[Updated: July 3] — The results are in: The Raspberry Pi 3 is the most desired maker SBC by a 4-to-1 margin. In other trends: x86 SBCs and Linux/Arduino hybrids get a boost.

More than ever, it’s a Raspberry Pi world, and other Linux hacker boards are just living in it. Our 2017 hacker board survey gives the Raspberry Pi 3 a total of 2,583 votes — four times the number of the second-ranked board, the Raspberry Pi Zero W.

 

2017 Hacker SBC Survey References This survey’s launch post Catalog with brief descriptions of all 98 SBCs Spreadsheet comparing key specs of all 98 SBCs Votes tallies for all 98 SBCs

 

Our total of 1,705 survey respondents is just shy of the 1,721 voters in the 2015 survey and about four times more than in our 2016 survey. Our voters — 26 of whom won community-backed Linux and Android single board computers as prizes — selected their favorite community-backed single board computers that run Linux or Android from a catalog of 98 open-spec SBCs. Only 23 of the 98 boards received at least 100 votes (by Borda ranking).

2017 Hacker Boards Survey results
(click image to enlarge; click here for detailed votes data)

Unlike last year, the second and third places also went to Raspberry Pi models — the new Raspberry Pi Zero W and a revised, Cortex-A53 version of the Raspberry Pi 2. The only other Raspberry Pi on our selection list — the Zero — ranked 9 out of 98, and no doubt would have done better if the Zero W hadn’t come along to eclipse it. All the Raspberry Pi boards, including older models not included here, have cumulatively sold over 3 million units, far more than the closest competitor.


This year’s top 10 hacker SBCs by Borda rankings
(click image to enlarge; click here for detailed votes data)

The Raspberry Pi’s success came despite the fact that it offers some of the weakest open source hardware support in terms of open specifications. This, however, matches up with our survey responses about buying criteria, which ranks open source software support and community over open hardware support (see farther below).


This year’s top 10 hacker SBCs by Borda rankings
(click image to enlarge; click here for detailed votes data)

Note that by “votes” we are referring to Borda rankings that combine 1st, 2nd, and 3rd choice rankings (see the detailed list of votes here). Farther below, we also look at how the picture changes when looking only at respondents’ first choices. (Yes, the RPi 3 wins again, and by a 7-to-1 margin instead of 4-to-1.)

A Raspberry Pi SBC has won in all four of our annual surveys, but never by such a high margin. In our 2016 survey with 81 boards, the same RPi 3 out-scored the second-ranked Odroid-C2 by only 387 to 227, and in the 2015 vote with a choice of 53 SBCs, the top-ranked Raspberry Pi 2 beat the BeagleBone Black by 2,455 votes. Back in the 2014 survey, with 32 boards, the original Raspberry Pi Model B beat the BeagleBone Black by 1,136 to 814.

Not only did the Raspberry Pi boards run the table, but they appear to have slowed the momentum of dozens of pseudo clones that offer a Raspberry Pi like 40-pin expansion connector claimed to support RPi add-on boards. Only one 40-pin model from a third party — the Odroid-C2 — ranked in the top 10, and it fell from 2nd to 7th place. Fourth place after the Raspberry Pi models went to the Odroid-XU4, an octa-core Exynos based SBC, which lacks a Pi connector, but recently received a major price cut to $59.

Several RPi-compatible boards appeared in the top 20, and they performed well overall, but they were less dominating compared to last year. Respondents appeared to focus as much on the quality and scope of the community projects — and especially their software support — than on specific SBC models.

The Banana Pi pseudo clones — both from SinoVoip and LeMaker — outshone Shenzhen Xunlong’s Orange Pi boards, the most popular of which was the 28th ranked Orange Pi Zero. Yet the Orange Pi boards outshone FriendlyElec’s NanoPi boards, which scored a high with its similarly minimalist, dirt-cheap NanoPi Neo and Neo2, which together ranked 31st.

Both of these China-based vendors of Raspberry Pi compatible boards shared the highest number of entries, and we had to combine several similar boards to keep them within our 10 board per vendor limit. One reason the Orange Pi boards may have performed better than the NanoPi models is that shipping to the U.S. costs under $4 while the NanoPi starts at around $16, or twice the cost of the Neo itself.

The Orange Pi and NanoPi SBCs, which on paper are among the leaders in price/performance/feature attributes, performed surprisingly poorly. In part this may have to do with spotty software support. However, it could also be that they’re more popular in China than elsewhere, and China ranks only 17th among LinuxGizmos reader home countries, at about 1.4 percent. The survey participation was even lower: only eight respondents came from China, and three of those were from Hong Kong.

2017 Hacker SBC Survey participation by country
(click image to enlarge)

SurveyMonkey is blocked in China, but that still doesn’t explain why there were only nine respondents from Japan, which is 9th ranked in LG readership, two respondents from South Korea (13th ranked), or 10 from Russia (14th ranked). We will look into whether firewalls, language barriers, or other issues are limiting participation, and consider potential solutions. The picture is complicated by the fact that a survey announcement also appeared on Linux.com, which co-sponsored the survey.


2017 Hacker SBC Survey participation by region
(click image to enlarge)

Despite these gaps, 90 countries were represented overall from all over the world. The U.S. amounted to about a quarter of respondents, while other major sources included Europe, India, Canada, Brazil, and Australia. Interestingly, Europe’s share was only slightly less North America’s — but then, it is home to half the boards in the top-ten list.

[quick jumps: SBC descriptions | SBC specs table | all SBC scores]

 
x86 gains and an Arduino resurgence

While there is no single processor vendor that dominates hacker boards, the ARM architecture is clearly in charge, representing 83 of the 98 boards in our catalog. Yet this year’s survey revealed much greater interest in x86-based boards, which totaled eight entries. There were seven MIPS-based boards, including the three Arduino models.

Seco’s Udoo x86 came in sixth place, and Aaeon’s similarly Intel Atom based UP Squared and earlier UP board came in 13th and 17th place, respectively. None of these were in last year’s survey. Last year’s highest ranked x86-based board was Intel’s MinnowBoard Turbot Dual, which fell from 16th place into the back of the pack. The new MinnowBoard Turbot Quad, however, was ranked 26th.

The Intel Edison Kit for Arduino advanced to 18th from last year’s 35th, and the Quark-based Intel Galileo Gen 2 ranked 43rd. However, both products, including the Intel Edison module itself, are being discontinued, according to a June 19 Hackaday story that points to some PDFs located on an Intel server. While the discontinuation of these older boards is unsurprising, the story also said that the newer, Atom-based Intel Joule module is also being discontinued.

There have always been plenty of x86 SBCs, but far fewer models aimed at makers. One reason for the growing popularity of x86-based SBCs sold in single quantities to hackers is that there are more powerful quad-core Atom options available this year. Due in large part to more expensive CPUs, the boards cost more than similarly configured ARM boards, but prices have dropped a bit, and there are relatively few ARM boards that can match the x86 boards’ performance and feature sets. If you want features like USB 3.0, SATA, M.2, or mini-PCIe, you’re far more likely to find that on an x86 SBC.

The survey also shows a resurgence of interest in official Arduino boards that also run Linux. The Arduino Industrial 101 came in 10th, up from its 30th place ranking last year, and the Arduino Yun/Yun Poe and Arduino Tian moved up to 14th and 15th place.

Most of the Linux-enabled Arduinos have come from the more commercially focused Arduino.org wing of the once feuding Arduino camps. It’s unclear to what extent Linux will factor in Arduino’s future. The only major new Arduino board announced since the reunification of the two Arduinos is the non-Linux MKRFOX1200 board. Yet, the new CEO of Arduino is Arduino.org’s Federico Musto, who at one time headed up Red Hat’s European business unit.

In addition to these official Arduino boards, there are many other SBCs in our catalog that offer Arduino shield accessory compatibility. Yet there are far fewer of these than of RPi-compatible SBCs, and there seem to be fewer new Linux boards adding Arduino links.

Along with Arduino boards, x86 boards, and the Pi pseudo clones, another major category of hacker boards includes those that comply with Linaro’s ARM-oriented 96Boards specs. There’s no processor requirement aside from ARM architecture, but the spec requires standardized expansion connectors and a few other basic features. The DragonBoard 410c was once again the leading the 96Boards contender, rising from 9th to 8th place, and the HiKey board came next in 29th place.

A final group consists of BeagleBoard.org blessed BeagleBone spinoffs, which are the closest to true clones here since they use the same TI Sitara processor and share most of the same features. The venerable BeagleBone Black fell only slightly to 5th place, and the BeagleBone Black Wireless came in 11th. The new robotics-focused BeagleBone Blue came in 23rd, and Seeed’s BeagleBone Green Wireless ranked 37th.

Beyond the rise of x86 and Arduino boards in the rankings, it’s hard to see many hard and fast trends in the data. As usual, the more affordable SBCs have did better than pricier boards, and IoT-oriented models aimed at home automation seem to outshine media player platforms, but there are plenty of exceptions. Despite the accelerating Raspberry Pi juggernaut, there’s still plenty of experimentation going on with new board models, and to a lesser extent, new board projects.

Requirements for our contest included support for Linux or Android, an under $200 price at least in base configurations, and a minimum requirement for a few real-world ports such as USB connections. The boards needed to be backed by community resources with optimized open source OSes available for download, and must offer at least some open hardware specifications to enable easier hacking and potentially, open source spinoffs. (See the “Selection Criteria” section of this survey’s SBC catalog, which includes product blurbs and a 98-product comparison spreadsheet for more details.)

[quick jumps: SBC descriptions | SBC specs table | all SBC scores]

 
First-choice comparison

We believe that the Borda ranking, which calculates a total based on 1st, 2nd, and 3rd rankings, is the best indicator of a board’s popularity. However, it’s also interesting to see how the SBCs ranked by order of respondents’ first choices. Almost three quarters of respondents — 74 percent — claimed to have hands-on experience with their primary choices, while only 58 percent had experienced their second choices, and only 46 percent had used their third choices.

Top 10 SBCs based on first choices only

SBC	1st choice votes
1. Raspberry Pi 3 Model B	624
2. Udoo X86	86
3. Odroid-XU4	78
4. BeagleBone Black, Rev C	75
5. Raspberry Pi Zero W	73
6. Raspberry Pi 2 Model B	67
7. Odroid-C2	54
8. DragonBoard 410c	53
9. UP Squared	36
10. Raspberry Pi Zero	34

When ranked by first choices only, the top 10 ranking is very similar, but with some key differences (see above table). The Raspberry Pi 3 is even more dominant here, earning more than seven times the total number of votes of the second-ranked UDOO X86, compared to just about four times more than the second (Borda) ranked RPi Zero W. Here, the Zero W and the RPi 2 both drop a few rankings to make room for the UDOO X86, Odroid-XU4, and BeagleBone Black.

The other major change in the top-ranked list is that the UP Squared jumped into the top 10 list, bumping the Arduino Industrial 101. Other boards that saw their popularity rise in the first-choice rankings include the Banana Pi BPI-M3, BeagleBone Blue, Nano Pi Neo/Neo2, and the Parallella, to name a few.

SBC prize preferences (Borda weighted)
(click image to enlarge)

When respondents were asked which of the six available board or board families they preferred if they won a prize, Intel’s MinnowBoard Turbot Quad was the clear first choice. This is curious considering it beat out several boards that scored much better in the general survey. The second highest preference was for the DragonBoard 410x, and the third was a selection of BeagleBones. The fourth-ranked option was to receive the UP Squared, UP Board, or soon to ship UP Core. True, the Turbot Quad only recently shipped, but the same is true of the UP Squared, which costs even more. The last two prize choices were the non-Linux Arduino Uno WiFi and the Cortex-A8-based Gumstix Pepper DVI-D.

[quick jumps: SBC descriptions | SBC specs table | all SBC scores]

 
Buying priorities: open source software is all important

To earn a chance to win a prize, respondents needed to answer a few more questions beyond ranking. When asked about the key criteria for selecting a hacker SBC, our 1,705 respondents were very consistent with last year’s choices. High-quality open source software again led the list, followed by community ecosystem and networking and wireless I/O.

Most important SBC features
(click image to enlarge)

Open source hardware support ranked fourth, bumping memory and storage farther down the list. Other answers were also ranked much like last year, with standard I/O ports outranking media connections in priority.


Likey SBC applications
(click image to enlarge)

Choices for intended applications were also very consistent with previous years. Home automation was again the clear winner, followed by education, which jumped from fourth to second place. This was followed by home multimedia and special function servers.


Nature of SBC-based projects
(click image to enlarge)

In a separate question about more general usage categories, the maker/hobbyist segment increased to 63 percent, and education rose to 9 percent. Those who considered themselves commercial developers dropped sharply from about 22 percent to 14 percent.

[quick jumps: SBC descriptions | SBC specs table | all SBC scores]

 
Prize winners

Congratulations to the 26 winners of our drawing for one of six SBCs (or selections from a family of SBCs). The prizes this time around include several BeagleBone models, including the new BeagleBone Blue robotics kit. Other giveaways include Arrow’s Qualcomm-backed DragonBoard 410c, the Gumstix Pepper DVI-D, ADI’s Intel-backed MinnowBoard Turbot Quad-core, and several Aaeon UP board models including the new UP Squared and one of the first UP Core boards, which has just launched on Kickstarter. Finally, there was a lone non-Linux giveaway, the nifty Arduino Uno WiFi.

Top left to bottom right: Arduino Uno WiFi, BeagleBone Blue, Qualcomm DragonBoard 410c, Gumstix Pepper DVI-D, Intel MinnowBoard Turbot Quad-core, Aaeon UP, and Aaeon UP Squared.
(click each image for board details)

Speaking of our survey’s prizes, we want to express a giant “thank you” to Aaeon, Arduino, Beagleboard.org, Gumstix, Intel, and Qualcomm for collectively donating the 26 SBC prizes for this year’s hacker SBC survey. We also want to express our gratitude to the 1,705 survey participants who gave a few minutes of their time to help us collect this interesting data on the hacker SBC market.

 
Did you win?

The 26 winners randomly picked from the 1,705 participants in our annual Hacker SBC Survey have now been confirmed. They’re located in 12 countries. Please see our announcement of the 26 winners for a list of the winners and their 12 countries, organized by SBC supplier.

Thank you for participating in this year’s survey, and we look forward to seeing you next year!

 

2017 Hacker SBC Survey References This survey’s launch post Catalog with brief descriptions of all 98 SBCs Spreadsheet comparing key specs of all 98 SBCs Votes tallies for all 98 SBCs