Dedicated to helping readers in the broad IT field, master more professional, more practical knowledge and skills, and quickly enhance the market competitiveness.
In 2021, the semiconductor industry has always been the focus of attention. The semiconductor industry continues to grow in 2022, even as the global economy continues to be overshadowed by the COVID-19 pandemic. Behind the chip shortage, some semiconductor companies are happy and some are worried. In the face of the piled order demand and the growing market scale, what specific plans or actions do the semiconductor companies have in 2022?
Number 1: 3nm process for mass production, 2nm on the schedule
- TSMC: 3nm will be mass-produced in the second half of 2022. Fab 20, built for 2nm production, is also scheduled to start construction in 2022 and is planned for mass production in 2024.
- Samsung: The 3nm first-generation process will be mass-produced in the first half of 2022, and the second generation is expected in 2023; 2nm is planned for mass production in 2025;
- Intel: Plans to invest $20 billion in the first phase to build a 3nm wafer fab and start mass production in 2025. 20A and 18A (Emi-class) processes will be mass-produced in 2025.
Number 2: DDR5 standard memory for mass production and commercial
Demand for super-scale data centers continues to be strong around the world, with industry giants such as Samsung, SK Hynix, and Micron unveiling a Roadmap for mass production and commercial use of DDR5-standard memory products.
Number 3: The DPU market will explode
Data Processing Unit (DPU) is dedicated to solving the problems such as network protocol Processing, Data security, and algorithm acceleration that the CPU cannot do well and the GPU cannot do well. DPU is a dedicated computing chip that meets the requirements of high bandwidth and low latency. On the DPU R&D track, Shuangying (Nvidia and Intel) technology leads the way. Domestic DPU accumulates and develops steadily: Zhongkeyu Number, Xinqiyuan, Xingyun Zhilian, and other companies are committed to establishing domestic brands.
Number 4: Storage and computing integration over the “storage wall” and “power consumption wall”
Von Neumann architecture chip and system is a kind of storage and computing separation architecture, there are “memory wall” and “power wall” bottlenecks. Processing in-memory (PIM) technology is designed to break through the bottleneck. In addition to Samsung and Micron, many enterprises such as Myhtic, Zhizun, Jiutianrui, Smartchip, Hmo Intelligence, Apple, and Xinyuan Semiconductor have engaged in research and development. It is estimated that it will take at least 5 to 10 years for the application to mature.
Number 5: 5G construction will focus on independent networking and millimeter wave
China has built the world’s largest and most technologically advanced 5G independent network, setting a global example. 5G millimeter wave technology is an important addition to the sub-6ghz band and is most commonly used in “hotspot coverage”, including enterprise indoor deployment, indoor/outdoor venues, transportation hubs, fixed wireless access, industrial Internet of Things, and other fields. Both Qualcomm and MediaTek plan to launch relevant chip products in 2022.
Number 6: EDA tools are starting to use AI to design chips
International EDA giant Cadence, Synopsys, Siemens and domestic EDA enterprises Guoweissier core, core Hua Zhang, Hongxin micro and nano, the introduction of AI technology, Google, Amazon, Alibaba, Tencent, Baidu, ByteDance, and other companies that make Fabless cores for Internet crossover have also successfully applied AI to chip design by using their established research institutes or artificial intelligence laboratories.
Number 7: Matter is driving the convergence of connected standards for the Internet of Things and the smart home
The Matter standard is already supported by Internet giants (Amazon, Apple, and Google), chip manufacturers (Silicon Labs, NXP, and ESPRESSIF), Internet of Things and smart home device vendors (Ikea, Huawei, and OPPO), and smart home platforms (Doodle and Wulian), It is expected to be publicly released in 2022 and rapidly grow and gain popularity worldwide, becoming a unified interconnection standard for the Internet of Things and smart homes.
Number 8: RISC-V architecture processors enter the field of high-performance computing applications
Global chip giants, Fabless startups, and IP developers are all trying to bring RISC-V instruction set architecture to high-performance computing applications such as data centers, AI, 5G, and servers. RISC-V has the potential to compete with x86 and Arm.
SiFive’s Performance series P550 kernel has been selected by Intel’s 7nm Horse Creek platform; Startup Esperanto unveiled ET-SOC-1, an AI accelerator chip that integrates more than 1,000 RISC-V cores and is designed for data center AI reasoning; Ali’s Xuantie 910 RISC-V processor points to 5G, artificial intelligence, network communication, and autonomous driving and other fields; RISC Pygmy is a high-performance low-power RISC-V multi-core heterogeneous AI SOC for AIOT. It includes one low-power ORV32 core, four high-performance ORV64 cores, and hardware accelerated AI vector units independently designed by RISC, which can efficiently and flexibly handle a variety of AI tasks in terminal devices. The first generation of RISC-V single board computer “Fang · Xingguang VisionFive” has been put on sale; Another high-performance RISC-V SoC “Fang · Jinghong 7110 (JH7110)” is scheduled for mass production in Q2 2022;
Number 9: Advanced Packaging Technology Becomes “New Moore’s Law”
In addition to FinFET, which is evolving into GAA process technology, Yole classifies Flip-Chip, Fan-in WLP, 3D stacking, Fan-out, and Embedded Die technologies as “advanced packaging”. Substrate-level SiP packaging technology and Chiplet chip design technology will be the two paths that drive the development of “advanced packaging”. Innovation requires synergy from the whole industry chain. Wafer-level SiP packaging technology has to work with TSMC, Intel, Samsung, and other foundry factories, and at present, foundry factories are deeply involved in certain processes of substrate-level SiP and Chiplet. In addition to the packaging and testing manufacturers such as Sun and Moon, SiP, Anjin, Changdian, Tom Tomo Microelectronics, Huatian Technology, YongSi Electronics, JingFang Technology, and Huazin Semiconductor, there are also new entrants such as Huanxu Electronics, Shennan Electric Road (Tianxin Interconnect), Geer, Ovation, Quti, Lixin Precision, Foxconn and Wintel Technology (Anshi Semiconductor).
Number 10: “Software-defined cars” are becoming possible
The future of automotive electronic and electrical architecture is a three-domain control architecture, i.e. smart cabin, smart computing, and smart driving. In addition to higher peak arithmetic power, the car brain chip should also take into account the overall balance, heterogeneous architecture, lower power consumption, more security, different data type processing, thermal management, OTA, and many other aspects. More and more car manufacturers will develop their own chips, leaving enough redundancy space to cope with the changing architecture of the car and AI algorithms, in order to truly realize the “software-defined car”. Nvidia, Qualcomm, Mobileye, Amba, NXP, Texas Instruments, ON Semiconductor, Renesas, STMicroelectronics, Infineon, CoreChip, CorePower, Heisei, BYD, CoreWant, Jayfa, Horizon, Samsung, and other groups of masters and intensified competition, automotive chips into the era of warring states.