In the previous sections, we looked at a series of microcontroller boards with varied features, which have all of the components embedded in a single chip. What if you want your robot or drone to perform modern-day computations for path planning or obstacle avoidance? Microcontroller boards help us read sensor signals and control the actuator signals, provided they follow simple math libraries for understanding and processing sensors.

What if you want to perform intense computation, store incoming sensor values over time, and then perform computations on such stored values? That is where SBC comes into play. SBCs can be loaded with an OS and could perform computations like a modern-day desktop, in a small form factor setup. They can't immediately replicate a modern-day computer as desktop computers nowadays are attached to additional components such as a GPU or powerful network cards that help solve our problems.

SBC, on the other hand, can be found in simple applications such as robotics, drones, ATM machines, virtual slot machines, or advertisement display televisions. Today, there is a wide number of such SBCs available and they are still being researched and developed with more features, such as GPU support or a volatile memory of more than 2 GB. In fact, some boards support modern computation techniques such as machine learning algorithms or are preloaded with deep learning neural networks.

Let's look into some interesting SBCs and their features. We have classified them into two categories:

• CPU boards
• GPU boards

Let's begin with CPU boards.