Benchmarking the Raspberry Pi 3 Model A+

Back in March, the release of the Raspberry Pi 3 Model B+—the Pi 3 B+ to its friends—brought a chance to take stock and review just how far the project had come since its launch via a series of benchmarks. Now the launch of the Raspberry Pi 3 Model A+ brings a bold claim: a dramatic drop in size, weight, and price over the Pi 3 B+, but without any loss in performance.

In other words: it’s benchmark time once again.

Specifications

SoC: Broadcom BCM2837B0 quad-core A53 (ARMv8) 64-bit @ 1.4GHz
GPU: Broadcom Videocore-IV
RAM: 512MB LPDDR2 SDRAM
Networking: 2.4GHz and 5GHz 802.11b/g/n/ac Wi-Fi
Bluetooth: Bluetooth 4.2, Bluetooth Low Energy (BLE)
Storage: MicroSD
GPIO: 40-pin GPIO header, populated
Ports: HDMI, 3.5mm analogue audio-video jack, 1x USB 2.0, Camera Serial Interface (CSI), Display Serial Interface (DSI)
Dimensions: 67mm x 56mm x 11.5mm, 29g

It’s not hard to see where the design of the Pi 3 A+ differs from the Pi 3 B+: its smaller form factor, last seen in the original Raspberry Pi Model A+ which launched four years ago, has led to the removal of the Ethernet port and associated combination LAN/USB chip, which in turn drops the available USB ports down to just one with a direct connection to the SoC. The loss of the Ethernet port has another knock-on effect, removing support for Power over Ethernet (PoE)—available on the Pi 3 B+ via an add-on board.

Raspberry Pi 3 A+ RAM Chip

The memory, too, has been halved over its larger stable-mate—for reasons of price, the Raspberry Pi Foundation’s Eben Upton explains, not board space—down to 512MB. The system-on-chip, though, is entirely unchanged: it’s still the B0 spin of Broadcom’s BCM2837, which means it’s a quad-core Arm Cortex-A53 running at 1.4GHz and a Videocore-IV GPU sharing system memory.

Impressively, the Pi 3 A+ includes the same radio module as the more expensive 3 B+, making it the first Model A/A+ to include any form of on-board networking. It’s not the cheapest Pi variant with wireless connectivity, though: that title goes to the Pi Zero W. Moving up to the Pi 3 A+, though, gives you a considerable boost in performance plus analogue audio and video output as well as full-size USB, camera (CSI), and display (CSI) ports.

Raspberry Pi 3 A+ Radio Module

The move to a smaller form factor has an obvious effect on size and weight, addressed later, but brings with it one point of concern: the potential for a loss of performance due to there being less circuit board mass into which the SoC can bleed heat.

Thermal Management Benchmark

Raspberry Pi 3 B+ Thermal Photograph
Raspberry Pi 3 A+ Thermal Photograph

Any concerns over the smaller circuit board leading to higher core temperatures, though, are quelled by a quick look at the Pi 3 A+ and Pi 3 B+ under a thermal camera. These images, taken after ten minutes of a CPU-heavy benchmark, show that the SoC on the smaller Pi 3 A+ actually reaches a marginally lower peak temperature compared to the Pi 3 B+—likely a result of the removal of the LAN/USB chip, to the right on the B+, which shows up as a clear contributor of considerable heat even when no Ethernet cable is connected.

At the same time, the Pi 3 B+ image can be compared to its equivalent in the original benchmark to show how changes made in the thermal management firmware post-launch have helped to bring down the SoC’s temperature considerably—improvements which also benefit the Pi 3 A+.

Whetstone Benchmark

The classic Whetstone benchmark measures a processor’s floating point performance, in Whetstone instructions per second (WIPS). Here the performance is given in millions of WIPS, or MWIPS. It’s a promising start for the Pi 3 A+, here, returning a result well within the margin of error which puts it neck-and-neck with the larger and more expensive Pi 3 B+. The Foundation’s promise of no performance loss between the models would appear to ring true.

Dhrystone Benchmark

Where Whetstone offers floating-point performance measurement, Dhrystone looks at integer performance. The measurement is given here in millions of instructions per second, or MIPS. Again, allowing for expected variances between two distinct boards, it’s even-Stevens between the Pi 3 A+ and Pi 3 B+.

SysBench CPU Benchmark

The SysBench application offers a range of synthetic benchmarks, and its CPU test demonstrates the difference multi-threading can make. Here two measurements are given: the number of seconds it takes to complete the benchmark in single-threaded and multi-threaded modes, with the latter only available to the quad-core Pi 2 and Pi 3 families.

The Pi 3 A+ is still level-pegging with the larger Pi 3 B+, but a more interesting comparison here is to the original Pi A+: between the processor itself being both faster and more efficient and the presence of three additional processing cores, it’s clear that any project currently using a Pi A+ could see a serious performance boost using a Pi 3 A+ as a drop-in replacement.

Linpack Benchmark

Designed for use in high-performance (HPC) computing environments, Linpack is a more generalised test than either Dhrystone or Whetstone. Here three readings are taken: double-precision complexity, single-precision complexity, and a special version of the single-precision benchmark compiled to use the NEON accelerated instruction set available on the Pi 2 and Pi 3 families. All measurements are in millions of floating-point operations per second (MFLOPS), but don’t represent an absolute maximum performance measurement: with software optimisations specific to particular models it’s possible to get considerably higher figures.

Still no surprises: the Pi 3 A+ draws level with the Pi 3 B+, allowing for expected board variance.

Python GPIO Benchmark

Many Raspberry Pis spend much of their life running Python applications, so this custom benchmark measures each model’s capability at exactly that workload. A simple program switches a pin on the general-purpose input/output (GPIO) port on and off as quickly as possible, while an external frequency counter measures its switching rate. The result is express in kilohertz (kHz), with 1kHz equalling a thousand switches per second. It’s a fairly synthetic test—it’s unlikely you’d actually want to simply turn a pin on and off repeatedly without doing anything useful in-between—but does offer a hint of how shifting to each model up the rankings will affect CPU-bound GPIO applications.

Here, we see the first surprise: the Pi 3 A+ appears to offer a marginal, though measurable, improvement over the Pi 3 B+. It’s not enough to get excited about, though—and nowhere close to the difference between the original Pi A+ and Pi B+—but it’s worth noting, and could be chalked up to the system having slightly less load at idle due to the removal of the LAN/USB chip.

SysBench RAM Benchmark

The CPU isn’t the be-all and end-all of performance, though, and here the SysBench utility measures the memory throughput for reading and writing data. Figures are given in megabytes per second (MB/s). A shift from 1GB of RAM to 512MB of RAM could, potentially, have resulted in a performance drop; thankfully, that’s not the case with the Pi 3 A+ again proving itself the equal of the Pi 3 B+.

Matched performance between the models, though, is only a guarantee for workloads that don’t exceed 512MB of RAM; if you’re running something bigger, like a web browser with a fistful of open tabs, you’ll find things slow down considerably sooner on the Pi 3 A+ than the Pi 3 B+. Naturally, if your workload exceeds 1GB of RAM, then you’re going to be hitting a performance wall on either model.

Power Draw Benchmark

Performance comes at a cost, and the power draw benchmark shows exactly what that cost is: energy usage. Each model is configured with an HDMI-connected display, USB wireless keyboard dongle, plus Wi-Fi and Ethernet connected on models where they are available, then two measurements are taken: one where the Pi is sat idle at the Raspbian desktop, and another while running a CPU-intensive application.

Here we see the first signs of the Pi 3 A+ drawing away from the Pi 3 B+: the loss of the Ethernet port and reduction in USB ports, together allowing for the removal of the power-hungry LAN/USB chip, sees the power draw at both idle and load dropping considerably—impressive, given it retains full compute performance. For those looking to upgrade a project from an original Pi A+, though, it’s clear from the results that power draw may prove a problem.

Footprint Benchmark

This benchmark doesn’t rely on any software: it’s simply a look at the footprint of each board, measured in square millimetres. The footprint is taken from each board’s widest points—to include connectors, like USB and HDMI, which sit proud of the circuit board itself—and represents an absolute maximum size. Anyone who has seen the various models side-by-side will find nothing surprising here: the compact Pi Zero range is by far the smallest, with the Pi 3 A+ level-pegging with the original Pi A+ as a mid-range option.

A look at the footprint of the original Pi Model A and Model B is interesting, making clear the size improvements that were made when the Model A+ and Model B+ launched the Raspberry Pi’s new, and enduring, design with its enlarged 40-pin GPIO header yet more compact layout.

Weight Benchmark

They say size isn’t everything, and for embedded projects that’s true: from high-altitude ballooning to wearables, weight is a key consideration when choosing a single-board computer. With the Pi 3 B+ tipping the scales at 50g—making it, as a point of interest, the heaviest Raspberry Pi model ever commercially released, beaten only by the limited-run Alpha prototypes—there was clear room for improvement.

While the Pi 3 A+ retains the thicker circuit board and metal shielding of the Pi 3 B+, the component and footprint reductions have it weighing in at a svelte 29g. That’s heavier than the original Pi 3 A+—and, naturally, the tiny Pi Zero and Pi Zero W—but for many projects the performance gains may be worth the trade-off.

Conclusion

The Pi 3 A+ has launched with the promise of all the power of the Pi 3 B+ in a smaller, cheaper design. The benchmarks certainly bear that out: at no point does the Pi 3 A+ underperform compared to the Pi 3 B+, which given the size, weight, and power draw reductions is no small achievement.

Raspberry Pi 3 A+

The retention of the radio makes the Pi 3 A+ a more promising device than the original Pi A+, too: while it still only has a single USB port, you don’t need to tie that up with a Wi-Fi or Ethernet dongle; you don’t even need to use it for a keyboard or mouse, if you’re planning on running one as a desktop, as you can easily connect both via Bluetooth to leave the USB port free for other purposes.

Having full-size ports is also an improvement over the Pi Zero family, removing the need to buy adapters for the USB, CSI, and HDMI ports—thus offsetting some of the cost difference between the Pi Zero W and the Pi 3 A+, in case the considerably improved performance wasn’t enough for you. The larger board also adds in the ability to connect analogue speakers, a composite video display, or devices like the Raspberry Pi Touchscreen Display via the DSI port missing from the Pi Zero range.

Raspberry Pi 3 A+ DSI Connector

If you’re looking for a Raspberry Pi which offers you as few limitations as possible, the Pi 3 B+ is still the model to buy; if you want something smaller, lighter, lower-powered, and cheaper, but don’t fancy compromising on performance, the Pi 3 A+ is a clear winner.

2 thoughts on “Benchmarking the Raspberry Pi 3 Model A+

  1. Andre says:

    There aren’t any permormance difference between 3B+ and 3A+, even the 3A+ with half the memory? None of the bechmarks use RAM memory, only CPU?

    • Gareth says:

      You’ll see performance differences if your workload exceeds free RAM on either model; these benchmarks concentrate on the CPU performance and memory throughput, not total memory size. If your in-memory workload is over 512MB, you’re going to need the 3B+; if it’s over 1GB you’re going to need the 4B; if it’s over 4GB, the Raspberry Pi isn’t a suitable device for what you’re trying to do.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.