Focus on artificial intelligence edge computing
Strengthen cooperation with the Chinese market
'Artificial intelligence allows machines to perceive the environment and enable them to respond better to human commands. As processor technology advances, artificial intelligence is shifting from cloud computing to embedded applications while still evolving rapidly. NXP is very concerned about the development of artificial intelligence, and hopes to introduce artificial intelligence technology into the intelligent Internet of Things. ' Geoff Lees pointed out.
According to Geoff Lees, NXP emphasizes cost and application in the development of artificial intelligence. 'No one will do AI for AI. Artificial intelligence technology will eventually lead to productization. The key to landing is cost, and second, Applications. NXP's products and solutions are used in a wide range of applications in the edge computing of the Internet of Things. As the demand for AI technology increases, we will select the appropriate AI core to join existing products, and users do not have to develop it themselves. Application software. The overall solution of 'turnkey' is our development direction and our strength. ' Geoff Lees said.
At the same time, Geoff Lees is very optimistic about the development of artificial intelligence in China. 'I think China's artificial intelligence and machine learning market is leading the world. Although American companies first started to develop AI technology, now the breadth of Chinese companies developing AI products is other countries. There is no way to compare. Chinese AI products are diverse, including toys, which have begun to emphasize the concept of AI. China is the most active area in AI applications.
Therefore, NXP will strengthen its cooperation with Chinese companies in the field of artificial intelligence in edge computing. According to Geoff Lees's concept, cooperation can be divided into three phases: The first phase is to cooperate with the artificial intelligence company of China's voice-side, related operations. The system and ecosystem are ported to NXP's products, not only in i.MX 7ULP, i.MX 8M and other application processors, i.MX RT and other cross-border processors, even in lower-end microprocessors. , the relevant ecological environment and operating system will be transplanted. The second phase is optimization. NXP will work with Chinese universities and research institutes to transplant relevant research results into products, including tools, language algorithms, etc. Not all Data processing operations must be done in the cloud, many machine learning algorithms can be optimized locally and more efficiently and safely. The third stage is to transfer more advanced machine learning, algorithms to low-cost micro On the processor, on the microcontroller, the application range of machine learning is further extended in the field of edge computing.
NXP's focus on artificial intelligence focuses on edge computing, while emphasizing cost, application and collaboration with the Chinese market.
Optimistic about FD-SOI process
Applicable to the smart Internet of Things market
The development and transformation of artificial intelligence Internet of Things is promoted by the combination of microprocessor technology and process technology. In terms of manufacturing process, FD-SOI route becomes the focus of NXP's intelligent Internet of Things layout.
In the 2017 media event, Geoff Lees said that NXP's embedded application processor will focus on FD-SOI technology. There are two reasons for adopting FD-SOI: First, with the development of technology, chips The complexity is getting higher and higher, such as integrated analog circuits, RF circuits, etc., and also requires lower power consumption. Some applications require fast wake-up function to communicate with the network cloud in real time. 'I personally think FD- SOI is the most suitable chip manufacturing process for IoT applications. ' Geoff Lees said.
As a semiconductor manufacturing process, FD-SOI has many technical advantages, such as reducing parasitic capacitance and increasing device frequency. Compared with bulk silicon, the frequency of SOI devices is increased by 20% to 35%. Due to the reduction of parasitic capacitance, leakage is reduced. Current, the power consumption of SOI devices is reduced by 35%~70%; the latch-up effect is eliminated; the pulse current interference of the substrate is suppressed, and the occurrence of soft errors is reduced; and the process is compatible with the silicon process, which can reduce the process by 13%~20%, etc. In the low-power IoT chip products, the FD-SOI technology has its own characteristics.
It is understood that NXP's i.MX 7ULP has been mass-produced with Samsung's 28nm FD-SOI, and six chips are in preparation, including i.MX 8 and i.MX 8X, and iMX RT will also be transferred. To the FD-SOI process.
At present, there are two major OEM companies in the world with FD-SOI production lines - Samsung and Grofund. In addition to the 12FDX process development in Dresden, Germany, Grofund is expected to be put into mass production in 2019. It is also building a FD-SOI production line in Chengdu, Sichuan, China. Shanghai Huali Microelectronics, a foundry company in China, has also reported the development of FD-SOI production process.
Geoff Lees said that when the technology is mature, it does not rule out cooperation with other foundries.