Skip to content
One of the most common questions we receive at Nvixeon is simple: which board should I use? 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 so you can choose without wasting money on hardware you do not need.
Quick Decision Guide
If you are learning STM32, on a tight budget, or building simple projects, choose the NVX-F4 from GH₵220.
If you need motor control, industrial IoT, or real-time tasks alongside connectivity, choose the NVX-H7 Compact from GH₵480.
If you need every peripheral simultaneously, or are doing R and D with complex systems, choose the NVX-H7 Pro from GH₵750.
If you are building on-device AI, computer vision, or running neural networks, choose the NVX-N6 Vision from GH₵950.
STM32F4 — The Workhorse
The STM32F4 family has been the backbone of embedded development for over a decade. The F411CEU6 on the NVX-F4 board packs a Cortex-M4 at up to 100MHz with FPU into a 48-pin package. It has a mature ecosystem with more tutorials, libraries, and community answers than any other STM32 family. It is affordable, straightforward to get started with, and well-supported by STM32CubeMX. Its limitations are 128KB SRAM, no Ethernet MAC, no CAN FD, and a single core. It is perfect for sensor nodes, data loggers, SPI display projects, USB HID devices, and any application where its memory is sufficient.
STM32H7 — The Powerhouse
The STM32H745 combines two independent cores running simultaneously — Cortex-M7 at 480MHz and Cortex-M4 at 240MHz. It has 1MB SRAM, 2MB dual-bank flash, Ethernet MAC, USB HS, dual FDCAN, three ADCs, LTDC display controller, and DCMI camera interface. The NVX-H7 Compact exposes this in a manageable footprint while the NVX-H7 Pro breaks out every pin of the TFBGA240 package. It does not have a hardware NPU, consumes more power than the F4, and the clock tree is more complex to configure. It is best for industrial motor drives, CAN bus gateways, Ethernet instruments, display HMI systems, robotics controllers, and audio DSP.
STM32N6 — The AI Platform
The STM32N6 is a different category entirely. It combines a Cortex-M55 at 800MHz with a dedicated hardware NPU capable of 600 GOPS at INT8. This means the NPU runs neural network inference while the M55 handles everything else — sensor fusion, communication, display, control — simultaneously. Where a Cortex-M7 might take 180ms to run MobileNetV2, the STM32N6 NPU does it in under 8ms with only 5 percent CPU load. It supports MIPI-DSI display, DCMI and CSI2 camera, and LPDDR4 memory. Note that the N6 is flashless — firmware and model weights live in external flash rather than on-chip. It is the most expensive board in our lineup and overkill if your project has no AI component. It is the right choice for object detection cameras, face recognition, gesture control, industrial vision, and any application where a neural network is part of the product.
Side-by-Side Comparison
NVX-F4: M4 at up to 100MHz, 1 core, 128KB SRAM, no NPU, SPI display, from GH₵220.
NVX-H7C: M7 plus M4 dual-core, 1MB SRAM, no NPU, RGB display, Ethernet optional, from GH₵480.
NVX-H7 Pro: M7 plus M4 dual-core, 1MB SRAM plus SDRAM, no NPU, LTDC RGB, 168 GPIOs, from GH₵750.
NVX-N6: M55 at 800MHz plus hardware NPU 600 GOPS, 4.2MB SRAM plus LPDDR4, MIPI-DSI, from GH₵950.
The Upgrade Path
Most Nvixeon customers start on the NVX-F4, build confidence with the STM32 ecosystem, then move to the NVX-H7 Compact when their project needs more memory or requires Ethernet and CAN. Engineers building complex industrial systems with many simultaneous peripherals land on the NVX-H7 Pro. Teams building vision-based products move directly to the NVX-N6.
Every board ships with a pinout diagram, schematic PDF, and example firmware. Browse the full lineup at nvixeon.com/products or contact us if you need help choosing the right platform for your project.
