With the gradual disappearance of the Internet dividend, the Internet of things in the overall trend of scientific and technological development, has become a new generation of industry development direction.
Among them, the application of Internet of things in industry is called industrial Internet of things. Industrial Internet of things continuously integrates various kinds of sensing or control sensors or controllers with sensing and monitoring capabilities, as well as ubiquitous technology, mobile communication, intelligent analysis and other technologies into all aspects of industrial production process, so as to greatly improve manufacturing efficiency, improve product quality, and reduce product cost and resource consumption, Finally, the traditional industry will be promoted to a new stage of intelligence. Industrial Internet of things is essentially machine to machine (M2M) support extended to the cloud.
At present, the industrial Internet of things is still in the early stage of development, but because of its broad application prospects and huge revenue potential, many companies and organizations have made a lot of investment in the industrial Internet of things.
Achronix semiconductor is one of them. Alok sanghavi, senior product marketing manager of achronix semiconductor, gave an interview with non internet on industrial Internet of things related topics. Achronix believes that the industrial Internet of things (iiot) is a new driving force for the global economy. The industrial Internet of things not only means the communication connection between machines, tools, automation systems, control systems, but also the operation support systems such as ERP systems running in the local data center or cloud. At the same time, the industrial Internet of things adds and redistributes the intelligence in the application fields such as intelligent factory, energy production and management, business and building automation, intelligent logistics and asset tracking, process automation and intelligent agriculture. Therefore, connectivity and intelligence are two features of industrial Internet of things.
These industrial Internet of things markets have some common characteristics. First, many standards and protocols in these areas have not yet been established; Secondly, new technologies such as artificial intelligence or edge computing are still in the early stage of adoption in these fields; Third, the new artificial intelligence or edge computing accompanying with industrial Internet of things need advanced hardware acceleration.
Alok told reporters from www.anhronix.com that with its programmability and flexibility, various FPGA products will play a key role in the new economy driven by the industrial Internet of things. Computing power is also particularly important. The special design resources of embedded FPGA (efpga) used in ASIC or SOC will add super hardware acceleration to the system, such as machine learning processor (MLP) in the semiconductor intellectual property rights (efpgaip) of acronix speedcoregen4 embedded FPGA.
Therefore, achronix’s strategy is to provide customers with a brand-new and most complete FPGA product portfolio, including independent FPGA chips, chip manufacturing services based on FPGA core grains, and embedded FPGA Semiconductor Intellectual Property Rights (efpgaip), so as to help customers or their partners to develop FPGA products without significantly changing their algorithms and development environment, Successfully upgrade the efficiency and capability of its chip.
Alok goes on to say that for the industrial Internet of things (iiot), connectivity and computing are like two sides of the same coin. With the support of connection technologies including wireless and wired solutions, edge computing and cloud computing will become important technologies for industrial Internet of things applications. Achronix’s high-performance and high-density FPGA chips, chip based on FPGA grains and efpgaip provide strong support for the two fields of industrial Internet of things.
In the field of communication and connection, there are many new standards and communication protocols emerging, such as 5g is entering the factory. Therefore, it is necessary to transfer and transform continuous sensor data between different protocols and standards, including emerging protocols, traditional protocols and private protocols. FPGA has a historical advantage in this aspect. With the emergence of new communication protocols, FPGA chips and SoC / ASIC with efpgaip will continue to serve these markets.
In another area, edge computing and related AI technologies are increasingly deployed in different networks. For example, the application of artificial intelligence in factories has expanded from machine vision based part quality detection to sensor fusion and predictive maintenance…… The wider application of these technologies brings new challenges to the traditional ASIC methods; And edge intelligence in industrial Internet of things needs hardware acceleration to achieve low latency data processing.
Therefore, new devices need to be reprogrammable to cope with the evolution of standards or protocols, and / or new algorithms for artificial intelligence and other embedded intelligence. They require that new architecture innovation must be introduced into SOC designed for smart factory and many other applications. Chip designers should pay attention to embedded FPGA IP products with second level acceleration function, such as machine learning processor (MLP) in speedcoregen4efpgaip of acronix, It can improve the performance of AI / machine learning by 300%.
FPGA in edge computing
As for the specific role of edge computing in the industrial Internet of things, Alok added that for many industries, edge computing is a very important trend. Therefore, all stakeholders, including factory owners, equipment manufacturers, system integrators, software vendors and related chip designers, must have a vision of their hardware acceleration infrastructure. They should choose a hardware infrastructure with flexibility and scalability, and be able to migrate or evolve from standard devices to SOC for category applications, without making architectural changes to their edge computing solutions. Achronix provides a complete solution based on FPGA for edge computing.
For example, industrial Internet of things solution providers can choose achronix’s independent FPGA chip to develop algorithms and quickly start. After that, they can obtain FPGA chip grain and system level package (SIP) services from achronix for manufacturing their own SIP multi grain chips, in which FPGA devices are directly connected to processors and / or other devices, thus achieving higher performance. When the demand of AI devices is increasing, and they need more computing power to support a set of intelligent factory functions, they can develop a system on chip (SOC) by using the semiconductor intellectual property (IP) of acronix speedcoregen4e FPGA, so that they can be more directly oriented to the application and get the highest cost performance.
In speedcore fpgaip based on gen4 architecture, the new machine learning processor (MLP) module for artificial intelligence / machine learning (AI / ml) is a complete AI / ml computing engine in addition to the cost performance and higher bandwidth of efga. Each MLP includes a circular register file that reuses stored / cached weights or data using temporal locality, thereby improving performance by significantly reducing data movement for various calculations. MLP is tightly coupled with its adjacent MLP and larger memory units to maximize processing performance and provide the maximum number of operations per second with the lowest power configuration. MLP supports fixed-point and floating-point formats (bfloat16; 16 bit, half precision; And block floating point numbers). Users can dynamically select the best data precision according to the requirements of each application, so as to trade-off between precision and performance.
In the industrial Internet of things market, data has become the most critical part. As the front-end data collection and processor, sensors are becoming more and more important, and sensor fusion has gradually become the general trend.
Alok explained that although achronix does not provide any sensor products, achronix’s most complete FPGA product portfolio in the industry can support the needs of sensor fusion and related data communication. For example, in today’s factories, due to the continuous improvement of accuracy, the demand for optical detection of precision parts is becoming more and more common. For human workers, this is a challenging and repetitive task. Therefore, optical detection is increasingly processed by intelligent machines, which need a variety of FPGA products, including independent FPGA chips, FPGA grain based SIP package chips and efpgaip.
There are two driving forces for FPGA products, especially for efpgaip. On the one hand, sensor fusion in some industrial Internet of things (iiot) applications may need to optimize the system architecture, because the sensor fusion needs low power consumption and higher performance. SOC or ASIC integrated with FPGA will have excellent power efficiency, programmability and unit cost. On the other hand, when the data from sensors is large, such as machine vision and lidar data, and the algorithm is constantly improving, front-end data processing or sensor fusion really needs FPGA devices or FPGA logic array to speed up the processing process.
Nowadays and in the future, data security has always been the most concerned and sensitive issue for enterprises and users. Security and security protection capability are important aspects of industrial Internet of things (iiot). Because any intrusion or data leakage may cause huge losses. Therefore, it is necessary to consider the safety planning from the perspective of the whole life cycle, especially for the fixed assets such as the equipment in the intelligent factory. Equipment suppliers must be able to support any customized or unique security algorithms they use in different customers’ factories, even if they all plan to use the same manufacturing equipment products. In the whole life cycle of equipment, encryption, authentication or other information security means in industrial Internet of things applications should have a certain degree of flexibility and scalability.
For these reasons, reprogrammable devices such as FPGA have a long history of application in information security. Achronix has supported many users in this area, covering different levels of security. Therefore, chip suppliers who develop SOC for industrial IOT applications can use efpgaip such as speedcore, and use some ASIC resources to develop flexible and scalable security modules. In addition, speedcore fpgaip can provide custom blocks to speed up the security algorithm, so as to achieve hardware speed level security.
At the end of the interview, regarding the current ecological construction problems of the industrial Internet of things and the efforts made by achronix, Alok believes that “as the ecosystem and standards of the industrial Internet of things (iiot) are still developing, solution providers should choose a method that can evolve forward in different stages of development.
Developers can start by choosing achronix’s independent FPGA chip to develop the algorithm and quickly start the project. After that, they can obtain FPGA chip grain and system level package (SIP) services from achronix for manufacturing their own SIP multi grain chips, where FPGA devices are directly connected to processors and / or other devices for better internal connectivity and performance. When the demand for devices with edge computing is increasing, and they need more computing power for a set of industrial IOT functions, they can use the achronix speedcoregen4e FPGA Semiconductor Intellectual Property (IP) to develop a system on chip (SOC) to meet today’s and future needs.