Artificial Intelligence (AI) is rapidly redefining embedded computing to provide core drivers for next-generation solutions in areas such as robotics, industrial automation, and medical devices. As data processing increasingly moves to the edge, a platform with efficient AI performance, low latency and high customization has become a rigid requirement in various industry sectors.
IWave's iW-RainboW-G58M system module (Figure 1) is based on Altera's Agilex ™ 5 SoC FPGA， Leap through edge AI design. This compact module combines Altera's first FPGA architecture that injects AI functionality directly into Fabrics, enabling a unique combination of flexibility, computing power, and efficiency for AI-driven embedded systems.
Figure 1: iWave iW-RainboW-G58M Modular System. Image source: iWave)
Compact design and powerful calculation force
IW-RainboW-G58M SoM 60 mm x 70 mm overall dimensions, B32A encapsulated, Altera Agilex supported ™ SoC FPGA E series products. The series of modules are divided into two groups:
Group A: A5E06AA/052A/043A/028A/013A SoC FPGA
Group B: A5E065B/052B/043B/028B/013B/008B SoC FPGA
Agilex ™ Series 5 module integrates asymmetric heterogeneous processing system, carries dual Arm Cortex-A76 and dual Cortex-A55 cores, and achieves optimal balance between performance and energy efficiency. Enhanced DSP blocks integrated with AI tensor blocks accelerate deep learning workloads, while GTS transceivers up to 28.1 Gbps PCIe ® X 8 and display interfaces such as HDMI and DisplayPort provide high-speed connectivity for complex multimedia and AI applications.
AI Software Ecosystem and Development Tools
Agilex ™ The seamless integration of platform 5 with the leading AI framework (Figure 2) simplifies the development of custom AI reasoning solutions. The platform supports TennorFlow, Pyr, and Open ™ Toolkit with Altera's FPGA AI Suite and Quartus ® Prime design software optimizes the in-depth learning model for FPGA reasoning.
Figure 2: Agile ™ Seamless integration of platform and leading AI framework. Image source: iWave)
With Altera FPGA AI suite, all developers from hardware engineers to data experts can efficiently build and deploy optimized AI pipelines using familiar workflows. The suite successfully interfaces FPGA design with the high-level AI framework, reducing product time to market while maintaining deterministic performance and hardware-level energy efficiency.
Cloud AI and Edge AI: A Balance
In modern architectures, AI processing is divided into cloud processing and edge processing (Figure 3). Cloud AI offers enormous scalability, but often faces delays and privacy restrictions. In contrast, edge AI performs real-time processing locally, enabling instant decision making, stronger security protection, and continuous and stable operation even without network connection.
Figure 3: Cloud AI is scalable, but often has latency and privacy limitations. Edge AI performs real-time processing locally for instant decision making, enhanced security, and continuous operation even without network connection. Image source: iWave)
In a hybrid deployment architecture, edge nodes handle time-critical analysis tasks, while the data center summarizes and processes larger datasets for trend analysis or predictive modeling.- g. The iW-RainboW-G58M is designed for this hybrid model, enabling flexible computing at the edge and seamless integration with cloud infrastructure.