The full name of FPGA is Field Programmable Gate Array, which is a type of logic chip. The logic chip also includes general-purpose processing chips such as CPU, GPU, and DSP, as well as application-specific integrated circuit chips ASIC.
With the development of the information industry and microelectronics technology, programmable logic embedded system design technology has become one of the most popular technologies for information products, with applications including artificial intelligence, data centers, medical care, communications, 5G, security, automotive electronics, Industry and many other popular fields.
And with the advancement of technology and the development of technology, it will expand to more and wider application fields. More and more designs are also starting to shift from ASIC to FPGA, and FPGA is entering every corner of our daily life in the form of various electronic products.
Advantages of FPGA Chips:
After the logic chips such as ASIC, CPU and GPU are manufactured, their chip functions have been fixed, and users cannot make any modifications to their hardware functions.
In contrast, the advantage of FPGA chip is that the connection and logic layout of the underlying logic operation unit are not solidified. Users can program the logic unit and switch array through EDA software to configure the function, so as to realize the integrated circuit chip with specific functions. Therefore, it is also suitable for computing fields where the underlying algorithm needs to be continuously changed, such as artificial intelligence algorithm optimization.
At the same time, this also means that FPGA chip companies not only need to provide chips, but also need to provide FPGA professional EDA software for chip configuration. FPGA chip companies are not only integrated circuit design companies, but must also be integrated circuit EDA software companies.
For artificial intelligence, algorithms are in rapid iteration. Although the ASIC chip has high area utilization, high speed and low power consumption, it can obtain the best performance; however, the development risk of ASIC is relatively high, and a large enough market is required to ensure the cost price, and the cycle from R&D to market is long. , not suitable for areas where algorithms such as Deep School and CNN are rapidly iterating.
FPGA can just be applied to the field of artificial intelligence to meet the needs of high-speed parallel computing. Based on the popularity of the global artificial intelligence market, it has also stimulated the FPGA market and is conducive to the development of the FPGA industry.
FPGA Market Competitive Landscape:
Demand in the FPGA chip market has been strong in recent years. Major chip makers are also increasingly competing in the market.
According to industry body forecasts, global FPGA demand will grow from $6.86 billion in 2021 to $12.58 billion in 2025, with a compound annual growth rate of about 16.4%.
In the global FPGA market, the market is highly concentrated due to high technical barriers and fast replacement. AMD Xilinx and Intel are the two leading suppliers in the industry. In 2019, the two companies together accounted for more than 85% of the global market share.
5 Directions of FPGA Development:
(1) FPGA-based embedded system (SOPC) technology
System on Chip (SOC) technology is more and more widely used in the field of chip design, and SOPC technology is the application of SOC technology in the programmable field. At the heart of this technology is building a processor inside an FPGA chip.
AMD Xilinx mainly provides hard core solutions based on Power PC, while Intel provides soft core solutions based on NIOSII.
(2) Design method based on IP library
In the future, the density of FPGA chips will continue to increase, and the traditional HDL-based code design method is difficult to meet the design needs of ultra-large FPGAs.
With the increasing number of professional IP library design companies, the types of commercial IP libraries will become more and more comprehensive, and the supported FPGA devices will become more and more extensive.
As an FPGA designer, the main job is to find IP library resources that suit the needs of the project, and then integrate these IPs to complete the top-level module design. Since all commercial IP libraries are verified, the simulation and verification work of the entire project is mainly to verify the correctness of the interface logic design of the IP library.
(3) Dynamic reconfigurable technology of FPGA
FPGA dynamic reconfiguration technology mainly refers to the realization of high-speed functional transformation of all and part of the logic resources of the chip for an FPGA chip with a specific structure, driven by a certain control logic, so as to realize time-division multiplexing of hardware and save logic resources.
As densities continue to increase, the functions that can be implemented by FPGAs have become more and more complex. The time required to configure all the logic of the FPGA at one time is also longer, which reduces the real-time performance of the system. The configuration function of local logic can realize "dynamic reconstruction on demand", which greatly improves the efficiency of configuration.
(4) CPU+FPGA is used in cloud data center
At present, the demand for image processing is rapidly increasing. The demand for image thumbnail generation, pixel processing, image transcoding, and intelligent analysis and processing of user-generated content, video image capture, and other methods continues to increase.
Many applications urgently need high-performance, cost-effective image processing solutions. In this case, the data center faces a core challenge - the balance between user experience and service cost.
In general, the existing pure CPU processing solutions are relatively expensive in terms of servers, electricity bills, broadband, venue personnel, and the user experience is relatively poor.
The solution based on FPGA+CPU heterogeneous computing, share CPU processing tasks through high-performance FPGA, prompt throughput performance, reduce latency, and achieve cost savings and energy efficiency improvement. FPGA+CPU is used in cloud data centers, and it is applied in fields with highly sensitive information. It is more secure and controllable to use self-designed chips.
(5) FPGA chips are developing in the direction of high performance, high density, low voltage and low power consumption
With the continuous improvement of chip production technology, the performance and density of FPGA chips are constantly improving. The early FPGA mainly completed the interface logic design, such as the glue logic of AD/DA and DSP. Today's FPGA is becoming the core component of the circuit to complete key functions.
In high-performance computing and high-throughput 1/0 applications, FPGAs have replaced dedicated DSP chips as the best implementation. Therefore, high performance and high density have also become important indicators to measure the design capabilities of FPGA chip manufacturers.
With the improvement of FPGA performance and density, power consumption has gradually become the bottleneck of FPGA applications. Although the power consumption of an FPGA is lower than that of a processor such as a DSP, it is significantly higher than that of an application-specific chip (ASIC).
FPGA manufacturers are also using various new processes and technologies to reduce the power consumption of FPGAs, and have achieved some obvious results.
