Comprehensive Overview of FPGA: Introduction, History, Architecture, Development Flow, and Application Scenarios

1. FPGA Introduction

FPGA (Field‑Programmable Gate Array) is a semi‑custom circuit that evolved from earlier programmable devices such as PLA, PAL, GAL, and CPLD. It bridges the gap between fully custom ASICs and limited‑capacity programmable logic, offering both flexibility and high logic density.

2. History of FPGA

Compared with PROM, PAL/GAL, and CPLD, FPGA devices have grown larger and more powerful. Major manufacturers include Xilinx, Altera (now part of Intel), Lattice, and Microsemi, with Xilinx and Altera together holding about 88% of the market. In 2015 Intel acquired Altera for $16.7 billion, launching a roadmap that integrates Intel CPUs with FPGA fabric to support emerging workloads such as artificial intelligence.

Figure 1: FPGA Development Timeline

Figure 2: FPGA Applications in the Telecom Industry

3. Current Status of Leading FPGA Companies

Xilinx focuses on cutting‑edge acceleration solutions and open‑strategy support for major cloud platforms, maintaining a lead in data‑center deployments with its UltraScale+ family and the VU9P device used by AWS, Baidu, Alibaba, Tencent, and Huawei.

Xilinx has also introduced the next‑generation ACAP architecture and a 7 nm Everest device that integrates ARM cores, DSP blocks, and math engines, promising up to 20× AI performance improvement over VU9P.

Intel provides a full‑stack solution from silicon to platform to applications, keeping its hardware and software closed to avoid ecosystem fragmentation, which results in slower progress despite large investments.

Figure 3: Xilinx Product Portfolio

Figure 4: Intel Stratix Product Generations

4. Challenges of FPGA in Data‑Center Servers

High programming barrier: hardware description languages (HDL) require deep hardware knowledge, limiting the talent pool (estimated >20 k FPGA developers in China).

Complex integration: co‑design of hardware and software, system modeling, HDL coding, simulation, driver development, and board‑level debugging.

Longer development cycles compared with pure software projects.

Difficulties in obtaining independent IP cores.

5. FPGA Architecture

An FPGA consists of Configurable Logic Blocks (CLB), Input/Output Blocks (IOB), interconnect routing, and embedded hard macros such as RAM, DSP, and DCM.

Each CLB contains a small switch matrix of 4‑ or 6‑input lookup tables, multiplexers, and flip‑flops, enabling implementation of combinational logic, shift registers, or small RAMs.

IOBs are grouped into banks, each supporting independent I/O standards; modern devices provide over a dozen banks for flexible interfacing.

Routing resources include short local lines for CLB‑to‑CLB connections, long horizontal/vertical lines for high‑speed paths, and global low‑skew networks for clocks and control signals. CAD tools hide most routing details from the user.

Embedded hard macros (RAM, DSP, DCM, etc.) are integrated into the fabric, as illustrated in the internal block diagram.

Figure 5: FPGA Internal Block Diagram

6. FPGA Development Flow

The typical FPGA design flow includes functional definition & device selection, design entry, functional simulation, synthesis, place‑and‑route, and programming/debug.

Functional definition / device selection: Define system functions, partition modules, and evaluate performance, resource, cost, and routing constraints to choose an appropriate FPGA.

Design entry: Write HDL code (usually Verilog) to describe the desired hardware.

Functional simulation: Verify logic using testbenches; tools such as ModelSim or VCS generate waveforms and reports.

Synthesis: Convert HDL to a gate‑level netlist optimized for the target device; tools include Synplify and vendor‑specific synthesizers.

Place‑and‑route & implementation: Map the netlist onto the FPGA fabric, generate configuration bitstream, and produce reports.

Programming & debugging: Load the bitstream (.bit file) onto the board, perform board‑level tests, and iterate as needed.

7. How to Use an FPGA

After generating a verified bitstream, the typical software‑controlled startup sequence is:

Load the bitstream into the FPGA.

Reset the logic.

Wait for PLLs to lock.

Run self‑tests on external RAM, internal Block RAM, DDRC, etc.

Initialize all writable RAM and registers.

Configure registers according to the chip manual.

Begin normal business processing.

8. Typical FPGA Application Scenarios

FPGA excels in irregular, highly parallel, compute‑intensive, and protocol‑parsing workloads such as artificial intelligence, genomic sequencing, video encoding, data compression, image processing, and network processing.

Note: Scan the QR code below to follow the public account and access more cloud‑computing, micro‑service, and architecture resources.