The Development Trend of FPGA in AI Era
As a programmable logic device, FPGA has gradually evolved from the peripheral device of electronic design to the core of digital system in more than 20 years. With the advent of the era of cloud computing and artificial intelligence, FPGAs that are good at data-parallel computing, more flexible and low-latency will receive more attention, and FPGA manufacturers will also launch different types of solutions to deal with them—heterogeneous computing platform, IP FPGA is showing more new development trends
Driven by AI applications, the FPGA market is growing rapidly
The application development of cloud computing and artificial intelligence has put forward higher requirements for the computing power of chips. FPGA can perform data parallel and task parallel computing at the same time, and has greater flexibility. Many low-level hardware control operations that are difficult to implement using general-purpose processors or ASICs can be easily implemented using FPGAs. Therefore, FPGAs have received more and more attention and applications in recent years.
“The exponential growth in AI inference demand requires a specific architecture (DSA) for maximum efficiency. However, the DSA development cycle for fixed hardware is too long compared to the speed of AI innovation.” Vice President, Segment Core Markets, Xilinx Yousef Khalilollahi pointed out: “FPGAs provide a flexible platform that supports DSAs that can be customized to the latest AI technologies at any time, without waiting for long silicon development cycles.”
Under this circumstance, in recent years, the application field of FPGA is constantly expanding, and the market scale is also expanding. According to a report by research firm Global Market Insights, the scale of FPGAs is expected to exceed $9.98 billion in 2022. According to Semico Research forecast data, in artificial intelligence applications alone, the market size of FPGAs will triple in the next four years to $5.2 billion.
Platformization + Heterogeneous Integration, FPGA Shows New Trend
In the face of expanding application trends, various FPGA manufacturers are also introducing different solutions. These solutions reflect the different development strategies of different manufacturers in the face of new market demands, and also have different impacts on the technology trend of FPGA.
For Intel, the development strategy for FPGA is to integrate more processor product portfolios and provide comprehensive computing solutions for the cloud.
The Speedster7t recently released by Achronix represents another development trend, which strives to combine the programmability of FPGA with the wiring structure and computing engine of ASIC, through the two-dimensional network-on-chip (2D NoC), and high-density new machine learning The integration of processor (MLP) module arrays, etc., simplifies user designs to meet the needs of AI machine learning applications.
The Speedster7t system integrates mathematical functions, memory and programmability into the machine learning processor, combined with the cross-chip, two-dimensional NoC structure, to ensure the free flow of data throughout the device. In AI machine learning applications, where memory bandwidth is everything, the Speedster7t delivers impressive performance numbers.
Embedded FPGA, increasing market acceptance
Unlike the development trend of CPU, GPU, etc., which rapidly realize IP in the mobile era, embedded FPGA (eFPGA) has not become the mainstream of the market. However, with the advent of the era of cloud computing and artificial intelligence, the market acceptance of eFPGA is expected to be further improved, becoming one of the main development trends of FPGA products.
In this regard, Robert Blake pointed out that FPGA can effectively realize IP. eFPGA is easy to use, greatly reduces the threshold for use, and can quickly meet the diverse needs of users, whether it is system-level or chip-level. This is very important in the age of artificial intelligence, especially in edge-side computing. In fact, many users are now asking about FPGA IP. It is believed that more and more IP-based FPGAs will be applied in the future.
eFPGA refers to a development model in which one or more FPGAs are embedded in chips such as ASIC, ASSP or SoC in the form of IP. One of the characteristics of eFPGA is ease of use, which lowers the threshold for customers to integrate FPGA accelerators.
Customers can combine multiple IPs to create optimized programmable functions for specific applications by customizing their logic, Ram and Dsp resource requirements. This development model is widely used in chips such as CPUs and GPUs, but it has not been popularized in FPGAs.
Haoxinshengic is a pprofessional FPGA and IC chip supplier in China. We have more than 15 years in this field。 If you need chips or other electronic components and other products, please contact us in time. We have an ultra-high cost performance spot chip supply and look forward to cooperating with you.
If you want to know more about FPGA or want to purchase related chip products, please contact our senior technical experts, we will answer relevant questions for you as soon as possible