawei and its subsidiaries.
After reading this news, Tiger Smell also learns part of the situation from relevant people in ARM China - the cancellation of ARM Chinas new product launch conference is also due to pressure from the British parent company, which obviously wants to avoid suspicion completely.
Huawei also issued a brief statement later:
We value close relationships with our partners, but we also recognize the pressure some of them are under because of politically motivated decisions. We believe that this regrettable situation can be solved. Our first task remains to continue to provide world-class technology and products to our global customers.
So far, the complete suspension of cooperation between ARM and Huawei has become a foregone conclusion. But what do we really think about it?
First, what is the existence of ARM?
For a few people, ARMs name is not unfamiliar, after all, its a company that has boosted the whole smartphone era with its own technology.
ARM companys fist product is ARM processor, which is characterized by the use of RISC (Reduced Instruction Set, relative to the CISC complex instruction set used by PC), through the simplification of the underlying operation of the computer, so that the whole processor is more efficient. This also makes it the best choice in the era of smart terminals with limited power supply and mobile use.
Of course, it is also very important that ARM companys own operating mechanism - ARM does not sell physical products, but only technical authorization, from the most basic instruction set, to specific Core design, and even follow-up processor optimization, there are special technical authorization projects.
This model is a great invention in itself. To some extent, it is comparable to TSMCs wafer-on-wafer business. There are two main reasons for this.u2014u2014
First, the entry barrier of mobile processors has been greatly reduced.
Designing chips is not an easy and interesting thing, not to mention how to recruit enough talents for cutting-edge technology research and development, there is also a energy consumption of tens of millions of streaming, testing, and finally software and hardware compatibility, optimization and so on.
However, due to the existence of ARM, the burden of handset processor manufacturers is lightened, because they no longer need to carry out the most risky underlying processor development, but can buy authorization directly from ARM, and then take this authorization to their own wafer factory or other wafer generation factory, using ARM chip design information fast (compared with self-developed chips). )u201d Produce CPU, SoC and other chips that you need.
Moreover, this model is also under great pressure. ARM also needs to shoulder the expectation of the whole ecosystem for processor performance progress. If we want to earn more revenue, we can only make ARM ecosystem bigger. This is undoubtedly an admirable company (saying that ARMs money-collecting media can go back to wash and sleep).
Secondly, the authorization mechanism of ARM ensures the order of the whole ecosystem and maintains the good compatibility of hardware and software ecology.
Unlike X86 architecture in PC era, the RISC architecture of ARM has been evolving greatly. In the face of huge performance requirements, the instruction set of ARM has been evolved more than ten times. If we want to give full play to the hardware performance of a processor, the software system must support these instruction sets well.
It is precisely because of the existence of ARM that the evolution rhythm of the instruction set of the whole ARM processor has been well controlled. The system does not need to adapt to the new instruction set frequently, but also can optimize the instruction set more deeply.
As for why ARM is now turning a blind eye to Huawei, how many technologies from the United States are included in the existing ARM processors? Up to the time of publication, Tiger Sniff had not been able to get a comment from the relevant people. But I believe it is still under direct pressure from the U.S. government.
Appointment is not just an answer in an interview the day before yesterday: The United States is a country ruled by law, American enterprises must abide by the law, the real economy must abide by the law. The media should not always scold American companies. Most of us speak for American companies and scold American politicians if we want to scold them.
2. What is the situation of ARM China?
In May last year, Hu Wen made a special report on the establishment of ARM China Joint Venture Company (China Cores First Breakthrough War).
Previous outflows of ARM China fundraising PPT
The biggest bright spot of the joint venture is that Chinese investors account for 51% of the shares, and ARM accounts for 49% of the shares, and it directly takes over all of ARMs business in China.
Previous outflows of ARM China fundraising PPT
Unfortunately, management dominance does not equal the transfer of technology as a whole. In an ARM China fundraising PPT, it clearly showed the authorization relationship between ARM China and ARM UK parent company and Chinese customers: ARM UK parent company authorized ARM China first, then ARM China could authorize Chinese customers. This order can not be disturbed.
Considering that the British parent company of ARM has issued a notice to suspend cooperation, the Chinese subsidiary of ARM can not cross the permission of the British parent company to authorize Huawei technology accordingly.
It is a little profound that the establishment of ARM China is closely related to the rise of Haisi. First, Huaweis smartphones continue to grow rapidly. Second, Huawei itself is the leader in the whole field of communication. Now it has developed an ARM server CPU such as Kunpeng 920. The establishment of a Chinese subsidiary company will undoubtedly make the cooperation between the two sides more convenient.
Third, how much impact will Huawei suffer?
Objectively speaking, it should be bigger than most people think.
According to ARMs own financial report, sales in China accounted for about 20% of ARMs global sales in fiscal year 2018, and most of that figure was contributed by Huawei Haith.
As we mentioned above, ARM licenses processors in a variety of ways, but from the lowest infrastructure to the final processor, there are several main categories:
1. First is the instruction set of the processor. Instruction set is actually part of the hardware design logic of the ARM processor, and it is also the most basic part of the whole ARM processor. You can think of the real CPU as a lamp, and the instruction set is a detailed description. If you are strong enough, you can also follow the instructions to make your own lamp. As long as you follow the instruction set as a processor, then you will be able to find compatible systems, and ultimately with the help of the ecological application.
2. The second is Cores core authorization. Because the instruction set is still too basic, ARM will also provide actual processor design drawings to customers. For example, the latest Cortex-A76 and Cortex-A55. These design drawings are already available directly to wafer factories at the production level.
3. The last is to authorize the improvement and adjustment of ARM processor. For example, Huawei, Qualcomm and Samsung all have certain technology accumulation for ARM processors. They will choose to develop new processors on ARM, even participate in the architecture stage, and optimize their processors. This is why Qualcomm and Samsung do not have their own code name for the core of their SoC chips.
Based on existing information, Huawei is likely to have obtained permanent authorization for the ARMv8 instruction set. In other words, Huawei can at least develop its own ARM processor and continue to enjoy the compatibility of the entire ecosystem with this instruction set.
But unfortunately, the ARMCore authorization that is still iterating and updating, as well as processor improvement and optimization, deep collaboration and so on, will certainly be affected by the suspension of cooperation.
On the actual product level, if the relevant ban in the United States continues, and if Huawei still wants to keep its chips competitive enough in its own products, Huawei will actually have to complete the original ARM + competitors workload. You know, the former is an old semiconductor company with more than 6,000 employees, which has been in the field of RISC processors for nearly 30 years.
Huawei Haisi probably has a hard time facing such a big challenge all of a sudden.
However, the Kirin 985 processor, which has been exposed in the past few months and is rumored to be released next month, will probably not be affected much. After all, according to the usual research and development time in the semiconductor industry, Kirin 985 should have entered the stage of large-scale film production. As for Kirin 990 after 985, there is not enough information to judge the impact.
But it is certain that the project after Kirin 990 will encounter a lot of trouble. Not to mention the new areas that Huaweis own chips are involved in, such as server processors and base station chips.
IV. Personal Views and Summary
To some extent, I personally think that the suspension of ARM cooperation is not the most serious.
Take mobile SoC chip for example, Huaweis Kirin 980 has actually reached the worlds leading level (performance, communication capacity, energy consumption), and we are now in a period of widespread popularization of 4G network, and 5G promotion is still in the process of gap. And the smartphone itself has entered a relatively flat development period, if you really want to take Kirin 980 will take a year or two is not impossible.
By contrast, chip production is far more deadly. Kirin 980 also owes its outstanding performance to the 7-nm process of TSMC to some extent. At present, the stable output level of domestic wafer factories is still around 14 nm. To make a simple algorithm, the area will be four times larger if the process is twice as bad (the process is edge length, the area needs square), and the power consumption and heating will rise simultaneously. After becoming fat, Kirin 980 is absolutely impossible to plug into your mobile phone.
There is no doubt that there will be a wave of independent innovation in China today because of the suspension of cooperation between ARM, but it is also at this time that we need to go back and read the two paragraphs of Ren Zhengfeis interview the day before yesterday.
Why not take a cold bath? I think the most important thing for us is to be calm and calm. Blood boiling, slogans flying all over the sky, and its useless not to be able to fight in the end. Its true to win in the end.
Independent innovation as a spirit is worth encouraging. Innovation on the basis of human civilization is correct. But we should see that technological innovation needs to stand on the shoulders of our predecessors. For example, Haisi has not started independent innovation from the source, but also paid a lot of intellectual property fees to others. Some have signed cross-licensing agreements, some agreements are permanently authorized, some of you have me, some of you have formed our own innovation on the basis of others.
Because semiconductors are really an industry that has to be globalized.
Pictures from: Moore Elite
In 2017 alone, the total output value of the semiconductor industry is as high as 430 billion US dollars, and the resulting electronic industry is as high as 1.5 trillion US dollars. This is not to mention the high cost of materials and R&D investment in the semiconductor industry.
In such a high-speed iterative and competitive frontier technology industry, the pursuit of independent innovation is perhaps the worst outcome.
Source: Editor-in-Charge of Tiger Sniffing Network: Qiao Junjing_NBJ11279