PC/104 has taken a lot of flak lately. PC/104 in this case means the whole concept of stacking 90 x 96 mm modules – regardless of the bus interface between the modules (represented in SFF-SIG by the SUMIT-ISM standard). PC/104 is one of the most successful architectures in the history of embedded computing, with module shipments well into the millions. What has happened to bring on these attacks from industry analysts, the trade press, and non-PC/104 embedded board suppliers. Are the attackers right, and what does the future hold for this venerable architecture?
COM and PC/104: Boxing gloves raised
There are two main sources for the concern about the future of PC/104. The first involves significant inroads, particularly in higher-volume applications, of the Computer-On-Module (COM) architecture. COMs are tiny CPU modules designed to plug into a custom, application- specific baseboard. (Interestingly, this was one of the original uses for PC/104 CPUs.) The COM provides the bus interface and basic I/O to the baseboard. The baseboard adds the unique I/O required for each application. COMs are typically smaller than PC/104 CPUs, less costly, and much easier to use on a baseboard because the module-to-baseboard I/O interface is standardized and well defined. COMs have had widespread acceptance in high-volume applications, particularly where the OEM has board-design expertise in-house.
PC/104, however, continues to retain an advantage by providing a shorter time to market and much lower development cost. PC/104 has retained a strong position in lower-volume, high-complexity, and higher-value applications. And the PC/104 ecosystem is enormous, encompassing off-the-shelf I/O cards, enclosures, and more. A large number of embedded applications simply cannot use a COM solution because the OEM does not have the resources or budget to design a baseboard. Murmurings of PC/104’s demise are dead wrong.
PC/104: Kicked off the bus?
The other significant reason for the attack on PC/104 is that the architecture has stumbled in its move to adopt new bus technology – in this case PCI Express. PC/104 was originally designed around the ISA bus, and ISA-based PC/104 modules continue to dominate the ecosystem, ship in very high volumes, and receive new design wins. In the late 1990s, PC/104-Plus was created to add the PCI bus to PC/104 modules. PC/104-Plus has achieved only a fraction of ISA-based PC/104 shipments.
In the past couple of years, significant differences among suppliers regarding how to add PCI Express to PC/104 stacks have resulted in two different standards (one from the original PC/104 Consortium and one from SFF-SIG). While a number of suppliers have moved to implement each of these standards, the market has been fragmented and the creation of a deep ecosystem has fallen behind. Recent concerns about the original technical implementation of the PC/104 Express solution might yield yet a third standard, further fragmenting the market.
In contrast, SFF-SIG’s SUMIT-ISM solution raises no such concerns. SUMIT-ISM is now a proven, robust, next-generation stackable architecture providing a multibus, interboard interface with PCI Express x1 and x4 lanes, USB, LPC, SPI, and I2C interfaces on the SUMIT connectors. Plus SUMIT-ISM supports both the PC/104 (ISA) bus and PC/104-Plus (PCI) bus to protect legacy investment. More than a dozen SUMIT-ISM products have already been brought to market with more on the way. In view of the concerns raised about the future of PC/104, SFF-SIG members ask the PC/104 community to rally around SUMIT-ISM as the sole future architecture for PC/104 to help build the ecosystem as rapidly as possible. By uniting around SUMIT-ISM, this community can take the steps to retain and grow the overall PC/104 market into the future and stave off further erosion.
PC/104 lives on
Regardless of future path, PC/104 is far from dead. Legacy PC/104 applications will continue to ship in high volume for years to come. OEMs with a short time to market or without in-house board design expertise will continue to depend on PC/104 solutions. But OEMs with new designs requiring the higher speeds of PCI Express might pause to consider other architectural approaches unless this community can come together around a single solution. It’s gettin’ late. The time to do this is now.
Small Form Factor SIG 408-480-7900 firstname.lastname@example.org www.sff-sig.org