As AIoT continues to evolve, demand for intelligent edge applications is rising rapidly. Use cases such as real-time object detection in smart surveillance, ultra-HD content processing in commercial displays, edge-based data analytics, and interactive portable devices all require stronger AI performance and advanced multimedia capabilities from system-on-modules. These applications increasingly rely on powerful edge AI inference, support for high-resolution video encoding and decoding, and an efficient balance between performance and energy consumption.
To address these needs, Forlinx Embedded, a key partner of Rockchip, has launched the FET3572-C SoM based on the RK3572 processor. It integrates a hardware-accelerated 4 TOPS NPU with an octa-core heterogeneous architecture and a well-developed ecosystem, providing a flexible and high-performance platform for edge AI applications and enabling new innovations in embedded intelligence.
Empowered by 4TOPS NPU for More Efficient Edge AI Inference
Equipped with a dedicated NPU offering 4 TOPS (INT8) computing power, the FET3572-C SoM addresses the critical bottlenecks of AI performance and inference speed at the edge. Its compatibility with INT4/INT8/INT16 mixed precision and frameworks such as TensorFlow and PyTorch facilitates straightforward deployment of a wide range of edge AI inference tasks.
From facial recognition in smart security to anomaly detection in industrial settings, or intelligent noise reduction in consumer electronics—the Forlinx Embedded FET3572‑C SoM enables real-time responses with its powerful computing capability. By eliminating reliance on cloud computing, it dramatically cuts data transmission costs and latency, delivering true ”local intelligence” for edge devices.
Robust Foundation Balancing Performance and Power Consumption
Powerful AI and multimedia capabilities require a solid hardware foundation. Built on an octa-core heterogeneous architecture (2×Cortex‑A73 + 6×Cortex‑A53) and an advanced 8nm process, the board offers over 100% higher performance than previous mid-range platforms while reducing typical power consumption by more than 50%.
The dual A73 cores handle demanding AI inference and ultra-HD video processing, while the six A53 cores manage lighter tasks—ensuring peak power when needed and optimal efficiency during low loads.
Test results include an Antutu v10 score above 310,000 and standby power below 10mW, enabling support for high-performance loads like the 4 TOPS NPU and 8K codec while meeting the low-power demands of portable devices.
8K Codec: The Ultimate Ultra-HD Experience
Equipped with a high-performance multimedia unit, the board supports 8K decoding and 4K encoding across mainstream and open-source formats.
It integrates a 12 MP ISP for accurate color reproduction and supports 5 x camera input, ideal for multi‑channel capture and monitoring. With dual‑screen independent display (4K@60fps + 2K@60fps), it suits commercial displays, POS systems, digital signage, and more.
An efficient decoding architecture lowers system bandwidth and power use—delivering stunning 8K quality without compromising stability or energy efficiency.
Full-Scenario Adaptability
Beyond AI and multimedia, the board provides extensive connectivity:
- High‑speed interfaces: PCIe 2.1, dual Gigabit Ethernet
- Industrial buses: CAN‑FD, I2C
- Native support for LPDDR5/5X memory
It adapts seamlessly to smart security, commercial displays, industrial control, edge computing, and vehicle‑mounted systems—enabling everything from HD surveillance with real‑time AI analysis to local data processing and intelligent interaction.
Comprehensive Ecosystem & Support
To speed up development, the board supports Linux and Android and comes with full resources: driver source code, development manuals, and technical examples.
Backed by Forlinx Embedded’s expertise, customers receive end‑to‑end support—from solution design and prototyping to mass production—helping shorten R&D cycles and accelerate time‑to‑market.
The All‑ound Edge AI Platform
As edge AI and ultra‑HD demands grow, the Forlinx Embedded FET3572‑C SoM stands out with its 4 TOPS NPU, 8K codec performance, and balanced power efficiency.
It delivers an efficient, stable, and cost‑effective core solution for upgrading intelligent devices across industries.
The FET3572‑C SoM is now open for pre‑orders. Developers and manufacturers are invited to explore collaboration—let’s build more competitive intelligent products and drive the edge AI industry forward.
