Almost all embedded engineers today will have some familiarity with the Arm reduced instruction set (risc) architecture; it is arguably the most widely licensed and implemented third party IP in the industry and, in one form or another, is present in the portfolio of some 40 different Arm licensees – predominantly semiconductor vendors.
Just as many engineers will be familiar with Intel's x86 family; the instruction set architecture (ISA) that empowers the majority of desktop and laptop computers today. But as the x86 is predominantly used in PC platforms – which aren't considered to be deeply embedded – it may be known more by reputation than experience amongst embedded engineers.
Both architectures are massively popular and their application areas have predominantly differed in the past, but they do have some similarities. It's true to say that, while an Arm-powered processor lies at the heart of almost every mobile phone in circulation today, it isn't the Arm brand that sells the phone. Similarly, and while Intel's persuasive advertising jingle may suggest otherwise, it is the OEMs who sell PCs, not necessarily the Intel architecture inside.
From that point of view, the companies behind these popular and, some would say, equally successful architectures have in the past distanced themselves from the front line in terms of their customers' customers. However, recently those boundaries have become less defined, as both companies are now vying for a slice of the industrial embedded control market and to achieve that, they must now promote themselves as the right product for embedded applications. So which one wins?
When embarking on a new project, the choice of architecture is huge, particularly if the application requires something relatively simple. If that is the case, chances are it can be accommodated by a low cost 8 or 16bit architecture, of which there are many. Perhaps king among these is the venerable 8051, which is now available in both 8 and 16bit forms, from a number of licensed manufacturers. There are also a great number of proprietary architectures available, which have the full support of their respective parent companies and strong roadmaps for future development.
Any engineering team would need a good reason to move away from any one particular architecture in favour of another, and consideration for future support plays an important part in that reasoning. Arm has been pursuing the microcontroller market for some time and offers a number of excellent cores to meet its needs. To some extent it has battled against 16bit architectures on unequal ground, but it is inescapable that an increasing number of engineers are realising the benefits of moving to 32bit architectures; manufacturing costs are now so low that the extra horsepower and peripherals far outweigh the slightly higher unit prices.
Recently, Intel launched the Atom processor family. As it is based on the same x86/IA-32 ISA it is capable of natively running the familiar desktop oriented Windows operating systems; something the Arm architecture remains incapable of doing – at least for the moment. But while Windows compatibility has been significant in one market sector where the giants are now competing – the netbook market – it hasn't traditionally been important in the deeply embedded market. For those applications that do not require a Windows experience, the choice would appear to be simple. However, it is the emerging 32bit industrial market that has attracted Intel's interest and it is here that they are set to clash with Arm.
The target application has always been a prominent consideration when selecting a device for industrial use, mainly because if you know the mix of power and peripherals the application needs, it is possible to select a device that will almost exactly meet them. Here, the Arm architecture wins hands down over the Atom; thanks to the large (40+) vendors using the architecture, an engineer is almost guaranteed to find a device with just the right mix of onboard processing power, memory, IO and peripherals. As a result, it is very widely supported with a large base of freely available code.
Although the Intel device has less flexibility, it provides greater processing power than a low-end Arm device, with a standardised interface. This leads to relatively simple integration but it may not be perfect for every deeply embedded application. Additionally, with embedded applications in mind the Atom has a range of sophisticated low power modes, which for the first time make Intel's offering more comparable to the advanced power saving techniques used in their competitors' embedded processors. Atom's enhanced low-power states (C1E, C2E, C4E) use Intel's Speedstep technology, which forces the processor's core to run at a lower frequency and lower voltage during inactive periods, under software control.
The Atom has several other significant advantages. Intel devices have strong support from industrial PC providers, specifically in single board computers (SBCs) such as those from Congatech and AValue. Many industrial applications are relatively low volume and so fit well with the cost model of using an SBC. They come largely preconfigured and have the additional benefit of essentially being a PC. This means the development of the application software can be carried out on a PC and easily ported to an SBC. For many applications this will provide significant cost savings in terms of the time required to develop an application.
Another consideration is the use of Microsoft's embedded platforms, many of which will also run on an Arm processor (such as Windows Mobile or Windows CE), but there is one that will still only run on an x86 architecture; Windows Embedded Standard. This is a componentised version of the desktop platform, which allows it to be optimised for a specific application, using only those components needed by the application (of which there are in excess of 12,000). While this requires some experience of Windows Embedded Standard, it does deliver a more optimised platform for deeply embedded applications.
For Arm there are far fewer SBCs though vendors support their Arm implementations with hardware and software development kits. However, with a strong eco-system of software providers targeting the embedded space, and an extensive network of silicon providers to choose from, configuring an optimised hardware and software product is arguably much simpler when using an Arm-based processor.
It may seem clear that Intel is more focussed on the commercial sector and for its mainstream, powerful and power-hungry processors that is true. However, the Atom is the first family from Intel to support the industrial temperature specification; when it was launched in early 2009 there was a flurry of complementary announcements from SBC vendors announcing their support for the Atom, in boards that could support the industrial temperature specification. This is significant because up until the Atom, SBCs based on an Intel device were either restricted in their exposure to industrial temperature ranges (which required expensive cooling systems) or they needed to be individually screened to ensure they could function at industrial temperatures – which increased their price. With the availability of the Atom processor, SBC vendors are able to provide boards that were specified to industrial temperature ranges at much more competitive prices. It also provides a route to higher volume deployment, where the OEM moves from an SBC to a bespoke hardware platform based on the Atom.
Intel and Arm share a common goal, to penetrate the deeply embedded market further. For Arm this is familiar territory but it still faces fierce competition. Intel is less familiar with competition and may not find the territory quite so welcoming. However, while both companies continue to develop leading edge products for the embedded market, it is the engineering community that is sure to win.
Odo Akaji is product manager for semiconductors at RS Components