2385978_en-US

[#1]

Building AI/ML Devices at the Edge? Start with FRDM

From smart sensing and anomaly detection to computer vision, voice interfaces, and even multimodal GenAI—AI/ML use cases are rapidly moving onto the device.

But if you're an engineer getting started, the real questions usually are:

• What hardware should I use?
• What tools actually work for embedded AI?
• Do I need the cloud for everything?

Let’s break it down !

---

Why Edge AI?

Running AI directly on-device enables:

• Real-time decisions (low latency)
• Better privacy
• Offline operation (no cloud dependency)
• Optimized power consumption

That’s exactly where the FRDM development platform comes in.

Quick Positioning: From MCU to Edge AI Processor

Category	Boards	AI Capability Level	Typical Use Case
MCU (Low Power Edge ML)	FRDM-MCXA156	+	sensor AI, TinyML
MCU + Neural Acceleration	FRDM-MCXN947	++	Edge AI with vision/audio, TinyML
Application Processor (Entry Edge AI)	FRDM-IMX93	+++	HMI + AI inference
High-Performance Edge AI	FRDM-IMX8MPLUS	++++	Vision AI, Advanced HMI
Next-Gen AI + Safety	FRDM-IMX95 / PRO	+++++	Gen AI, Advanced Edge Computing

AI/ML applications can run on general-purpose hardware, but leveraging dedicated hardware acceleration significantly improves performance, enables faster results, and reduces power consumption. Below is a reference list of FRDM development boards that support AI/ML applications, helping you choose the right platform based on your target use case

Board	Positioning	AI Acceleration	Hardware Capabilities (AI-relevant)	Best For
FRDM-MCXA156	Entry-level MCU (TinyML)	No NPU (CPU-only)	Sensors via Expansion headers (Arduino, MikroBUS, Pmod) Parallel display support	Sensor ML Anomaly detection Basic TinyML
FRDM-MCXN947	MCU with neural acceleration	NPU	Parallel camera interface (basic) Parallel display Audio (PDM/I2S)	Voice AI  Low-res vision Object classification Anomaly Detection
FRDM-IMX93	Entry Edge AI MPU	NPU	MIPI CSI camera Display (MIPI DSI/LVDS) Audio + connectivity	Smart HMI Light vision AI Edge gateways
FRDM-IMX8MPLUS	Advanced Multimedia + Edge AI platform	NPU	Multi-camera (MIPI CSI) High-res display (HDMI/DSI)  Audio DSP Connectivity (Wi-Fi, BLE, Ethernet) Some PCIe expansion	Computer vision Object detection Industrial AI
FRDM-IMX95	Next-gen AI + real-time MPU	Next-gen NPU + heterogeneous compute	Multi-camera pipelines Advanced HMI Industrial connectivity M.2 expansion(Up to 1 AI accelerators)	Robotics Industrial AI Safety applications
FRDM-IMX95-PRO	Full-featured AI dev platform	High-performance NPU + scalable AI (Ara240 Discrete NPU)	Multi-camera Advanced display M.2 expansion (Up to 2 AI accelerators)	Advanced AI prototyping Gen AI Edge servers Edge computing

 Software and tools for ML/AI applications

 

GoPoint

GoPoint accelerates AI/ML evaluation on FRDM platforms powered by i.MX application processors by providing a ready-to-use, graphical environment with pre-integrated demos. Developers can quickly run applications such as image classification, object detection, and voice recognition directly on the hardware without complex setup. These demos are already optimized for available compute resources—including CPU, GPU, DSP, and NPU—allowing users to immediately visualize performance and understand how AI workloads map to the system. This makes GoPoint an ideal starting point for exploring edge AI capabilities and validating use cases before moving into full application development.

---

Application Code Hub (ACH)

Application Code Hub complements rapid evaluation tools by offering a centralized repository of reusable, production-oriented software examples for FRDM boards. It provides full application projects, source code, and documentation that developers can directly import into MCUXpresso IDE or VS Code. With filtering based on use case—such as vision AI, audio processing, or anomaly detection, ACH enables developers to quickly find and customize reference implementations. This helps bridge the gap between proof-of-concept and real product development, significantly reducing development time while enabling scalable AI/ML application design.

---

eIQ Time Series Studio (TSS)

eIQ Time Series Studio is purpose-built for developing AI models based on sensor and time-series data, making it highly relevant for FRDM-based edge intelligence applications. It provides a guided workflow for data collection, labeling, model training, and validation, all optimized for MCU-class devices. Developers can easily transform raw sensor data—such as vibration, motion, or environmental signals—into deployable machine learning models for use cases like predictive maintenance, anomaly detection, and condition monitoring. With built-in analytics and seamless deployment to FRDM boards, TSS simplifies the path from data to intelligent behavior on the edge.

---

eIQ AI/ML Software Environment

The eIQ software environment is the foundation that enables AI/ML development across the entire FRDM ecosystem, providing an end-to-end workflow from model creation to on-device inference. It supports importing and optimizing models from popular frameworks such as TensorFlow, PyTorch, and ONNX, and integrates tightly with MCUXpresso and Linux-based environments. eIQ includes tools for model optimization—such as quantization and pruning—as well as runtime engines designed for efficient execution on CPUs, DSPs, and NPUs. By combining these capabilities with hardware acceleration available on FRDM boards, eIQ allows developers to build, deploy, and run real-time AI applications directly on embedded devices with minimal reliance on cloud computing.

FQA

What is an NPU in the FRDM platform?

A Neural Processing Unit (NPU) in the FRDM platform is a dedicated hardware accelerator integrated into certain microcontrollers (such as the MCX-N family) that is specifically designed to execute machine learning and neural network workloads efficiently. Unlike general-purpose CPUs, the NPU is optimized for the mathematical operations used in AI models, enabling significantly faster inference—up to tens of times higher throughput—while consuming less power. In FRDM boards, the NPU works alongside the CPU and DSP to offload complex AI computations, allowing real-time processing for applications such as image recognition, voice detection, and sensor-based anomaly detection directly on the device. Combined with NXP’s eIQ® software environment, the NPU becomes the core execution engine that transforms FRDM platforms into efficient, low-power edge AI systems capable of running intelligent applications without relying on the cloud

What is a Discrete NPU?

A Discrete Neural Processing Unit (DNPU) is a standalone AI accelerator designed specifically to execute machine learning and neural network workloads efficiently. Unlike integrated NPUs that are built into a processor, a DNPU exists as a separate chip or module that can be added to a system. It offloads compute-intensive AI operations, such as matrix multiplications and deep learning inference from the main CPU or GPU, delivering significantly higher performance and better energy efficiency. This makes DNPUs ideal for advanced edge AI applications like computer vision, generative AI, and real-time multimodal processing.

---

How do I use Ara modules (DNPU) with FRDM boards?

Ara modules, based on NXP’s DNPU technology, can be used with compatible FRDM boards to extend AI processing capabilities. On supported i.MX-based FRDM platforms such as FRDM-IMX95 or FRDM-IMX95-PRO—developers can connect Ara modules (e.g., Ara240) through the M.2 expansion interface. Once connected, the Ara module works alongside the main processor to offload complex AI workloads, enabling faster inference, lower latency, and improved power efficiency. Using the eIQ® AI software environment, developers can prototype and validate models on FRDM, then scale performance by enabling Ara acceleration, creating a seamless path from development to high-performance edge AI deployment.

From TinyML to advanced edge AI and GenAI, discover how to build intelligent systems directly on-device with FRDM, no cloud dependency required.

FRDM-IMX8FRDM-IMX8MPFRDM-IMX9FRDM-MCXNi.MX Application ProcessorsMCU