Raspberry Pi: Five ways to upgrade your $35 computer

If you want to add more pep to your Pi, fast mobile internet, or a suite of sensors, these are the add-ons you need.

The Raspberry Pi may pack in a lot of computer for $35, but there are times you’ll want to do more than the vanilla board allows.

Fortunately the success of the Pi, more than 20 million sold and counting, means a huge ecosystem of add-ons are available to extend the Pi’s capabilities.

Whether it’s faster storage or mobile internet connectivity, all can be added, and it can be as easy as slotting a new HAT board on top of the Pi.

Here are five ways you can upgrade your Raspberry Pi using hardware add-ons.

1. Fast mobile connectivity

If you’re interested in fast mobile internet on the Raspberry Pi, the 4GPi is the Pi’s first commercially produced add-on board offering CAT4 LTE connectivity, with 150Mbps download and 50Mbps upload speeds.

The add-on, produced by Japanese company MechaTracks, is compatible with any Raspberry Pi model that has a 40-pin GPIO header: including the Pi 1 A+ and B+, Pi 2 Model B, Pi 3 Model B and B+, as well as the Pi Zero and Zero W.

While the board is expensive, $222 / €195, TechRepublic’s James Sanders writes it is “substantially more affordable than other LTE-connected development devices”, making it attractive when developing hardware that requires fast mobile connectivity. He says competing products suffer from being bandwidth limited, or are adapters for LTE modems used in laptops, with smaller antenna connectors.


The 4GPi attached to a Raspberry Pi.

Image: MechaTracks

2. Fast SSD storage

If you’re sick of the Pi being bottlenecked by its slow SD card storage, you’ve got options, such as adding a fast Solid-State Drive (SSD).

Unfortunately you won’t get the full speed of the SSD, but you can still achieve a noticeable speed up over an SD card.

There are various options for adding an SSD to the Raspberry Pi, but one recent, well-received add-on is the Geekworm X850 mSATA storage expansion kit.

The $23 kit contains everything you need to mount a tiny mSATA SSD on top of the Pi, including the adapter board, cable, screws and washers, although obviously, you’ll still need to add your own mSATA SSD. It is compatible with Raspberry Pi 3 Model B+, 3B, 2B and the 1B+ and setting up the Pi to boot from the SSD will require delving into the terminal, as outlined here.

While the speed of the drive will be squeezed by connecting via a USB 2.0 connection, this reviewer still describes the kit as delivering “a very noticeable improvement”, saying the system boots faster and runs more smoothly.

On paper, the Pi-Desktop Kit should also be able to boot the Pi from an SSD, although a recent hands-on by TechRepublic sister site ZDNet was unable to getting this feature to work.

3. Machine learning accelerators

Machine learning has proved its worth in teaching computers how to recognise images, speech and language.

While the Raspberry Pi struggles to carry out machine learning on its own, there are USB accelerators that give the Pi sufficient power to handle a range of computer-vision tasks.

Intel’s Movidius Compute Stick boosts the rate at which the Pi carries out vision-related tasks such as facial and object recognition, using its 12-core Myriad 2 Vision Processing Unit.

The $79 / £84 USB stick is capable of 100 gigaflops (one thousand-million, floating-point operations per second) and generally consumes a single watt. Rough estimates of performance online say the stick’s VPU can do 10 inferences per second using a GoogLeNet convolutional neural network (CNN), a machine-learning model commonly used for image recognition. That’s compared to about 2 inferences per second using Google’s Inception CNN architecture on an unaided Raspberry Pi.

Google also plans to launch its own USB stick that it says will dramatically accelerate the rate at which the Pi runs trained computer-vision models, even more so than the Movidius Neural Compute Stick.

Google’s says the Edge TPU Accelerator will support a range of computer-vision models, and be capable of running state-of-the-art mobile vision models, such as MobileNet V2, at 100+ fps. You can register here to find out when it’s available.

As indicated, these accelerators allow the Pi to offer acceptable performance when running trained computer-vision models. However, the training of these models still needs to be carried out using a more powerful system.

4. Power over Ethernet

The official Power over Ethernet (PoE) HAT is an add-on board that sits on top of the Raspberry Pi 3 Model B+ and allows the computer to be powered using an Ethernet connection.

This capability is particularly useful for people wanting to run Pi boards in remote areas where mains power isn’t available and, if combined with network booting, makes it simpler to use multiple Pis in a factory or other workplaces.

The HAT uses the 802.3af PoE standard, which allows 15W to be delivered to the board, providing it is connected to a network with the necessary 802.3af PoE sourcing equipment.

The £18 HAT currently appears to be out of stock, following earlier reports of problems with some not delivering enough power, but retail sites will let you sign up to be notified when it’s back in stock.

5. Sensors and camera

The $40/ £32 SenseHAT simplifies getting started tinkering with hardware by bundling together various sensors, a joystick and an LED matrix that the Pi can interact with.

You’ll be able to use the SenseHAT to read data about the Pi’s orientation or how fast it’s travelling — via the accelerometer, 3D gyroscope and magnetometer — as well as measure the surrounding pressure, humidity, and temperature.

SenseHATs have been used with the Pi to create various gadgets, including simple games, weather stations and, as part of a wheeled robot. It is also a fundamental component in the Raspberry Pi-based systems that run the Astro Pi experiments on board the International Space Station.

If you’re so inclined, the $42 / £24 official Raspberry Pi camera is worth picking up too, due to being capable of capturing 3280 x 2464 images, and 1080p@30FPS, 720p@60FPS, and 640x480p@90FPS video.

Also see