Low-Power Congatec SMARC Modules for Powerful AI Inference

Discover how NXP’s i.MX 95 application processors—with an integrated NPU and congatec’s Smart Mobility ARChitecture (SMARC) modules—enable low-power, high-performance AI vision at the edge. Ideal for industrial and embedded applications, this platform supports real-time inference, robust design and flexible AI model deployment.

Low-power SMARC modules, powered by NXP’s i.MX 95 processor and its integrated Neutron Neural Processing Unit (NPU) are transforming AI inference at the industrial edge. Drawing inspiration from the human eye’s efficiency, this solution emphasizes the need for real-time, local processing to reduce latency, ensure data privacy and eliminate reliance on cloud connectivity.

At the heart of this transformation is the NPU, a neuromorphic processor designed to handle deep learning (DL) and machine learning (ML) tasks with minimal power consumption. The i.MX 95 processor combines this NPU with Arm Cortex-A55 cores, Cortex-M7 and M33 controllers, and a Mali 3D graphics processing unit (GPU), delivering up to 2 Tera Operations Per Second (TOPS) of AI performance. This architecture supports high-resolution image processing, real-time decision-making and immersive graphics, making it ideal for applications in robotics, surveillance, medical imaging and industrial automation.

The SMARC 2.2 Computer-on-Module (CoM) conga i.MX95 serves as the hardware foundation, offering robust performance, low power consumption and wide-temperature tolerance. It integrates seamlessly with industrial cameras like the Basler dart, enabling compact, efficient AI vision systems. The module supports multiple input-output (I/O) interfaces and can operate in harsh environments, making it suitable for edge deployments in agriculture, manufacturing, retail and more.



The brain behind AI vision: The conga-SMX95 SMARC 2.2 module brings neuromorphic intelligence to the edge.

Simplified AI Development with eIQ
On the software side, NXP’s eIQ machine learning toolkit simplifies the development and deployment of AI models. It supports both bring your own data (BYOD) and Bring Your Own Model (BYOM) workflows, allowing developers to train and optimize models using popular frameworks like TensorFlow, PyTorch and ONNX. The toolkit includes tools for model validation, quantization and performance tuning, ensuring efficient inference on edge devices.

The synergy between hardware and software enables developers to build intelligent systems capable of real-time object detection, anomaly recognition, voice and face recognition and complex diagnostics. These systems can operate autonomously, adapt to changing environments and interact naturally with humans through gestures and emotion recognition.

Ultimately, the i.MX 95-based SMARC modules are a powerful, scalable solution for AI-accelerated embedded vision. With application-ready building blocks and a comprehensive software ecosystem, developers can reduce time to market, enhance system reliability and unlock new possibilities in edge AI.