SMALL FORM FACTOR FORUM
Do We Really Need COM Standards?
COLIN MCCRACKEN & PAUL ROSENFELD
Page 1 of 1
Do we really need COM standards? COM (Computer-on-Module) products have taken the embedded market by storm over the past few years. They are becoming the mainstay of the exploding small form factor market. The market has been driven by popular standards such as COM Express, ETX, Qseven and others. Some of these standards have many incompatible sub-standards (e.g., pinout types), effectively driving the number of distinctly different COM architectures into the dozens.
What benefits accrue to the market as a result of this standardization effort? Typically, these benefits would include: a broad ecosystem of ancillary products (I/O boards, enclosures, cooling solutions, etc.), second sourcing, lower pricing through competitive bid for each purchase, and upgradeability to higher performance. All of these are truly worthy of the standards effort.
The trouble begins when examining the performance of COM solutions with respect to these benefits. First, there is no ecosystem per se for COM products. COM products by definition have a single interface to an application-specific baseboard that provides I/O and bus interfaces. Baseboards are custom to each application and/or implementation in form factor and functionality. There are no third-party I/O cards, enclosures, or cooling solutions for COM products (vendor-supplied heatspreaders don’t count). Secondly, the alternate sourcing issue fails due to the widely reported and well documented failure of interchangeability between modules implemented to the same COM standard. The issues, ranging from BIOS incompatibility to power sequencing to undocumented uses of “reserved” pins, mean that it is difficult if not impossible to change modules in production without also making changes to the baseboard.
This failure shoots big holes in the other potential benefits of COM standards—lower pricing through competitive bid as well as the ability to upgrade performance by swapping the COM module. To be fair, some of these difficulties can be mitigated by designing the baseboard from the get-go to support more than one module, but this is rarely done and it frequently proves exceedingly difficult to resolve the differences on a single baseboard. And any BIOS incompatibilities must be resolved by the COM vendors.
This leads to the question of whether we would be better off (or at least no worse off) if every COM manufacturer went their own way with their own unique pin definitions, connectors and form factor. Probably not, because there is comfort in the illusion of compatibility provided by the standards that exist today. And because switching between manufacturers to a different pin definition, connector and form factor is a much more significant undertaking than simply trying to accommodate the differences that exist between modules today. That does not say, however, that what exists today has much value. Let’s face it—every COM solution is a custom solution. Period.
Does this give the green light to COM suppliers to develop their own custom implementations with modified or tweaked pin definitions or form factor extensions?
There’s a better way. Interestingly enough, the answer comes from the chip community where some of these benefits (second source, price competition) have been a big issue for decades. In this community, if true interchangeability/second sourcing is desired, the design is licensed to another chip company to produce. It’s time the COM manufacturers woke up and looked at the best way to perpetuate their technology. Because of the rigid pin definition, there is virtually no opportunity to incorporate any unique features or functionality. Every COM built to the same standard has the same components and layout and the same features and functions, with only those annoying little implementation details discussed earlier that make them different. It’s time for one COM manufacturer to take the lead, do the first design with a new processor / chipset combo, and share the design at no charge with other members of their standards group. Members could take turns with new designs so that no one member is burdened every time. This approach provides a rock solid, ironclad second sourcing and the opportunity to seek alternate suppliers with each purchase. It does not, however, deal with the upgrade issue.
Ever wonder why virtually any memory module (DIMM and SODIMM) that you buy works in virtually every desktop or laptop? Because many are made from the same design shared among memory suppliers. COMs can do this too. Or could it be that COM suppliers only want the illusion of compatibility and not the reality. That they want to avoid second sourcing and direct price competition. Ouch! Hello users—are you out there?