Internet of Things, artificial intelligence, AR, cloud computing, etc. are subverting our world. On November 8th, the most influential industry leaders and technology giants in the semiconductor industry gathered in Shenzhen, China, to discuss the impact and opportunities of artificial intelligence and the Internet of Things at the global CEO summit held by ASPENCORE, the world's largest technology information group. Hot topics such as roadmaps, networking standards, and forward-looking changes to the world's future technologies.
Dai Weimin: From the Internet of Everything to the Reality of All ThingsIn a keynote speech, Dr. Wei Weimin, Chairman and President of VeriSilicon, said that China’s current car ownership is very high, which is very conducive to the era of electric vehicles. Although there may be a long way to go in the traditional car, in the case of electric cars and smart cars, we can completely achieve cornering overtaking.
It is well known that L2 and L3 self-driving cars require people to respond quickly in the event of an accident, but in more advanced autonomous driving (L3, L4), people can basically not participate. Although companies now have L4 and L5 autopilots, they are limited to certain scenarios, such as warehouses, agriculture and cargo handling, airport shuttle buses, mines and other fixed lines.
“To achieve driverlessness, vehicles need sensors, 5G communications, and many other devices that need to be well connected to the surrounding cars and smart cities, but we can’t put all the devices in the car, and with the electronics occupying the car. The proportion of total cost is increasing year by year, and cost is also a very important consideration, which makes our industry development direction need to be adjusted.” Dr. Dai Weimin talked about AI engine, NB-IOT, intelligent video, FinFET/FD-SOI process, etc. The elements, and remind the industry that there may be only special scenes in the next 2-3 years that require training for the driverless system. These systems are still programmable, and the cost and energy consumption may increase significantly.
At the same time, he also mentioned the concept of "molar pressure." With 28nm as the boundary, everything is very good before 28nm, but everything changes after 28nm, especially the cost, the cost of each node below 28nm will be significantly improved. Therefore, he hopes that the industry can walk on two legs in the face of different applications, both the FD-SOI process and the FinFET process. Especially the former can help us develop more and better products, which will be very popular for many years to come.
Finally, Dai Weimin appealed to the industry that “when proposing products, ideas and intellectual property that promote R&D and innovation, there must be beneficial tools to protect it, which is very important for future sustainable development.”
Xilinx: Designing Next-Generation Architecture for Adaptive IntelligenceXilinx CEO Victor Peng pointed out at the summit that "we are undergoing a transition from the mainframe era, the PC era to the adaptive intelligence era", the data explosion, artificial intelligence and the calculation of the post-Moore's Law era, will To become the three most important trends in the future, Xilinx is ready for the arrival of this era.
In such an era, infrastructure and architecture will change. For example, from 2000 to 2015, the number of wireless standard updates and AI papers published has grown very rapidly. More and more neural network engines have been added to AI applications, which makes it impossible for previous CPU/GPU/ASIC chips to respond directly. Therefore, innovation must be made at the architectural level.
“Data Center Acceleration” was listed by Victor Peng as a priority strategy for the company's development. Xilinx is strengthening its collaboration with key data center customers, ecosystem partners and software application developers to further drive innovation and deployment in computing acceleration, compute storage and network acceleration. The data center is an area of rapid adoption technology that focuses on enabling customers to quickly benefit from the performance and unit power performance benefits that Xilinx technology brings to a variety of applications, including artificial intelligence inference, Video and image processing, genomics and other applications.
In order to better meet the needs of users, Xilinx also launched a new product category - Adaptive Compute Acceleration Platform (ACAP), its first product Versal uses TSMC 7nm FinFET process technology, currently plans a total of six Series: AI core series, AI Edge series, AI RF series, basic series, flagship series and HBM series, built-in engine including scalar processing engine, flexible hardware engine, intelligent engine, advanced protocol engine, etc., can handle different workloads .
ST: Semiconductors lead the Internet of Things from vision to reality“In the Internet of Things era, our vision is to enable billions of smart IoT devices to be interconnected, enabling almost any system to leverage the Internet and cloud computing ecosystem to innovate, making IoT products smarter and more aware. Strong.” Jean-Marc Chery, President and CEO of STMicroelectronics, divides the Internet of Things into smart phones, smart homes/smart cities, smart factories, and smart driving, indicating that semiconductor technology will move from vision to reality in the Internet of Things. Play an important role in the process.
Chery pointed out that smart products, smart homes/smart cities, smart factories and smart driving will all achieve significant growth from 2017 to 2021. The Internet of Things has made the world smarter. It not only brings convenience and efficiency, it makes us safer when using devices, and it makes our interaction with the world stronger and stronger.
In terms of smart driving, the automotive industry will make driving safer, greener and more intelligent by continuously applying IoT products. Simply put, the Internet of Vehicles can achieve more services, the networked car will set off a change in vehicle safety, V2X (vehicle and outside communication) will achieve safer, faster travel, and environmental protection, SiC will be in the field of electric vehicles To an important role.
In the smart industry, whether it is China Manufacturing 2025, German Industry 4.0, and the US Industrial Internet of Things (IIoT), they all share a common theme, even if factories and machines become smarter and more conscious, and sensors are for them. Give these abilities. The combination of sensors and locally distributed intelligence provides very powerful features that facilitate human-machine collaboration. As a result, the role of workers will change dramatically, and factories will become more connected, enabling better data collection and uploading it to the cloud. In this process, sensing, driving, processing, security, connectivity, regulation, protection, motor control, and power management all bring opportunities to the semiconductor industry.
Another great opportunity comes from 5G. 5G networks offer significant improvements in data rates, latency, and connection quality, which are critical for car networking, industrial robots and drones, and NB-IoT applications.
“Of course, what I just mentioned is largely a concept and an idea that needs to be chosen with the right technology and product portfolio. In order to properly manufacture the product, the correct production process, such as embedded NVM, microprocessor, is required. "Microcontrollers, which processes should be used for image processing, are all issues that need to be addressed," said Jean-Marc Chery.
Silicon Labs: Simplify the Internet of Things with a platform designed for end nodes"If we deploy billions of IoT devices, we will encounter a series of problems and challenges. For example, how to better communicate and exchange information between machines, how to balance in cost and power consumption, etc." Silicon Labs CEO Tyson Tuttle said that in this process, we need to consider software and hardware, but the software is more important because it defines the characteristics and attributes of the device, and people must consider the different protocols that the device faces when it connects to the device. Stacks, applications, middleware and wireless updates, interface interactions and more.
He also introduced Silicon Labs' IoT "integrated software + hardware platform" - Gecko. The platform not only meets the needs of customers on the hardware, but also cooperates with the software. “For the design of the software, we have an expert level engineer who has a good understanding of the relevant hardware, design a better software platform, and establish a complete product portfolio.” Tyson Tuttle said, “At the same time, we are quite similar in software. Be cautious, we will ensure the interconnection of products, whether in the WiFi environment or other network environment, can achieve interconnection with different devices. Of course, we will also ensure good maintenance and maintenance of software and hardware."
IoT devices will also become more and more complex, and the implementation of many functions requires the combination of different software, in order to achieve the intended purpose, meet the needs of customers, and be recognized by customers. “So we have to have good tools and integrate all the tools to ensure that our customers can connect and use the equipment. At the same time, we have to make these tasks as simple as possible. This requires our engineers. I know a lot about our tools, know which tools to change, which tools to upgrade, etc.," said Tyson Tuttle.
Nowadays, more and more companies are committed to developing more software and hardware to meet the needs of the Internet of Things, which makes the rapid development of the Internet of Things possible. “The hardware and software of the Internet of Things is not a simple combination, but a highly optimized and tightly integrated platform to address the challenges of the Internet of Things.”
Lu Chaoqun: Heterogeneous integration will push the semiconductor industry to fight for another 30 years!Dr. Lu Chaoqun, CEO, Chairman and Founder of Etron Tech, and Director of the Taiwan Semiconductor Industry Association, presented at the meeting "Emerging Index Economic Growth: New Heterogeneous Integrated Systems of AI Addition ICs Power's speech.
"Artificial intelligence did not happen 10 years ago, but today's emergence is because humans have such a demand. In all kinds of applications of AI, semiconductors are playing the role of intelligent core. But many predictions show that Moore's Law is related to semiconductor kinetic energy. It may end in 2025.” Dr. Lu said that the first era of semiconductors is silicon 1.0, which is to increase the number of transistors by one time in the unit node area every two years. The density of each node is 0.7 times, 30 micrometers to 28 nanometers. 20 nodes, this is the hero who promoted the positive growth of the IC industry in the past decade: Index Moore's Law.
However, the silicon 1.0 era ended at the 28-nm node, from below the 28-nm node, to today's semiconductor technology to the 7-nanometer generation, this time is called the silicon 2.0 generation. Dr. Lu believes that in the Silicon Generation 2.0 to promote the Effective Moore's Law Economy (EME) to maintain economic investment efficiency, mainly rely on Area Scaling.
In the Silicon 3.0 era, Volumetric Scaling relies on the area of the bottom plus the encapsulation technique to pull out a 3D space and then back into a unit area. Through this approach, the spirit of Moore's Law has been renewed. In a very small package, multi-layer stacking, 3D NAND, which is currently in high demand in the market, uses this technology.
In the era of Silicon 4.0, Dr. Lu believes that the Heterogeneous Integration Design Architecture (HIDAS) will greatly promote IC innovation. For example, the system-on-chip will adopt a new architecture and use a variety of spatial layouts for various types of dies. He promotes this concept because the semiconductor industry must at least be obsessed with the miniaturization of process nodes; in order to achieve growth momentum, the industry must be different. Innovation by heterogeneous integration of technology.
The Silicon 4.0 era is characterized by the integration of peripheral MEMS, lenses, sensors, biosensors, etc. into Nano Systems products that are not semiconductors but made of silicon, and no longer use linear miniaturization to create value. Instead, the value of the terminal product is magnified, which is a concept of “reduction of the value of the function X”.
