Intel’s big fish in a little pond – A Minnow Interview
We chat to Scott Garman about Intel’s first foray into the world of Linux-powered open hardware development systems
Scott Garman, the man Intel has appointed as the ‘evangelist’ for its first steps into open hardware development, has a history of embedded development. “I’m not a hardcore kernel hacker,” he tells us during an interview to commemorate the launch of the MinnowBoard single-board computer, “but a generalist who enjoys working with the big picture in mind.”
A software engineer for the Yocto Project, a collaborative effort to make the development of Linux distributions for embedded platforms as simple as possible, working under Intel’s Open Source Technology Center, Garman has already given plenty back to the community. He’s responsible for Yocto-related training materials, including one of the most popular screencast introductions to the project, and is now leading the effort to introduce the first open hardware platform designed with Yocto firmly in mind: the Intel MinnowBoard.
The MinnowBoard, owes a clear debt of gratitude to devices like the Raspberry Pi and the BeagleBoard. Designed around Intel’s ‘Queens Bay’ platform, which has at its heart a low-power 32-bit Atom processor, the board is designed as a development platform to bring the flexibility and familiarity of the x86 instruction set architecture to embedded systems – and to do so at a low cost.
Most surprisingly of all, the device is open hardware: all specifications, schematics, board layouts and firmware packages are available for download and reuse under a permissive licence – a first for an Intel hardware product. Clearly, this is a departure from the norm, and one of which Garman is understandably proud.
How did the MinnowBoard come about?
“I’d say development in earnest started happening at the end of 2012. MinnowBoard was an unusual project because of the way Intel and CircuitCo [the company behind the BeagleBoard and BeagleBone development platforms] collaborated on it. The Intel Yocto Project team provided input to help design the requirements for MinnowBoard around performance, openness, flexibility and standards. CircuitCo then used an Intel reference design – the ‘Queens Bay’ platform – and adapted it as needed. They also manufacture and sell the MinnowBoard, so it’s primarily their product, not Intel’s. The result is a win-win situation where our team got a great development platform and more, and CircuitCo has a compelling product platform to sell.
“We’ve been really happy to work with CircuitCo on this, because they understand the open hardware/open community model so well, but we both had plenty of things to learn from each other during the process of bringing up this first board. It’s turned out to be a great collaboration and we’re really pleased with the results.
“In one sense, the Intel team working on this is extremely small. At the same time, I wouldn’t want to discount the contributions that many people have done to help make this happen, even if they weren’t involved with the project full- time on a daily basis.”
What led to the decision to choose the Queens Bay platform, with its ‘Tunnel Creek’ Atom processor, in the MinnowBoard?
“Two main issues led us to use Tunnel Creek CPU for MinnowBoard. One, we were looking for longevity of the platform. Going with an open hardware model means that this is going to be an attractive board for embedded product companies to use and adapt for commercial products. I can tell you from experience, there is nothing more frustrating than developing a successful embedded product and then finding out that your components are about to enter end-of-life. Queens Bay is a platform used in IVI [In-Vehicle- Infotainment] devices, and was designed with a long lifespan – since people tend to own cars for a long time. It still has four to five years left in the product’s life cycle.
“Two, the time to execute was now, so we weren’t about to wait for upcoming Intel platforms to roll out. Tunnel Creek met our main requirements and was available immediately, so we ran with it.”
What advantages does the x86 architecture used in the Atom processor offer in the embedded space, compared to ARM or other RISC platforms?
“Compatibility is likely the biggest advantage. Linux originated on the Intel 386, and whether you’re working on embedded or desktop applications, the toolchain, libraries etc were designed on and are pretty much guaranteed to work on x86.
“Performance is another significant one. Countless person-years have been put into optimising compilers such as GCC to take full advantage of x86 platforms, so you can leverage that to its greatest benefit.”
Do you agree that the success of the ARM- based Raspberry Pi has demonstrated a clear demand for low-cost development boards from hobbyists?
“Yes, absolutely. One of the things I appreciate and respect about the Raspberry Pi project is that it’s introducing so many new people to embedded Linux development. It used to be that embedded was a niche thing that software engineers specialised in, and now the embedded Linux community is becoming much more diverse, especially with younger people. I love that.”
What does the MinnowBoard offer that other, more established development boards lack?
“MinnowBoard stands out in its combination of performance, flexibility, openness and standards. The Intel Atom CPU provides plenty of performance for high computational workloads. It also has strong I/O performance due to PCI Express powering its Gigabit Ethernet and SATA disk features. We also make PCI Express available through MinnowBoard’s expansion connector to enable high-speed I/O to custom hardware projects, for example interfacing with FPGAs or other hardware.
“The MinnowBoard can be used for fun hobby projects, yet scales up to higher workloads. It’s expandable via several PC and embedded bus standards, and offers an environment for custom firmware development.
MinnowBoard includes PC architecture standards including PCIe, USB and SATA, as well as embedded standards such as SPI, I2C, GPIO and even [a CAN bus] for automotive applications. It was designed with the Yocto Project in mind, which is an industry-wide
standard for embedded Linux.”
The MinnowBoard is open hardware – how important do you think that will be to its success in the market?
“The open hardware model is very attractive in empowering your customers and allowing them to innovate in ways you can’t anticipate. Since people are free to reproduce and customise the MinnowBoard, the sky is the limit with this board’s potential. I think this will be fundamental to [its] success.
“Keep in mind that open hardware is meaningful not only if the design files are available, but when all the parts are available through accessible distribution channels and the board’s price is not subsidised. If someone wants to rebuild it, or rebuild something similar, they can at about the cost it’s being sold for now. Not many boards can offer that potential.”
What made Intel get involved in hobbyist-level embedded computing development?
“We wanted MinnowBoard to become a useful platform for Yocto Project development that encouraged experimentation and the pursuit of fun projects in addition to more serious embedded applications, as a board we would use in Yocto Project training courses. And we needed it to be flexible enough so you could do interesting things in a classroom-style setting, which lines up with what hobbyists want, too.”
With the board now in the hands of early adopters, have you seen any particularly innovative or exciting applications for the MinnowBoard emerge from the community?
“It’s still a bit early for me to have good visibility into this, as the board has only been available for about three weeks [at the time of the interview]. As I hear about community projects, I do intend to highlight them on our social media channels.
“I am aware of someone who plans to use a MinnowBoard as part of a quality control system for 3D printers by making use of computer vision capabilities. I’ve also heard that folks in the FGPA design community are quite interested in the MinnowBoard due to its high-speed expansion capabilities. There’s also a group of students interested in building out learning exercises on the MinnowBoard, as a way to advance their embedded design skills.”
Intel has something of a mixed history with the open source community – in particular in failing to release graphics drivers for selected Atom processor models. With the MinnowBoard being truly open, does this signal a shift in attitude towards the open source and open hardware communities at Intel?
“The Open Source Technology Center at Intel is full of incredibly talented, well-known and passionate people who care deeply about openness. In recent years this group has grown dramatically in both size and influence, and it reflects the reality of the increased influence open source plays in the global software ecosystem.
“Intel’s Core graphics have open source drivers that work with hardware acceleration, and the upcoming Bay Trail Atom platform makes use of it. This will address the issue you mentioned with graphics drivers on Atom. I’m looking forward to this, and I think Intel is definitely moving in the right direction in this regard.”
Intel and CircuitCo have published a list of current and proposed add-on boards, dubbed Lures, for the MinnowBoard. Is there a particular add-on you would like to highlight?
“One of the Lures that I’m sure will be quite popular is the Trainer Lure – the one based on an Arduino [microcontroller]. Combining the computational power of MinnowBoard with the embedded input/output capabilities and community of Arduino will open up a lot of interesting possibilities.”
How critical is Linux and the Yocto Project to the MinnowBoard’s success?
“We designed the board to be a Yocto Project development platform, and Yocto produces embedded Linux distros. So it’s pretty core to our purpose. It was the primary motivation behind the project itself – to my knowledge, MinnowBoard is the first hardware platform designed with the Yocto Project in mind. But most people using the MinnowBoard may have no knowledge of or interest in the Yocto Project – that’s okay, too.”
Is the MinnowBoard likely to be the first in a family of open development boards from Intel, or is it merely an experiment for the company?
“There are a lot of people within Intel who are excited about open hardware. I can’t speak [about] any specific future product plans, but I think the future is very, very bright.”