A few days ago, Gartner listed edge computing as a top ten strategic technology for 2020. The core logic of edge computing’s information processing and content collection and delivery is to keep traffic localized and distributed to reduce latency and bring critical applications and services closer to the people and devices that use them. By 2023, the number of smart devices at the network edge could be more than 20 times that of traditional IT.
In a nutshell, the main reasons for the rise of edge computing include the following three points: first, it can reduce the total cost of ownership, not always having cloud access, thereby reducing local costs; second, it can protect digital privacy and improve security; third, It can reduce application latency, better support real-time applications, and enhance local experience quality.
Based on this, a lot of computing power in traditional data centers is shifting to the edge side. Can traditional MPUs meet the characteristics of emerging ecological applications? Are there any new requirements in terms of optimizations and improved performance?
In this regard, Mr. Ron Martino, senior vice president and general manager of the edge processing business unit of NXP, believes that edge computing has a very wide range of possibilities, whether it is its performance or capability, its capability comes from industrial interconnection, TSN or scaling computing to meet standards workload. A scalable solution has higher value because it can fully tap the capabilities of edge computing, from the terminal all the way to the highly integrated terminal, such as uploading data and interfaces to the cloud, if more local capabilities are required, such as in To perform AI learning locally, that is, to complete the learning action on the edge side without using the cloud, this capacity needs to be improved and a very wide range of devices can meet the capacity requirements of the software while taking into account the efficiency of the platform. In this way to provide solutions, which is why many customers want to cooperate with NXP.
In addition, security is a very important consideration. Although security has been very important in many applications before, it has never been a decisive factor in purchasing behavior. Now with the increase of access points on the edge side, a lot of critical data is generated and its value is getting higher and higher. Security and continuous integration of functions have become extremely important factors for everyone to consider, and this is also embedded computing. Key reasons for continuous evolution. As a result, more and more business models are evaluating how to improve security as a very important consideration when evaluating long-term reliability.
Data breaches in many industries can impact a customer’s brand reputation, including privacy breaches, theft of critical data, and unauthorized intrusions and access. Therefore, government regulatory authorities and industry safety certification play a very important role.
Ron Martino said: “NXP has a very long history in security, with outstanding security in banking, e-government and financial transactions. In addition, we leverage our capabilities to provide a wide range of security products that are scalable at the edge, while Both discrete and integrated products are available. In addition, we now offer certified EDGELOCK ASSURNACE, a high level of safety certification and service for devices with very high safety standards and NXP’s partners. NXP’s safety performance has now reached PSA Level2, CC and other high standards.
In recent years, NXP has been vigorously promoting EdgeVerse, mainly for edge computing, especially for the Chinese market. According to Ron Martino, the product can meet six of the seven key areas of digital network construction proposed by China: first, artificial intelligence, which can provide easy-to-use AI solutions for machine learning deployment; second, industrial Internet, which can provide very Low-latency industrialized protocol support; third, intercity transportation, which can provide new human-computer interaction functions in this field; fourth, 5G network, which can provide Layerscape software-defined baseband and scalable processors and controllers; The data center provides data shunting with excellent power consumption performance for IoT work in the data center; the sixth is the charging of new energy vehicles, which uses NXP’s i.MX human-computer interaction capabilities and cross-border MCUs.
In addition, NXP has a full set of software to support edge computing, including vertical optimization of network, industrial and IoT edge computing, support for multi-system processing, including tools and distributed BSP, the company has also developed many middleware to help customers quickly integrate product to market.
At the end of 2019, NXP completed the acquisition of Marvell’s Wi-Fi business, giving the company a better product portfolio in wireless connectivity, such as BLE, Zigbee and Thread that can support narrowband, which can provide a better edge calculation plan. In addition, the company’s MCU Xpresso SDK has better connectivity to the platform, and many firmware and drivers are easier to use.
From the perspective of application fields, NXP’s edge processing applications are mainly divided into three categories: one is to support 5G local network and data integration; the second is the industrial edge, including factory automation, infrastructure, transportation, medical care, etc.; the third is the Internet of Things Edge, including smart home, consumer and wearable fields.
IoT and Smart Home
The earliest IoT device was a toaster exhibited in 1990. At that time, the toaster was connected to a computer and could be controlled remotely through the computer, but the concept of IoT was only proposed in 1999.
Although different IoT devices have some things in common, for example, IoT devices use some technologies to monitor and control Electronic and electrical components, but different IoT applications have different requirements, so different numbers and diversifications are required. MCU and MPU control and support, there is no unified solution. Yu Xiujie, vice president of NXP’s edge processing business unit and general manager of the IoT business line, said: “Whether it is GPU or FPGA, based on these technologies, they can find their own areas of application and expertise, and it also depends on the market’s demand for application power consumption. Tolerance. Sometimes the content that needs to be optimized is different, depending on whether it is voice or video, whether to do learning or inference on the edge or the cloud. Therefore, there are many considerations, and not only from TOPS from the perspective of the market, because the market has different tolerances for costs.”
“Now, some customers are asking if it is possible to do free AI and machine learning, which actually refers to using CPU to do machine learning. Now our eIQ tool can deploy a model similar to this, for example, in home scenarios, because people It is an interactive object, so there is not high demand for speculation, and there is no need to buy acceleration products in machine learning to reduce costs. We are constantly doing this fine-tuning with partners and customers, especially in the early stages of development, there will be a lot of Iteration and strategic fine-tuning. I believe there will still be a lot of room for innovation in the market,” said Yu Xiujie.
NXP’s product portfolio is very rich, ranging from MCUs based on Arm architecture, to MCUs with connectivity, high integration, high performance and low power consumption, and even very large, high performance MPUs.
In the home environment, NXP’s products can support the better combination of many components, making the use of the device easier. Taking the washing machine as an example, it integrates various technologies of NXP, including MCU, MPU, NFC, Wi-Fi, etc. The company is able to provide a reference platform to help customers launch new products. Taking the i.MX RT106 crossover MCU as an example, NXP has further enhanced local voice control, face recognition and Alexa voice services. The company can share more software and data to help IoT customers jump-start product development.
5G and the data center
If you want to meet the needs of multiple applications from the cloud to the edge, you need high bandwidth, low latency and large-area coverage. NXP has corresponding products from antennas to processors to provide customers.
The company’s Layerscape Access can support four applications, a fixed wireless access platform, distributed units, radio units and integrated small cells, which can meet customers’ needs for coverage. Mr. Nikolay Guenov, Senior Director of Product Management at NXP, said: “In these applications, we pay great attention to cooperation with partners in terms of hardware and software. NXP’s product line can meet the needs of processors from single-core to 4-core to 16-core. Layerscape Access meets the needs of programmable baseband, combined with partner RF chips, we are able to fully support customer needs. In addition, our product line supports both sub-6G and mmWave antenna standards.”
From the perspective of meeting market needs, the company supports the 5G open ecosystem. In terms of radio frequency, in order to meet the needs of small base stations or CPEs, NXP provides baseband interfaces to partners.
In the industrial edge field, high security and reliability are required, especially low latency. NXP has integrated TSN technology into three products, including the i.MX RT1170 crossover MCU, the i.MX 8M Plus, and the LS1028A processor.
In the field of EV charging, NXP’s Kinetis MCUs for metering can achieve very high accuracy results. In addition, there is i.MX RT for host processors. In terms of security, SE050 secure element and tamper detection function in MPU and MCU are provided, and its Wi-Fi, Bluetooth and NFC can support wireless connections including wireless charging and payment.
At present, machine learning is very common in the field of edge computing. NXP already has a number of partners in this area, such as working with Canada’s Au-zone to develop a machine learning toolkit, and working with Arm to enable the Arm Ethos-U65 microNPU to provide scalable, high-performance machine learning use cases.
In addition, NXP is advancing open standards such as many device and wireless interoperability, operability, and CHIP initiative (an open source method for developing and implementing new unified connectivity protocols), working with many partners to support diversity better interaction between devices.
NXP continues to invest in China, build more local partnerships, and manufacture products in China for China. The company has established a good partnership with Tianjin University, Soochow University, Shanghai Jiaotong University, etc. In addition, there is cooperation with the Chinese government, including the establishment of an AIoT laboratory in Tianjin.