We have an inside joke among the editors here at OpenSystems Publishing: "If a form factor isn't five years old, it's not ready." This tenet has proven true repeatedly over the course of our collective careers in designing products, marketing solutions, and observing activity in the embedded computing industry. It seems that until an embedded form factor turns about five, it just doesn’t reach critical mass.
Paradoxically, even though critical mass is difficult to reach, the number of form factors being developed is increasing, not decreasing. We‚Äôve been expecting form factor consolidation, but we are seeing just the opposite - proliferation. So what‚Äôs driving this growth? And how are vendors responding to it?
Needs are diverging
The answer depends at least partially on underlying usage needs. Some form factors primarily target a specific application segment, with features tuned to optimize design and performance characteristics in that usage. Some are designed to capitalize on a particular aspect of technology, like an expansion bus. Some are designed for lower cost, smaller size, or more board space. Some exist simply because a vendor wants to drive a popular product into a standard.
The problem for designers - and vendors and OEMs - is one size cannot fit all. What works well in a rack-mount server won‚Äôt work at all in a small mobile device. What works well in a set-top box won‚Äôt necessarily work well in an automobile infotainment system.
Beyond that, in this age when every OEM wants customized products at commodity prices, the needs diverge even further into each company‚Äôs specific product offerings. Vendors are tempted to design custom products and try to find other buyers to spread out the development costs and reduce risks.
CEMs can change the equation
Needs are one thing; successfully filling them is a different proposition. Another issue driving form factor proliferation is that almost anything can be built at a Contract Electronics Manufacturer (CEM). Gone are the days when a company was required to set up manufacturing lines in-house, forcing standardization on a few form factors. Now, designs initiated in-house can be handed over to a CEM for manufacturing, as illustrated in Figure 1, courtesy of MHM.
CEMs provide ultimate flexibility in manufacturing products - it‚Äôs what they do. All shapes, sizes, and technologies are in play. Customized runs usually aren‚Äôt a problem if the OEM is willing to pay set-up charges or guarantee volumes.
While working with CEMs may lower the technical barriers of form factor designs, doing so may heighten the financial barriers. Depending on the climate and relationship, getting a CEM to run in small volumes and short lead times may be problematic. Some vendors that have been through this cycle returned to creating small in-house manufacturing lines for this reason, giving them control over both the design and delivery variables.
All good, until life-cycle costs are considered
But just because almost anything can be designed and built to meet a specific need doesn‚Äôt mean that it should be. Life-cycle costs are the embedded industry‚Äôs trump card that swings the balance back in favor of standard form factors.
In a commodity space like the venerable ATX board for 2U servers, low-cost vendors release new products very rapidly, usually as fast as new processors are introduced. As long as the hole patterns don‚Äôt move and connector locations don‚Äôt change much, an old board can be readily replaced by a new one with little or no change in the packaging.
However, in the embedded space, change means pain. Any experienced engineer knows that the problem isn‚Äôt usually getting an SBC that can meet the requirements for the project. It‚Äôs getting the rest of the system - power supply, chassis, I/O boards, storage, networking, and more - to integrate and stay integrated over the life of the program, often five years but increasingly seven, 10, or even more years.
I remember when I was a product manager proudly touting our upgrade plan for a VME SBC, so the new board would drop in the customer‚Äôs unit. During one visit, a customer took me to his lab, showed me his system, and asked, "What‚Äôs your plan for the other eight boards in my system?" As good as our board solution was, and as good as our relationship with that customer was, our solution didn‚Äôt take enough risk out of the change for him to consider an upgrade to the entire system.
If the entire system is implemented on a single board, the problem is simplified, but many systems can‚Äôt be; in fact, designers rely on form factor standards‚Äô openness to add features and functions. A form factor standard‚Äôs interoperability is important, but as I mentioned before, reaching critical mass and staying there for the embedded life cycle is more essential to a form factor‚Äôs ultimate success or failure.
Vendors looking for answers
Designing a bunch of different boards to meet a range of opportunities is obviously not as easy as designing a single board. Some companies with high volumes can afford to do in-house design and life-cycle maintenance, but many companies are turning to embedded vendors for answers to form factor problems. The vendors, in turn, are searching for answers that yield the best financial results.
Small vendors usually can‚Äôt make a wide range of products, but they can excel at making relatively small volumes of devices in a reasonable range of form factors. They tend to win on responsiveness, quickly addressing new requirements for "markets of one." VadaTech‚Äôs president Saeed Karamooz says one of the company‚Äôs recent AMC designs was executed and delivered to the customer in less than 30 days - starting with a phone request on December 19 (while the rest of us were gearing down for the holidays).
VadaTech‚Äôs keys to speed are using a standard qualified component list and keeping in-house manufacturing simple and fast, with quality levels that can be closely monitored. The company then churns out designs quickly to meet opportunities as they present themselves, customizing to win. In reference to the earlier point about CEMs, Karamooz confirms that he looked at that option, but since it didn‚Äôt provide enough control over volumes and delivery needs, he decided to go with in-house manufacturing instead.
Large vendors can produce a wider range of products but have to worry about achieving higher volumes and lower costs. Opportunity costs skyrocket in engaging the wrong project that doesn‚Äôt sell well in the market. To prevent that, large vendors measure markets before committing to a form factor and monitor market share after products are launched - both tasks several embedded market analyst firms freely admit are becoming difficult to do accurately.
The small ties that bind
Embedded vendors have to pay particular attention to their available silicon, looking at features, costs, and overall life cycle. Small form factors cause vendors to be even more selective.
Glenn Beck, industrial segment marketing manager at Freescale Semiconductor, says the company surveyed the small form factor space and chose COM Express for a development kit around its MPC8360. As to the reason why Freescale settled on COM Express, he explains that it wasn‚Äôt so much the benefits of COM Express as a Computer-On-Module per se, but the fact that it is a popular form factor and has a connector that can support enough pins for the MPC8360‚Äôs functions.
Christine Van De Graaf, product marketing manager for the embedded modules division at Kontron, states, "Even though you can pack in more, sometimes it means having to pay more ‚Ä¶ and customers don‚Äôt want to pay a whole lot more to reduce power by a watt or so. We can‚Äôt adopt every CPU variation that comes out. We‚Äôre only adopting the ones that make the most sense."
Looking for the ones that make sense, Kontron is diversifying its processor choices, working with Intel, AMD, and VIA Technologies for processors fitting different needs. Kevin Rhoads, VP of business development for the systems division at Kontron, remarks, "We look for solutions that give the right mix of I/O and processing power at the right price for the application. That‚Äôs why we‚Äôve introduced our first Mini-ITX AMD solution."
More room for more horsepower
For applications with more space, the choices get less constrained in terms of size, power, and cost - performance moves up to the top of the list. Larger form factors like AdvancedTCA can handle big processors, more of them, and both general-purpose processors and application processors.
"Networking processing is a major requirement for most of the customers we talk to," asserts Sven Freudenfeld, Kontron‚Äôs director of business development for telecom in North America, citing Cavium devices. He also thinks multicore processing is a strong fit in applications such as IPTV. "Our third-generation AdvancedTCA blades include three dual-core Intel processors. With that approach, you can get a lot of parallel processing done."
Todd Etchieson, product manager of AdvancedTCA for RadiSys, agrees and notes that IPTV requires deep packet inspection. RadiSys is also looking at Cavium network processing on its blades. He concurs that the five-year timeframe is about right, because that‚Äôs about the time when a platform‚Äôs readiness (for example, AdvancedTCA, which is just turning five) and OEMs‚Äô willingness and ability to port applications to it intersect.
Bigger form factors also can handle heavier operating systems and even virtualization with bigger processing, memory, networking, and storage support.
Consolidation not near
Just scanning around, eyeing the occupants in this crowded room, one can tell that the range of form factors isn‚Äôt going to consolidate anytime soon. Given choices and time, engineers will tend to optimize heavily. The trend toward using CEMs is making designing and building both standard and proprietary form factor boards possible (but maybe not financially attractive) for more entrants.
Larger vendors will probably continue to look for safety in numbers, picking form factors that have significant business tied to them or those that seem to be building interest. Buyers wanting to reduce risk will go with those choices, seeking areas supported by multiple vendors.
Smaller vendors will certainly continue to capitalize on gaps where the larger vendors don‚Äôt play as much, servicing needs on quick-hitting projects where the payoff is near. They‚Äôll also tend to pick on expansion form factors, which can be used with a large range of products. Buyers looking for fast customization will probably opt for smaller vendors.
Everyone should take a good, hard look at using industry standards when they choose a form factor. Why wouldn‚Äôt an industry-standard format available from multiple sources do the job? When is the right time to jump on a new form factor? Why does a legacy form factor that is losing ground in the market have to be used, when switching to a more modern choice on the upswing has long-term life-cycle cost benefits? Taking a longer-term perspective in choosing a form factor when the product life cycle is longer isn‚Äôt an option - it‚Äôs a requirement.