Field Programmable Gate Array (FPGA) is a powerful integrated circuit that allows users to program custom digital circuits. In this post, we introduce what FPGA is and how does it work. Keep reading!
What is FPGA?
Integrated circuits chips include digital chips and analog chips. Digital chips can be divided into memory chips and logic chips. Logic chips generally include CPU, GPU, DSP and other general-purpose processor chips, as well as application-specific integrated circuit chips ASIC. FPGA (Field Programmable Gate Array) is also a type of logic chip.
FPGA is a product of further development on the basis of traditional logic circuits and gate arrays such as PAL (Programmable Logic Array), GAL (General Array Logic), and CPLD (Complex Programmable Logic Device). It uses computer-aided design to draw schematic diagrams that realize user requirements, edit Boolean equations, or use hardware description languages as design inputs; then go through a series of conversion programs, automatic layout and wiring, and simulation processes; finally generate FPGA data file to initialize the FPGA device.
In this way, the application-specific integrated circuit that meets the user’s requirements is realized, and the purpose of the user’s self-designed, self-developed and self-produced integrated circuits is truly achieved.
FPGA vs. CPU vs. GPU vs. ASIC
The mian difference between FPGA vs. CPU vs. GPU vs. ASIC is that for FPGA, the wiring and logic layout of the underlying logical operation units are not solidified. Users can program logic units and switch arrays through EDA software, and perform functional configuration, so as to realize integrated circuit chips with specific functions.
For other types of logic chips, such as ASIC, CPU, and GPU, the calculation relationship of the physical underlying logic units is fixed and immutable. Simply put, if the CPU, GPU, ASIC, etc. are like a well-built building, the layout of the rooms, corridors, and stairs in the building has been fixed; while the interior of the FPGA is similar to the magic staircase in Hogwarts, which can change the route relationship from room to room at any time.
How does FPGA work?
FPGA is composed of three programmable circuits such as programmable logic block (CLB), input/output module (IOB), programmable interconnect resource (PIR), and static memory SRAM for storing programming data.
CLB is the basic unit for implementing logic functions, and it is usually regularly arranged in an array and scattered throughout the chip. The IOB mainly completes the interface between the logic on the chip and the external pins and is usually arranged around the chip. PIR provides a wealth of connection resources, including vertical and horizontal mesh connections, programmable switch matrices and programmable connection points, etc., which connect each CLB, CLB and IOB, and IOB to form a specific function circuit. The static memory SRAM is used to store the programming data of the internal IOB, CLB and PIR, and form the control of the IOB, CLB and PIR, thereby completing the system logic function.
FPGA is a logic chip, which is different from other logic chips. The biggest feature of FPGA is field programmability. This feature allows FPGA to realize the logic function of any chip through programming, such as ASIC, DSP or even PC processor. This is why The reason why FPGA is called “universal chip”.
What are the features of FPGA?
FPGA has three major characteristics, namely high programmable flexibility, short development cycle and high parallel computing efficiency.
(1) Short development cycle
In logic chips, for example, the ASIC manufacturing process includes multiple steps such as logic implementation, wiring processing, and tape-out; while FPGA does not require wiring, masks, and custom tape-out, etc., so chip development is greatly simplified. General logic chips, such as ASIC, DSP, SOC, etc., take 14-24 months or even longer to develop, while FPGAs only need 6-12 months, which is 55% less than other chip development cycles.
Just as Xilinx, the world’s largest FPGA manufacturer, believes that faster is more important than cheaper. If a product is launched 6 months late, it will reduce profits by 33% within 5 years.
(2) High parallel computing efficiency
FPGA belongs to parallel computing, that is, it can execute multiple instruction algorithms at one time. Traditional ASICs, DSPs, and CPUs are all serial computing, and can only process one instruction set at a time. Therefore, in some special tasks, the parallel computing efficiency of FPGA is higher than the serial computing efficiency.
What are the applications of FPGA?
Due to the characteristics of high programmable flexibility, short development cycle and high parallel computing efficiency, FPGA has a wide range of application scenarios.
FPGA can be used in fields including network communication, consumer electronics, data center, automotive electronics, artificial intelligence and so on.
(1) Communication field
The communication field is one of the main application markets of FPGA. According to Frost & Sullivan data, the sales of FPGA chips used in the communication field in China reached 6.21 billion RMB in 2020, accounting for 41.3% of China’s FPGA chip market share. The compound annual growth rate from 2021 to 2025 will reach 17.5%.
FPGA is currently widely used in wireless communication and wired communication equipment to realize various functions such as interface expansion, logic control, data processing, and single-chip systems.
(2) Industrial field
The industrial field is one of the main application markets of FPGA. Frost&Sullivan data shows that the sales of FPGA chips used in this field in China reached 4.74 billion RMB in 2020, accounting for 31.5% of China’s FPGA chip market share, and the compound annual growth rate from 2021 to 2025 will reach 16.1%.
In the industrial field, FPGA is mainly used in video processing, image processing, CNC machine tools and other fields to realize signal control and operation acceleration functions. With the development of intelligence and automation technology, the industrial field is gradually shifting from the micro-core elements of human resources to the intelligent unmanned factory with the micro-core elements of automation.
(3) Data center
Data centers are one of the emerging application markets for FPGA. Frost & Sullivan’s data show that the sales of FPGA chips used in this field in China reached 1.61 billion RMB in 2020, accounting for 10.7% of China’s FPGA chip market share. The compound annual growth rate from 2021 to 2025 will reach 16.6%.
A data center is a network of specific devices for global collaboration, used to transmit, accelerate, display, calculate, and store data information on network infrastructure. Servers and storage are common basic equipment in data centers. In order to cope with complex and changeable application scenarios, FPGA chips are required to implement functions such as logic control, data conversion, function expansion, and system upgrades.
(4) Automotive electronics
The intelligence of automobiles has opened up the FPGA incremental space. According to Frost & Sullivan data, the sales of FPGA chips used in this field in China reached 950 million RMB in 2020, and the compound annual growth rate from 2021 to 2025 will reach 22.7%.
Under the trend of intelligent development of automobiles, FPGA can be used to control and drive electric vehicle motor control systems, connect driving systems, dashboards, radars, ultrasonic sensors and other on-board equipment, and realize signal processing and control such as laser radar and millimeter wave radar.
(5) Artificial Intelligence
Frost & Sullivan data show that the sales of FPGA chips used in this field in China reached 580 million RMB in 2020, and the compound annual growth rate from 2021 to 2025 will reach 16.9%.
Artificial intelligence algorithm chip implementation is divided into cloud-side processing and device-side processing. When processing on the cloud side, compared with GPU and ASIC chips, the internal parallel processing units of FPGA chips can reach millions of levels, which can achieve true parallel computing. Its programmability can also realize the flexible construction of data processing pipelines, so the computing speed is fast. The data access delay is low, which is more suitable for the real-time decision-making needs of artificial intelligence.
In the field of end-side processing, FPGA chips can achieve fast inference and decision-making. In addition, its features such as field programmable, customizable functions, high throughput, and low latency effectively meet the user’s requirements for various neural network designs and become the best choice for deploying and upgrading various compressed and optimized neural networks.
Field Programmable Gate Arrays (FPGAs) are semiconductor devices that are based around a matrix of configurable logic blocks (CLBs) connected via programmable interconnects.
- Communication field
- Industrial field
- Data center
- Automotive electronics
- Artificial Intelligence
The mian difference between FPGA vs. CPU vs. GPU vs. ASIC is that for FPGA, the wiring and logic layout of the underlying logical operation units are not solidified.