The first half of 2018 was an eventful year for China's semiconductor industry, with several major events that made people think deeply. As a deserter in the IC industry, I have recently been thinking about many issues, including some immature ideas in the capital market, the integrated circuit industry and research institutes.
In 2008, I entered the Department of Electronics of Huazhong University of Science and Technology. After graduating from the Institute of Semiconductors of the Chinese Academy of Sciences, I entered MediaTek to develop mobile phone chips (yes, it is the X30 that is not selling well). Before I left MTK, there was an episode. The boss who recruited me to join MediaTek later came out to start his own business to make security chips. At that time, he called me to his team. In fact, from the perspective of a technical employee or the future, it is undoubtedly the best choice for the former boss. The way out (excellent boss, excellent technology, visible valuation growth). But in the end, I chose to become a second-level dog. In fact, the demand for technical personnel in the entire chip industry is still very high.
I want to divide the chip industry into several categories: digital chips, analog chips, radio frequency chips, and FPGAs. I hope that I will have the opportunity to communicate with everyone while summing up my knowledge.
Let’s start with FPGA today.FPGA-Field Programmable Gate Array, refers to all digital integrated circuits that modify and configure the internal connection structure and logic unit of the device through software means to complete the established design function. In a simple and popular way of introduction, it is like an all-around athlete, swimming like Sun Yang, hurdling like Liu Xiang, and tennis ability comparable to Li Na. FPGA is so magical that various complex functional circuits can be realized through settings. The core advantages of FPGA: high programmable flexibility, short development cycle, high programmable flexibility for parallel computing. At the same time, FPGA also has many problems that need to be solved, FPGA limiting factors: cost, power consumption and programming design.
According to data from the authoritative market research organization Gartner, the total global FPGA market in 2014 reached 5 billion US dollars, of which China's market share was 1.5 billion US dollars, and the Chinese market accounted for one third of the global market. The analysis organization predicts that the compound annual growth rate of the global FPGA market from 2015 to 2020 is 9%, and the global FPGA market will reach 8.4 billion US dollars by 2020.
In terms of specific subdivisions, before FPGA is used for deep learning, FPGA has three main application directions: (1) High-speed interface circuit design of communication equipment, FPGA can be used for high-speed signal processing, generally if the AD sampling rate is high , The data rate is high, then FPGA is needed to process the data, such as decimating and filtering the data, reducing the data rate, making the signal easier to process, transmit, and store; (2) Digital signal processing direction/mathematical calculation direction, including image processing , Radar signal processing, medical signal processing, etc., the advantage is that it has good real-time performance and uses area for speed, which is much faster than CPU; (3) SOPC, that is, the underlying hardware environment of an embedded system built using FPGA platform, and then design The person develops embedded software on it.
FPGA International MarketWhen Xilinx first created FPGA in 1984, it was still a simple glue logic chip, but now in signal processing and control applications, it has replaced custom application-specific integrated circuits (ASICs) and processors. In the long history of just 30 years, more than a hundred industry giants have entered this market, but in the end most of them have failed. These companies include Intel, Philips, Agere Systems, AMD and Motorola and other internationally renowned chip design manufacturers. This is because Xilinx and Altera have been cultivating in this field for decades, and the two have continued an arms race, occupying 90% of the market. The patent protection laid out in advance has formed a strong market barrier for latecomers and almost blocked all of them. The thoroughfare to FPGA commercial products. While Microsemi and latTIce, QuickLogic once challenged their dominance, but their combined market share is less than 10%. In the end, the market share continued to decline due to various reasons, and the breakthrough and challenge to the duopoly was not achieved.
It can be said that FPGA is one of the products most in need of technology and monopoly breakthrough in the global chip design industry, and it is the most difficult technology product to break through and break the pattern in all chip fields.
Xilinx: The invention of FPGA overturned the semiconductor world and created the Fabless (fabless) semiconductor model. Xilinx's product portfolio combines FPGA, SoC and 3DIC series All Programmable devices, as well as fully programmable development models, including software-defined development environments. The products support various intelligent, interconnected and differentiated applications driven by 5G wireless, embedded vision, industrial Internet of Things and cloud computing. In December 2014, Xilinx's 20nm chips were mass-produced. In 2015, Xilinx immediately launched new 16nm FPGAs and SoCs, and adopted a new type of memory UltraRAM. Therefore, following 28nm and 20nm, it continues to lead the industry. The company's products are vertically arranged in various manufacturing processes, because 20nm, 28nm, 40nm and other process products will coexist in the market to meet various applications with different complexity.
Altera: It is the advocate of "System on Programmable Chip" (SOPC) solutions in the world. Combining programmable logic technology with software tools, intellectual property (IP) and technical services, it provides high-quality programmable solutions to more than 14,000 customers worldwide. In 2015, Intel announced the acquisition of FPGA manufacturer Altera for USD 16.7 billion. This is the largest acquisition in Intel's history. With the completion of the acquisition, Altera will become Intel's Programmable Solutions Division. In the process battle with Xilinx, the two giants are leading the way. Altera's roadmap, the most recent product series "Cedar" (replacement of Cyclone) adopts TSMC's 16nm process and will be delivered in the first half of 2016. The "Oak" series uses Intel's 14nm process and will be delivered in the second half of 2016. "Sequoia" uses Intel's 10nm process and will be listed in 2018. Altera will win the battle for the 10nm process node.
LatTIce: LatTIce (is the leader in the global intelligent interconnection solutions market and the second largest FPGA manufacturer in the world. It provides market-leading IP and devices with low cost, low power consumption and small size. There are three main products: programmable Logic; video transmission; millimeter wave solutions. The company’s end users are mainly OEMs in the communications, computer, industrial, automotive, pharmaceutical, military and consumer goods markets. Under the attack of the twin brothers, LatTIce's road is also going It is getting harder and harder to compete with the first two in the mid-to-high end market.
China's breakTsinghua Unigroup wanted to develop rapidly through the purchase of Lattice. After being reviewed by the Committee on Foreign Investment in the United States (CFIUS) and opposed by the Trump administration, the acquisition ended in failure. Regarding how domestic FPGA breaks and develops, under the condition that it cannot introduce technology through extension, the country can only break the current pattern through human research and development and technology accumulation.
Jingwei Yage: The company has gathered a group of technical elites who were the first to cultivate and try in the FPGA industry. They adopted the SoC FPGA strategy and integrated the CME-GM7 series of DSP, Memory, MCU and other units on the chip, trying to take advantage of integration Break down the barriers of the FPGA market. The company actually has two product lines: 1. The Jinshan series developed from the ground up for the low-end market; 2. The product line of American CSwitch is acquired for the high-speed communication market.
Product model: M7 Huashan series; HR3 pure FPGA low-power series; M5 Jinshan series; M1 Hengshan series.
However, it is a pity that under the specific domestic market and market environment, after the company made its strongest voice, it also entered a difficult situation due to various reasons, making it difficult for itself. FPGA has a very high threshold for technical support. Since Jingwei Yage has no way to achieve Pin-Pin, each product must have a technical engineer to maintain and follow up for a long time, but the success rate of the product may not be 10%. Under such a market environment, the company will fall into a very passive situation. Long-term strategic errors have caused Jingwei Yage to fall into a deadlock in talent, capital and development.
There are many things to think about in the failure of Jingwei Yage. The success of the FPGA industry lies not only in the products, but also in the ecosystem platform construction of the product line. This ecosystem platform includes: FPGA chip, EDA tool, IP library...Indispensable. A complete ecosystem can provide users with more comprehensive design resources, thereby highlighting the advantages of the system, quickly adapting to various market application changes, and quickly seizing the market. At the same time, market recognition is an important factor that determines whether chip manufacturers can achieve technology and capital accumulation in high-speed product iterations. Any high-performance product that is not accepted by the market will fail.
The domestic situation after the Whampoa Military AcademyThe efforts and achievements of Jingwei Yage on FPGA have given great motivation to local successors; many technical researchers of Jingwei Yage have entered the subsequent establishment of Shanghai Anlu Technology, AGM and Gowin Semiconductor teams. In other words, Jingwei Yage's existence as the originator of the FPGA field in China makes the scattered technical talents scattered among the rest of the domestic companies, becoming the Whampoa Military Academy of talents and technology.
Gowin Semiconductor: Gowin Semiconductor’s CEO Zhu Jinghui and SVP Song Ning are both from the Lattice team, especially Zhu Jinghui. After graduating from Tsinghua University, he worked for Lattice from 1996 to 2011. After seven generations of FPGA product development, he has won 11 US awards. Patent, 5 Chinese patents, and currently the technical person in charge of the major program of programmable devices under the Ministry of Science and Technology 863 Program. In addition to being a senior engineer at Lattice, Song Ning has also served as a senior engineer at Cadence. He is currently responsible for the entire FPGA software development of Gowin Semiconductor, and has unique experience in the simultaneous development of FPGA architecture, hardware design, and software R&D. Therefore, Synopsys provides the former SynplifyPro Gaoyun version of the software for Gaoyun, which is the only FPGA front-end software authorized by Synopsys in China.
Low-density non-volatile FPGAs have completely replaced traditional CPLDs and become the absolute main force in the low-density FPGA market, with an annual sales volume of approximately US$500 million. It can be said that Gowin Semiconductor's entry and benchmarking are Altera MAX V10 and Lattice XO2/3; 1 million-5 million volatile FPGA products, using TSMC 55nm based SRAM, can be compared with Xillinx Spartan and Altera Cyclone series PK. Recently, Gao Yun announced that it has successfully developed GW3AT-100, which is the first 28-nanometer mid-to-high-density FPGA in China, manufactured by TSMC.
Shanghai Anlu Technology: Shanghai Anlu Information Technology Co., Ltd. was established in 2011, headquartered in Zhangjiang, Pudong, and its founder is Wen Yubo. The company's founders and core team are composed of overseas senior technical management talents, product development backbones of foreign FPGA companies, and senior FPGA researchers in academia. More than 60% of the company's R&D team are masters or doctors from universities at home and abroad, such as Fudan University, Jiaotong University, UCLA, UIUC, etc., with strong research capabilities and design standards. Most of the core members have been engaged in technology research and management for more than decades in the world's top five FPGA companies and EDA companies, and have participated in the development of many world-leading FPGA chips and the best EDA development systems. At present, Anlu has formed a small-scale CPLD (Elf-300, Elf-650) to 2 million gate FPGA (EG-4, AL3-6, AL3-10, AL3-S10, EG-20, EG-D20) A series of devices, and a tens of millions of gate-level FPGA IP core (AL3-130) that has been actually used. After Huada Semiconductor became a shareholder, as the largest domestic EDA software vendor in China, Huada Jiutian and As an independent FPGA manufacturer Anlu Technology, both sides have great imagination for the cooperation space in EDA tools, including complementarity.
AGM: Shanghai Aogexin Microelectronics Co., Ltd. (AGM) was established in 2012. It is a leading domestic semiconductor integrated circuit design company based on programmable logic technology that provides SoC chips for the application market. It was founded by a team from a well-known programmable logic chip company in Silicon Valley, USA and a senior domestic engineering team. AGM started with the development of compilation software with independent property rights, compatible with the ecological chain of existing FPGA software. Seeing the smart phone outlet, AGM has also lost the opportunity to release an FPGA chip for smart phones and IoT. When helping customers gradually withdraw from the low-end market, seize the opportunity and seize the low-end market through price advantages. Through continuous product iteration and market expansion in recent years, AGM has quietly accumulated relatively stable customers, and its product line has begun to enrich, becoming a group of fast-developing dark horses in domestic FPGAs.
Tongchuang Guoxin: The application market of military industry institutes with the largest localization rate of substitution and the highest profit is basically monopolized by Tongchuang Guoxin (Shenzhen Guowei). Even the competitor, Chengdu Huawei, was basically silent. In addition, Tongchuang Guoxin’s popular civilian product Titan PGT30G has been mass-produced. This series of chip programmable logic devices adopts a completely independent proprietary system structure and mainstream advanced manufacturing technology, with DDR3 and PCIe interfaces, which are rare in China. Ten thousand gate-level FPGA. In addition, it adopts the advanced 40nm process of Taiwan Lianhua UMC foundry, which is leading in China. It is being tried out by ZTE and Fiberhome.
The development of FPGA industry depends on self-relianceThe development of domestic FPGAs can only rely on independence, which inevitably requires a long wait and dedicated care, but there is no better choice. The independent design of the chip is the lowest level guarantee for information security. This is why the basic chips (mobile phone chips, PC processors, etc.) related to information processing need to be self-made. At present, almost all FPGA technology and supply come from the United States, and technology powerhouses such as Europe and Japan have not mastered the core technology.
For China, the country's promotion of integrated circuit development has been upgraded to a national strategy. At the same time, the FPGA required for special application scenarios (military industry, missiles, aerospace) is related to national security and territorial security, and the requirements for chips will be more stringent, but this high-performance product is embargoed on China abroad. , This also contributed to the opportunity for domestic FPGA to satisfy itself. At present, domestically produced FPGAs are mainly used in fields such as military industry, communications, and aerospace.
In the civilian field, the domestic market has the largest demand for FPGAs. The largest customers of Xilinx and Altera are now in China. The communications market Huawei ZTE Fiberhome has taken over 60% of the country's volume. At the same time, the demand for high-speed parallel computing of artificial intelligence chips requires FPGA chips on the order of billions of dollars, and the domestic market for artificial intelligence high-speed chips is the fastest growing and most in demand. The development dividend of China's FPGA is that the demand market is large enough, and if there is demand, there must be corresponding products to support it. This is an opportunity for domestic manufacturers. At present, Tongfangguo Chip Co., Ltd. has already cooperated with Huawei ZTE, and wants to realize a part of domestic substitution.
Finally, from a technical point of view, we are not like 10 years ago that basically did not understand the core technology. With the continuous maturity and improvement of the domestic semiconductor industry chain and the continuous enhancement of chip design capabilities, we can independently design and tape out ARM-based mobile phone CPUs (Hisilicon Kirin, Datang Lianxin), and successfully commercialize them. This was 10 years ago I can't believe it. The technological precipitation and innovation capabilities we have accumulated in the past have enabled us to achieve a certain degree of self-supply in the specific application fields of FPGA (military industry, communications). In the future, it may also be similar to the use of CPU+FPGA in the cloud data center section. These application fields are places with high information sensitivity, and the use of self-designed chips can be more secure and controllable.
Later, I would like to talk about my exchanges with friends in the IC industry and some of my own feelings.
1. Why does everyone's pursuit of technology stay? Because for many people, after technology produces products and finances funds, the accumulation of technology and capital is achieved. The value generated by technology will be much slower than the value generated by capital in a short period of time, and most people will switch from technology to the capital market. This way over and over again, so you can see that after the technology reaches a certain platform, it will stand still for a long time, and the latecomers will follow the footsteps of the previous ones.
2. The so-called industrial support must not turn the cart before the horse and counter-revolution under the banner. The chip is not simply the need for capital support, because imperfect and unreasonable capital speculation will instead make the industry a feather after the tide recedes. Chips are more of a technology accumulation industry. All products and parameters require technical staff to constantly break through and think at work. For such a group of engineers, if the remuneration does not match the labor expenditure, the technical talents in the industry will leave. And loss is a natural thing. Therefore, the support of the industry is not only the support of capital to the company level, but also the technical staff in the grassroots.
3. It's easy to get started in the chip industry, but it will be difficult to move up. After the chip has reached a certain level, it is no longer a simple accumulation of technology that can be solved. It must have a deeper understanding of physics and mathematics, and deal with the basic difficulties from a high perspective. The demand for talents is very high.
4. On the market demand side, any kind of chip needs several generations of product changes from launch to application. Domestic chips start late, and product performance is far from abroad. Therefore, a dilemma will be formed: domestic applications do not use domestic products, and domestic product changes that cannot keep up will cause the company to close down or continue to update products. The foreign chip industry has been working for decades, and the pits and traps on the technological route have been filled by the first movers. However, the chip industry does not have a late-mover advantage. Where will the pits left by others continue to exist, and even need to avoid foreign countries Thousands of new pits for intellectual property protection applied by manufacturers in the process of exploration.
5. The localization of the chip industry has a long way to go, and we need to explore it slowly throughout the process. The eagerness for quick success and empty shouting of slogans only give speculators room to drill. Caring for the chip industry must also improve the welfare of technical personnel.
36V Bms,Bms For E-Bike,Bms For Scooter,10S 36V 15A Bms
HuiZhou Superpower Technology Co.,Ltd. , https://www.spchargers.com