PC/104 Gets a Refresh …or Two

  • Page 1 of 1
    Bookmark and Share

After much discourse about Computer-on-Modules (COMs), let’s turn our attention to the stackable PC/104 architecture. Actually, versions of PC/104 CPUs with stack-through I/O pins in 1992 could be labeled the first COMs, ten years before ETX volumes ever began to ramp. The Small Form Factor arena exists today because of the pioneering efforts of PC/104.

The resemblance between PC/104 and ETX ends right there, however. A stackable architecture thrives because of a healthy off-the-shelf I/O ecosystem. The COM I/O ecosystem consists of circuits provided by third-party design services for each custom carrier board. PC/104, on the other hand, is marked by the availability of hundreds of off-the-shelf I/O cards.

After an extended lull in the PC/104 product evolution, marked only by a sprinkle of end-of-life notices for 386 and 486 class CPUs that have long been obsolete in other form factors, this boutique community is suddenly up in arms over competing notions of how to welcome PCI Express into the architecture. The sudden simultaneous emergence of divergent next-generation standards seems all-too-coincidental.

The harsh realities of the SFF world are to blame. Faced with processors and chipsets handed down from the rapidly churning I/O-and-bus-rich desktop / laptop arena, SBC vendors must evolve their technology with the components at hand. “Small” form factors require choices to be made about which buses and I/O are available for user applications. The PC/104 form factor simply doesn’t have enough space for the entire kitchen sink and oven, much less the added space for cooling them.

Worse still, the desktop and laptop communities can invent new buses and kill old ones at a whim, while embedded OEMs actually want to go into 3-5 years of production with the device they’ve been diligently developing, qualifying and certifying for the past 3 to 5 years. So why not just squeeze all of the buses and I/O, old and new, into a tiny fraction of the space of a desktop or laptop motherboard? Yeah, right.

In one corner, meet the defender: The venerable PC/104 Consortium. In the other, the challenger: The upstart Small Form Factor Special Interest Group (SFF-SIG). One standard emphasizes high bandwidth and a “legacy-free” future with PCI Express and PCI; another prioritizes easy connectivity and “legacy-friendly,” with PCI Express and ISA. At the core of the fight is not just disagreement over implementation of the primary bus for new CPU designs, but the ease of migration from the time-tested designs of the past decade. Considering that the majority of PC/104 I/O cards in production are still ISA-based, the latter may be the more important decision. ISA in 2008? How about 2012? Who would have thunk.

It looks like a ten-rounder, at least. But keep in mind that it’s not about right and wrong. In the embedded market, it’s not always winner-takes-all. Each approach has merits depending on your exact situation. The diverse market can support multiple standards. And many suppliers will hedge their bets and keep one foot securely in each camp, resulting in multiple product lines. Unfortunately, this approach burdens some relatively small manufacturers with double the R&D effort, but at least much of the design and layout will be re-used between the two approaches.

Regardless of which path you take as a user, several things are clear. You are fortunate to have choices, since there isn’t a one-size-fits-all solution in the SFF domain. Choices and competition are hallmarks of the free market. You will be able to address the unique requirements of your application, considering the business needs of what to upgrade and when.

Some might ask why the two sides can’t patch things up and agree to a single standard going forward. Aside from some bruised egos and entrenched positions that would make the Shi’ia and Sunni look friendly, peace would need to be made here very rapidly–before suppliers get too far down their implementation curves with both solutions. We’re clearly on the cusp of it being too late. And in the final analysis, it may be best that both solutions are offered to the market–they do uniquely address different sets of needs, and may both have homes for the long term.

Sure as the sun will rise, products representing both approaches will still be in production a decade or two from now. Look at VME, PICMG 1.x slot cards, and even the venerable STD bus. If you come, they will build it. As always, we welcome your perspective on this topic at