At the "2012 Annual Conference of China IC Design Industry and Chongqing IC Innovation and Development Summit Forum" held, the upstream and downstream enterprises of China's IC design industry focused on issues related to industry status, opportunities and challenges, adjustment and innovation, cooperation and win-win, etc. In-depth discussion of the opportunities and challenges faced by the integrated circuit design industry, to explore ways to improve the basic capabilities of enterprises and enhance the overall capabilities of the industry. This newspaper selected their wonderful views to share with readers.
There is a myth in the pursuit of cutting-edge craftsmanship
The target market of China's IC industry is different from that of European and American companies, so mainland China will use appropriate technology to develop appropriate products and create a fast-growing market.
The capital investment of 28nm and below processes is huge, and the cycle from planning, design to mass production is very long. Therefore, it is necessary to carefully discuss with customers and carefully weigh various factors when expanding production capacity. As the market's demand for advanced processes continues to rise, TSMC is also accelerating the pace of development, with plans to 20nm in January next yearSoC trial production, 16nm FinFET trial production in November.
In the evolution planning of advanced technology, the next node of TSMC is 10nm. As for the key to process miniaturization – lithography technology, it is necessary to continuously improve efficiency and reduce costs. The 28nm process adopts HKMG technology, and the 16nm process adopts FinFET technology. At the 10nm node, new technical challenges will inevitably appear. Judging from the process progress of the world's major semiconductor companies, Intel continues to lead in CPU technology, Samsung leads in the field of memory, and TSMC's product line is more comprehensive, maintaining advanced in terms of integration and line width. It can be said that TSMC has achieved the industry's smallest wiring width in the back-end process.
TSMC's revenue this year is about 17 billion US dollars, and its business in mainland China accounts for about 5% of the company's total, which is growing rapidly. Customers in mainland China have already entered mass production of 40nm products from many companies, and more are expected next year, which shows that the demand for advanced technology in mainland China is very large. 28nm requires more resources, greater assurance and market demand to be able to establish a project, and mainland Chinese customers enter 28nm quickly, which shows the market grasping ability of mainland Chinese IC design companies. The target market of Chinese IC design companies is different from that of European and American companies that only focus on high-function requirements. Therefore, mainland China will use appropriate technology to develop appropriate products and create a huge and fast-growing market.
There is currently a myth in the foundry industry in mainland China, which is to strive for cutting-edge technology, while TSMC is mature and advanced technology, forming a broad and comprehensive process for SoC, MEMS, high voltage, embedded memory, etc. platform. As the industry needs more and more product integration, performance and power consumption requirements are getting stronger and stronger, TSMC's broad and deep process platform can best meet customer needs.
The differentiation of the IC industry in mainland China focuses on the structure
The design must consider factors such as process, IP, packaging and testing, involving algorithm design, software development, circuit board design and other links, which accordingly pose a huge challenge to EDA suppliers.
The semiconductor industry is undergoing dramatic changes, with embedded processors and mixed-signal SoC product designs becoming increasingly complex, customers are looking to the EDA industry for the fastest path to design creation, integration and optimization. With FinFET, 3DAdvanced manufacturing processes represented by IC are constantly emerging, and new requirements are also put forward for EDA tools, from 28nm to 20nm
EDA tools have changed a lot, the requirements for EDA tools have changed even more with FinFET processes, and each foundry process offers a different set of complex conditions involving speed, power consumption, and cost. Factors such as process, IP, packaging and testing must be considered in the design, so that the logical design and the physical design can be interoperable, which also involves algorithm design, software development, circuit board design and other links, and accordingly poses a huge challenge to EDA suppliers.
Cadence is actively innovating to meet these challenges, accelerating the development of next-generation design tools, currently 20nm physical design and verification tools are maturing. In addition, Cadence is developing 16nm tools to better support foundries. In terms of analog circuits and fully customized digital circuits, Virtuoso-ADE will continue to promote new methods of "analog design constraints" to reduce design errors, and continue to strengthen PDK functions to improve design efficiency.
At the same time, Cadence is also committed to the cooperation of vertical integration and strengthens the cooperative development with ARM and TSMC. In terms of cooperation, ARM considers how to develop the potential of EDA tools, and Cadence considers the efficiency and optimization of the tools, and provides one-to-one personal service. In June of this year, Cadence cooperated with TSMC to produce the first 3D
IC experimental chip. In October, Cadence cooperated with IBM to realize the first ARM Cortex-M0 processor based on IBM 14nm and using its FinFET process design.
In the mobile age, ARM stays ahead by building a leading ecosystem. From the perspective of future development, there will be great development in cloud computing, servers, etc. Facebook or Google will define indicators such as server power consumption, and the X86 architecture will also have a place. In the past, the needs and progress of the terminal drove the progress of the chip industry, but now the semiconductor industry is dominated by software manufacturers such as Google, because the processor chip must support the Google operating system.
Although the IC design industry in mainland China faces the problem of IP core homogeneity, in terms of architecture, such as the connection between core and core, the connection with memory, whether the clock is 4-phase or 8-phase, how to wire and how to use EDA tools skillfully to improve development Efficiency and success rate, etc. can be differentiated, but it is not impossible to differentiate with the same IP and the same process.
Only by doing both depth and breadth can we meet the IC needs of mainland China
The top ten most concerned markets in the IC industry in mainland China are communications, smart cards, computing, multimedia, simulation, navigation, etc. EDA tools need to provide sufficient technical support.
Synopsys has recently made a series of acquisitions, such as the acquisition of EVE, which adds high-performance, high-capacity simulation acceleration capabilities to its verification platform. In addition, it also spent $400 million to acquire Siyuan Technology to further strengthen its powerful system-level chip error correction capabilities. Synopsys' market strategy is to fully integrate with China's IC industry and provide full system-level solutions for EDA tools, enabling customers to continuously respond to current complex IC design needs.
As the world's second largest IP supplier, Synopsys is also continuously strengthening its strength in IP cores. Now that the IC design industry has entered the SoC era, more than 90% of products require IP authorization, or integrated design through its own IP. Synopsys can provide complete solutions such as interface IP in terms of IP. At present, the IC design industry in mainland China is catching up with the advanced level of foreign countries, and the foreign IC design industry has developed for many years, with the accumulation of endogenous IP. The IC design industry in mainland China needs mature digital and analog IP to achieve leapfrog development. The IC design industry in mainland China is constantly expanding in terms of depth and breadth. In terms of depth, IC design in mainland China ranges from 32nm/28nm to 22nm; in terms of breadth, the process distribution covers a wide range, which requires EDA tool vendors to be able to provide targeted Sexual tool service and IP. The ten most concerned markets in the IC design industry in mainland China are communications, smart cards, computing, multimedia, analog, consumer, etc. Three-dimensional strategies are needed to meet diversified needs, because customer needs not only need depth, but also breadth, and it is necessary to provide a wide enough technology.
In 2012, Synopsys' revenue in EDA tools exceeded 1.7 billion US dollars. In the past few years, the average annual growth rate of IP has been 20% to 30%. IP is not easy to do well, but Xinsi insists on investing, and it only started to make profits in the first two years. Although EDA tools are more profitable than IP, in terms of the depth of cooperation with customers, IP providers can go deeper into customer projects than EDA tool vendors. At present, the IC design industry in mainland China is rapidly following up on advanced processes, and more companies are entering the 22nm and 20nm stages. The chip design is becoming more and more complex, and the requirements for verification are getting higher and higher, and the collaborative design of software and hardware is even more necessary. Synopsys is also stepping up efforts to support the development of advanced technology, and provide more comprehensive and systematic services in front-end design synthesis and back-end layout and wiring.
Hardware emulation accelerator has more advantages
The industry adopts three simulation acceleration methods: software simulation, FPGA simulation and hardware simulation. The hardware simulation accelerator absorbs the advantages of the former two and achieves a balance between efficiency and debugging.
The cost reduction brought by advanced technology is very tempting, and the industry is willing to adopt advanced technology, but the development of advanced technology often faces bottlenecks, so other means are needed to meet the cost-driven requirements. Mentor's strategy is not only to bundle advanced technology, but also to go beyond the role of EDA tool supplier to provide customers with more value-added services.
As the complexity of multi-core continues to increase, it brings huge challenges to software developers, and it is necessary to carry out software and hardware co-design in the prototyping stage. The difference between Mentor and its competitors is that software design can be started at the prototyping stage and implemented on the same platform until the verification stage, enabling rapid software development. As SoC designs become more and more complex, more and more IPs are required. If each IP has a bug, it will lead to an increase in SoC defects. Mentor's software-hardware co-design approach accelerates IP verification.
Mentor Graphics not only provides efficient tools in the prototyping stage, but also has strong tools in the physical verification process. In the physical verification stage, the foundry will provide corresponding rules, and it must be 100% in line with the plan before production. 3DThe parameters involved in advanced processes such as IC are different, such as parasitic parameter extraction, etc. Mentor is an early manufacturer involved in the development of 3D IC tools in the industry.
From the perspective of simulation acceleration methods, the industry mainly adopts software simulation, FPGA simulation and hardware simulation acceleration. The disadvantage of software simulation is that it is slow, but debugging is more convenient. As the complexity of the chip increases, the speed of software simulation will not keep up. FPGA simulation speed is the fastest, but there is no way to debug, only to analyze the problem, but can not analyze where the problem is. Generally, it is known that there is a problem through FPGA simulation, and then Debug is performed on the hardware emulator, and the two complement each other. The hardware emulation accelerator draws on the advantages of both, and can achieve a balance between efficiency and Debug.
The advantages of Mentor's latest hardware emulator are: First, it has a large capacity, which can support up to 200 million gates, and parallelism can reach one billion gates; second, low energy consumption; Verify hardware and now software. For customers, hardware accelerators will be selected when the design complexity and customer scale reach a certain stage. Mentor will also focus on promoting its hardware emulators in mainland China.
Profitability is a major test in the next stage of the IC industry in mainland China
The pursuit of the IC industry in mainland China is not limited to advanced manufacturing processes. Mainland IC companies must fight against homogeneity in terms of positioning, entry point and timing.
The characteristics of mainland China's electronics industry are fast, large, low, and short: fast means rapid market changes, more manufacturers means more manufacturers, low means low prices, and short means short market windows. Therefore, China's IC industry, which is in the upper reaches of the electronics industry, has developed a set of effective tactics: one is to specialize, concentrate product development resources on pioneer products, and specialize in the mainland market; Close localization support; The third is customer orientation, product specification positioning is accurate and not exceeding the standard, and the first time to respond to market changes. Therefore, even in the face of fierce competition from major international manufacturers, the mainland IC design industry can still play its home field advantage, gain market share, and maintain growth.
Looking at the next 2 to 3 years, the international manufacturers have already raised the competition in the IC market in mainland China to the system level. At the system level, in addition to the design process, the competition level also focuses on chip packaging and testing processes, parts reduction planning and even public version design, etc., and the room for cost optimization has increased a lot. Therefore, if the mainland IC design industry still sticks to the silicon level and blindly pursues the most advanced design process, then the competitive options and flexibility may be limited.
Referring to the industry development experience in other regions, I believe that the IC design industry in mainland China will go through three stages of development: first, the first stage of seizing the market and entering the market more quickly; after occupying a place, it will enter the second stage of improving profitability; The third stage is the stage of building sustainable operation capability. In recent years, the number of IC design manufacturers in mainland China has increased rapidly, so most of the current distribution may be in the first stage. It is believed that a considerable number of companies will enter the second or third stage in the next few years. IC design companies in mainland China should also fight against homogenization in terms of positioning, entry point and timing.
As a foundry, UMC does not believe that the pursuit of the IC industry in mainland China is limited to advanced manufacturing processes. At present, no more than 20% of IC designs using 40/28nm technology are used, and 80% of the designs use mature technology. Take the display driver IC as an example, from HVGA, WVGA/qHD, WXGA/HD, to FullHD and other display resolutions, and the process technology of driver ICs are based on bandwidth and memory requirements, and the high profit points are different (0.16μm, 0.11μm, 80nm, and 55nm, etc.), but these products all exist in the market at the same time. Therefore, the IC design industry can actually pursue the highest cost performance between the lowest cost and the highest process, taking into account market competitiveness and profitability. UMC hopes to actively respond to the comprehensive needs of IC design customers in mainland China for advanced and mature processes. We believe that the 28nm process will continue for a long time.
Design vendors need to focus on architecture and software
The FabLite and DesignLite models in mainland China are prevalent, the chip is becoming more and more a SoC, and software has become more important, requiring designers to focus on architecture and software.
Judging from the development in recent years, the R&D investment in the global IC design industry has an average annual growth rate of 2.2% to 2.5%, but the average annual growth rate of revenue is only 1.5% to 1.6%. This is a question worthy of consideration by the industry, because OEM's R&D investment is about 5% to 8% of revenue every year, while global IC giants are generally 16% to 17%.
From the perspective of the development characteristics of the IC design industry, one is the prevalence of FabLite and DesignLite models in mainland China, and the other is that the chip is increasingly becoming a SoC, and software has become more important, requiring designers to focus on architecture and software.
The most striking thing in the Chinese mainland market is the hot sale of smart terminals. At present, the BOM cost of smartphone terminals in mainland China is about 11% lower than that of foreign counterparts. We know that AP (application processor) development is more difficult than BP (baseband). For IC designers, if there is only a processor and no baseband chip, development will be more difficult. Recently, the 4-core 5-mode AP+BP chip based on 28nm and supporting LTE has been launched. It is believed that the LTE market will develop after 2015. In terms of tablet PCs, the combination of different chip manufacturers is required. For example, processor and baseband chip manufacturers can join forces to carry out SiP packaging to enhance competitiveness. However, IC design companies in mainland China are reluctant to join forces, and BP is more difficult to develop. Companies that only develop APs need to pay attention to this. In terms of smart TV, the development of smart Android TV boxes has attracted wide attention, and the human-machine interface will be further improved in terms of voice, gesture recognition, and user experience.
There is still a huge blue ocean market, such as the LEP (plasma light source) lighting chip, which uses a high-voltage process. The chip developed by VeriSilicon in cooperation with customers has been successfully taped out in three months. There are also medical electronics, such as injections for children, it is usually difficult to find blood vessels, but the micro-projection technology can identify the blood vessels at once. In addition, automotive electronics are also worthy of attention.
At present, VeriSilicon focuses on providing design services in application markets such as multimedia, voice and wireless communications, providing market-leading ZSP licensable cores and platforms, industry-standard semiconductor IP, and upgradeable ASIC turnkey services, which can accelerate customer ASIC design ( From preliminary specifications to silicon products), one-time tape-out success on time and on-spec in silicon products, and mass production of customer silicon products. VeriSilicon is not just an IP provider, VeriSilicon is a manufacturer that provides customers with customized services for comprehensive semiconductor solutions.
It takes at least five years to do IP
The IP ecology is not easy to do. It takes at least 5 years for IP to go from R&D to profitability. Only companies that are tough enough can survive.
At present, the number of high-end SoC chips with relatively complex designs by IC companies in mainland China is increasing significantly, which requires them to use high-quality, stable and commercialized processor IP. Taisilida is continuously increasing its investment in the Chinese market. The IP ecology is not easy to do. It takes at least 5 years for IP to go from R&D to profitability. Only companies that are tough enough can survive.
IP vendors need to be able to do things that others don't want to do, and they need to be able to do things that others can't. For example, to support the transmission of 1080P video by the new mobile standard LTE, a completely different architecture of the DSP core is required. Other manufacturers are reluctant to set foot in considering the technical difficulty and profit, but the DSP core solution of Taisilida can already support Such high definition transmission. In addition, for the DSP core of Taisilida, the customer can provide the software, and we will provide it to the customer after customization and verification, and the customer can save the verification process.
For SoC design companies in mainland China, they cannot only consider IP, but also control the industrial chain links such as design and manufacturing in their own hands. As can be seen from the recent acquisition of MIPS, it is difficult for MIPS to survive after ARM launches 64-bit cores. MIPS has not established its own complete ecosystem, and both Apple and Microsoft support ARM, so the development of MIPS is affected. I don't know why the Chinese mainland IC design companies did not participate in the MIPS acquisition. The Chinese mainland IC design companies should cooperate with the government to buy MIPS, otherwise it will be difficult for mainland Chinese companies to develop the value chain system. IC design companies in mainland China should improve their cognition and design capabilities.
Taisilida has been deeply cultivating in the field of DSP core, mainly focusing on communication, audio and video, and image processing applications. In 2012, 2 billion chips are expected to use our DSP cores. With the start of the LTE market, due to the large amount of data in LTE, DSPs with higher performance, lower power consumption, and design flexibility are generally required.IP to do the baseband. Tasileda has developed a new DSP for LTE-Advanced (LTE-A)The IP core can support not only LTE-A, but also existing communication methods including LTE and HSPA+.
Responsible editor: tzh