Single-Board Computers (SBC / SoC)
A single-board computer is a full computer on one board, built around a System-on-Chip (SoC) that integrates a multi-core CPU, GPU, RAM controller, and I/O. It boots a real operating system (usually Linux), so you get a filesystem, networking, package managers, and the ability to run many programs at once — at the cost of real-time precision and instant boot.
Most SBCs boot from a microSD card (or eMMC/SSD on newer models). Use a quality card, flash the OS with the official imager, and shut down cleanly — pulling power can corrupt the card.
Raspberry Pi​
The most common SBC in the lab. Drag to orbit the model below, scroll to zoom:
The relevant generations:
| Model | SoC / CPU | RAM | Notes |
|---|---|---|---|
| Pi 3 (B+) | Quad-core Cortex-A53 @ 1.4 GHz | 1 GB | Older, fine for light/headless tasks |
| Pi 4 | Quad-core Cortex-A72 @ 1.5 GHz | 1–8 GB | USB 3.0, dual HDMI; needs USB-C 5 V/3 A |
| Pi 5 | Quad-core Cortex-A76 @ 2.4 GHz | 4 / 8 GB | Much faster; needs 5 V/5 A (PD), active cooling recommended |
- Use for: Linux services, networking, cameras, light vision, ROS 2 nodes, dashboards.
- OS: Raspberry Pi OS (Debian-based) — flash with Raspberry Pi Imager.
- I/O: a 40-pin GPIO header (3.3 V logic) for sensors and HATs.
Under-powered supplies cause the on-screen lightning-bolt warning and random instability — use the official PSU for the model. The Pi 4/5 run hot under load; add a heatsink or fan, especially for vision or sustained compute.
The Raspberry Pi GPIO pins are not 5 V tolerant. Applying 5 V to a GPIO pin can permanently damage the SoC. Level-shift any 5 V signal before connecting it.
NVIDIA Jetson (edge AI)​
SoCs with an integrated CUDA-capable GPU for running ML/vision models on-device (e.g. Jetson Nano / Orin Nano).
- Use for: real-time computer vision, on-device inference, edge AI — see AI / Edge AI.
- Software: NVIDIA JetPack (Ubuntu + CUDA/cuDNN/TensorRT).
- Note: higher power draw than a Pi; budget for the right barrel/USB-C supply and cooling.
Pi vs. Jetson​
| Need | Reach for |
|---|---|
| General Linux, networking, ROS 2, light vision | Raspberry Pi |
| GPU-accelerated deep learning / heavy vision | Jetson |
For robotics and servers you rarely need a monitor. Enable SSH (and Wi-Fi) in the imager's advanced settings, then connect over the network — no keyboard/HDMI required.