For the longest time, application-specific integrated circuits (ASICs) have been designed to perform data processing for one application. Unlike general processors such as x86 or ARM, ASICs are hardwired to perform specific calculations or algorithms for a given task and can be manufactured in two ways: semi-custom or full-custom. Regardless, they can be expensive to develop, but become cost effective at scale, when designed for small, highly integrated devices that are mass-produced, resulting in a low cost per chip. But what if you required a little more versatility than an ASIC at a much lower cost than a general-purpose CPU?
Field-programmable gate arrays, or FPGAs, present a highly accessible solution to this need, since their logic can be customized for specific applications. Its advantage over ASIC, in terms of design change even after the product, has been deployed in the field has expanded the use of FPGAs. This feature allows the designer to upgrade from a remote location eliminating the need of fabrication from scratch. Therefore unlike ASICs, FPGAs can be electrically reprogrammed after they have been manufactured and can be re-reprogrammed depending on the application—like Lego blocks that can be taken apart and put back together.
Stay ahead of the industry with exclusive feature stories on the top companies, expert insights and the latest news delivered straight to your inbox. Subscribe today.
And while FPGAs aren't quite as powerful as purpose-built ASICs, their versatility has made them increasingly popular for machine learning applications as they can be optimized for different AI models and neural networks enabling them to outperform traditional CPUs and even GPUs.
One of the reports on the FPGA market Forecast till 2027, by Market Research Future (MRFR) states, “The global FPGA market was worth USD 9.0 billion in 2018 and is estimated to develop at a Compound Annual Growth Rate (CAGR) of 9.7 percent from 2020 to 2027.”
They are commonly used for applications that include streaming, data processing, and heavy data flow because they have low power consumption and high computing density. In recent years, FPGA-based accelerators have advanced as strong contenders to the conventional GPU-based accelerators in modern high-performance cloud and edge computing systems.
IaaS resources are being utilized by cloud customers across multiple industries including communication, computing, automotive, and consumer. High-Level Synthesis (HLS) is making the platform more accessible to developers by enabling the use of high-level languages like C, C++, SystemC, and OpenCL to configure FPGAs precisely. Cloud service providers are employing FPGAs to expedite various operations such as network encryption, deep learning, memory caching, webpage scoring, high-frequency trading, and video conversion.
Major firms such as Amazon and Microsoft are utilizing FPGAs in their virtual machines to provide customers with hardware accelerations, deep neural networks, software-defined networking, and Bing search result rankings. The automotive industry is a major driver of customer demand for FPGAs, with designers of Advanced Driver Assistance Systems (ADAS) using FPGAs for vision processing applications that require high-level processing and fine-grained parallelism.
Despite technical advancements, there are still constraints that could limit industry expansion, such as the high-power consumption required by embedded field-programmable gate arrays and the scarcity of uniform industry authentication techniques. However, companies like Intel, AMD, Nvidia, Xilinx, and Intel are providing full processing solutions with proprietary accelerators and SmartNICs, and Xilinx and Intel offer SoC-based FPGAs with ARM cores, which are extensively used in embedded applications.
The development of new FPGA programming software is crucial for the future, as programmable devices become more complex and contain one or more processors. The FPGA industry is growing in both market share and innovation, and as FPGAs with embedded processors become the primary means of designing embedded devices, hardware/ software co-design tools' potential will be realized, promoting further growth.
Companies like Lattice Semiconductor offer various design and verification tool suites, including Lattice Diamond Software and Lattice Radiant Software, as well as Lattice Mico, a graphic tool designed for soft microprocessor-based designs. Xilinx, Inc. introduced 16- nm FinFET+ automotive-qualified circuits targeting ADAS and self-driving cars, and Lattice Semiconductor announced new FPGAs for automotive applications like ADAS and In-Vehicle Infotainment (IVI) systems for the award-winning Lattice CrossLink-NX family. Xilinx, Inc. also launched the Kintex and Virtex portfolios of defence-grade FPGAs for space and military applications, which are anti-counterfeiting and designed to withstand harsh conditions while providing safety and reliability. FPGAs are also used in tactical vehicles such as cameras, radars, and electronic warfare systems to achieve higher coverage, data collection, and electronic countermeasures.
The FPGA industry is rapidly expanding and finding use cases in various industries. Although there are constraints such as high-power consumption requirements and the scarcity of uniform industry authentication techniques, the industry is innovating to overcome these challenges. The design of new FPGA programming software is essential to the industry's future, as programmable devices grow in size and complexity, and tools to take advantage of these capabilities and optimize designs are in high demand.
