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-IMX8 FRDM-IMX8MP FRDM-IMX9 FRDM-MCXN i.MX Application Processors MCU
記事全体を表示