<p>One of the most common questions we receive at Nvixeon is straightforward: which board should I start with? The answer depends entirely on what you are building. This guide gives you a clear, honest comparison of the three STM32 families we build boards around — F4, H7, and N6 — so you can make the right decision without wasting money on hardware you do not need.</p>
<h2>Quick Decision Guide</h2>
<table>
<thead>
<tr><th>If you need…</th><th>Choose</th></tr>
</thead>
<tbody>
<tr><td>Learning STM32, tight budget, simple projects</td><td>NVX-F4 (from $22)</td></tr>
<tr><td>Motor control, industrial IoT, real-time + connectivity</td><td>NVX-H7 Compact (from $48)</td></tr>
<tr><td>Every peripheral simultaneously, R&D, complex systems</td><td>NVX-H7 Pro (from $75)</td></tr>
<tr><td>On-device AI, computer vision, neural networks</td><td>NVX-N6 Vision (from $95)</td></tr>
</tbody>
</table>
<h2>STM32F4 — The Workhorse</h2>
<h3>What it is</h3>
<p>The STM32F4 family has been the backbone of embedded development for over a decade. The F4CEU6 on the NVX-F4 board packs a Cortex-M4 at 168MHz with FPU into a 48-pin package, making it one of the most compact capable microcontrollers available.</p>
<h3>Strengths</h3>
<ul>
<li>Mature ecosystem — more tutorials, libraries, and Stack Overflow answers than any other STM32</li>
<li>Low cost — the NVX-F4 Bare starts at $22</li>
<li>Simple to get started — CubeMX generates clean HAL code with minimal configuration</li>
<li>3.3V logic, 5V tolerant pins, USB Full Speed</li>
<li>Plenty of peripheral mix: 3× SPI, 3× I2C, 3× USART, 12-bit ADC</li>
</ul>
<h3>Limitations</h3>
<ul>
<li>128KB SRAM — tight for applications with large buffers</li>
<li>No Ethernet MAC</li>
<li>No CAN FD — only basic CAN on some F4 variants</li>
<li>Single core — everything runs sequentially unless you use an RTOS</li>
</ul>
<h3>Best for</h3>
<p>Learning embedded development, sensor nodes, data loggers, SPI display projects, USB HID devices, audio effects processors, and any application where 128KB SRAM and 512KB flash is sufficient.</p>
<h2>STM32H7 — The Powerhouse</h2>
<h3>What it is</h3>
<p>The STM32H745 is the top of the standard STM32 performance range. It combines two independent cores — a Cortex-M7 at 480MHz and a Cortex-M4 at 240MHz — both running simultaneously. The NVX-H7 Compact exposes this in a manageable footprint, while the NVX-H7 Pro breaks out every single pin of the TFBGA240 package.</p>
<h3>Strengths</h3>
<ul>
<li>Dual-core AMP — run two independent firmware images simultaneously</li>
<li>1MB SRAM — 8× more than the F4</li>
<li>2MB dual-bank flash with live firmware update support</li>
<li>Ethernet MAC (10/100), USB HS, 2× FDCAN</li>
<li>3× 16-bit ADC with up to 36 channels</li>
<li>LTDC display controller for RGB parallel displays</li>
<li>DCMI parallel camera interface</li>
<li>FMC bus for SDRAM, NOR Flash, NAND</li>
</ul>
<h3>Limitations</h3>
<ul>
<li>No hardware NPU — AI inference runs entirely on the M7</li>
<li>Higher power consumption than F4</li>
<li>More complex to configure correctly — clock tree has more options</li>
<li>TFBGA240 package requires a breakout board (the NVX-H7 Pro solves this)</li>
</ul>
<h3>Best for</h3>
<p>Industrial motor drives, CAN bus gateways, Ethernet-connected instruments, display HMI systems, robotics controllers, audio DSP with USB audio, and any application requiring simultaneous real-time and connectivity tasks.</p>
<h2>STM32N6 — The AI Platform</h2>
<h3>What it is</h3>
<p>The STM32N6 is a different category of device. It combines a Cortex-M55 at 800MHz with a dedicated hardware NPU capable of 600 GOPS at INT8. This makes it the only STM32 that can run real-time neural network inference while the processor remains largely free for other tasks.</p>
<h3>Strengths</h3>
<ul>
<li>Hardware NPU — 600 GOPS at INT8, offloads inference from the CPU entirely</li>
<li>Cortex-M55 with Helium MVE — fastest Cortex-M core available</li>
<li>MIPI-DSI display output with ChromART GPU</li>
<li>MIPI-CSI2 camera input + DCMI</li>
<li>4MB internal SRAM + LPDDR4 support for larger models</li>
<li>STM32Cube.AI deployment pipeline — ONNX and TFLite models in, C code out</li>
</ul>
<h3>Limitations</h3>
<ul>
<li>Most expensive — NVX-N6 Vision from $95</li>
<li>Newer ecosystem — fewer community tutorials compared to F4 and H7</li>
<li>Overkill if your project has no AI or vision component</li>
</ul>
<h3>Best for</h3>
<p>Object detection cameras, face recognition systems, gesture controllers, industrial vision inspection, keyword spotting, anomaly detection in sensor data, and any application where a neural network is part of the product.</p>
<h2>Side-by-Side Comparison</h2>
<table>
<thead>
<tr><th>Feature</th><th>NVX-F4</th><th>NVX-H7C</th><th>NVX-H7 Pro</th><th>NVX-N6</th></tr>
</thead>
<tbody>
<tr><td>Core Speed</td><td>168MHz</td><td>480+240MHz</td><td>480+240MHz</td><td>800MHz</td></tr>
<tr><td>Cores</td><td>1</td><td>2</td><td>2</td><td>1 + NPU</td></tr>
<tr><td>Hardware NPU</td><td>No</td><td>No</td><td>No</td><td>600 GOPS</td></tr>
<tr><td>SRAM</td><td>128KB</td><td>1MB</td><td>1MB</td><td>4MB</td></tr>
<tr><td>Ethernet</td><td>No</td><td>Yes</td><td>Yes</td><td>Yes</td></tr>
<tr><td>CAN FD</td><td>No</td><td>2×</td><td>2×</td><td>1×</td></tr>
<tr><td>Camera</td><td>No</td><td>No</td><td>DCMI</td><td>DCMI + CSI2</td></tr>
<tr><td>Display</td><td>SPI TFT</td><td>SPI/RGB</td><td>RGB LTDC</td><td>MIPI-DSI</td></tr>
<tr><td>Starting Price</td><td>$22</td><td>$48</td><td>$75</td><td>$95</td></tr>
</tbody>
</table>
<h2>The Upgrade Path</h2>
<p>The four boards form a natural progression. Many Nvixeon customers start on the NVX-F4, build confidence with the STM32 ecosystem, then move to the NVX-H7 Compact when their project outgrows the F4’s memory or needs Ethernet and CAN. Engineers working on industrial systems with complex peripheral requirements typically land on the NVX-H7 Pro. Teams building vision-based products move directly to the NVX-N6.</p>
<p>Every board ships with a pinout diagram, schematic PDF, and example firmware. You can <a href=”/products”>browse the full lineup here</a>, or <a href=”/contact-us”>contact us</a> if you need help choosing the right platform for your specific project.</p>