BROWSE BY TECHNOLOGY










RTC SUPPLEMENTS


SMALL FORM FACTOR FORUM

Bring Back the Stack

COLIN MCCRACKEN

  • Page 1 of 1
    Bookmark and Share

The recent Design West trade show in Silicon Valley was filled with all of the usual software and chip announcements, but a faint murmur was detectable from a surprising cross section of the show: Stackable single board computers. While most of the board-level exhibitors still carry this type of product, it takes a discriminating eye to spot these in some booths when surrounded by a sea of other board types and system-level products as well.

What does this say about the state of the union of SBCs that allow I/O cards to be plugged in, parallel to the SBC rather than the right-angle / vertical desktop PC slot type? The predecessor ESC shows featured plenty of expandable SBCs. At first glance, it appears that manufacturers have shifted their R&D investments to faster-growing architectures such as computer-on-modules (COMs) or carrier boards that COMs plug into. In some cases, manufacturers have simply moved along the integration food chain to focus on system-level products, sourcing the SBCs and I/O they need from other branch offices and/or companies. This viewpoint is supported by examining the stackable SBC companies that have been gobbled up during the last 10 years. Many of the SBC R&D teams located in Europe and the U.S. have been re-purposed or shut down completely.

Some of the large board suppliers already have 5 to 10 years of experience helping their system OEM customers migrate away from tall I/O stacks to COMs and custom carriers. This activity pre-dates PCI Express appearing on any COM or stackable SBC. A number of the medical device manufacturers moved from PC/104 or PC/104-Plus to the ETX form factor, which also supports PCI and ISA buses, so such a transition was transparent to peripherals and device drivers. In fact, the very first small form factor (SFF) off-the-shelf production-deployable carrier was from the ETX era, long before standard SFF production carriers for COM Express and Qseven arrived on the scene.

Is this trend a one-way ticket? Hard to say. But some of the PC/104 manufacturers at the show generated buzz by either exhibiting or walking around evangelizing myriad new approaches to bus expansion for SFF SBCs. While each expansion connector has different attributes, a common thread is the willingness to trade-off I/O stack height and flexibility (such as stack direction and legacy buses) in favor of lower costs. The belt-tightening results in instant relevance for other CPU architectures (ARM, etc.), not just for the latest generation Core i7 processor and custom heat pipe.

In fact, size, weight, power and cost (SWaP-C) of the entire system would be reduced for most I/O types, due to reducing the board surface area consumed by bus connectors on every single board in the stack. With these fresh new approaches, the word “mezzanine” comes to mind for I/O, just like the mSATA and PCI Express Mini Card Wi-Fi modules. Imagine an SBC with multiple I/O sites side-by-side, each having a very low cost connector. This can be extended to standard carriers for COMs, in fact some PC/104 companies have now gone this direction already. After all, the very successful COMs are just processor mezzanines (remember PrPMC?). SFFs are finally taking a page out of the decades-old backplane architecture playbook, it appears.

Meanwhile, it’s business as usual down at Stackables City Hall. Some might say this architecture type has given up so much ground that it would be hard to regain market share. An extinct medical market segment white paper comes to mind. But there’s nothing wrong with being ultra-focused on one thing. Most of the new COM and SBC form factors over the years have been launched without a trade group. COMs are for custom carriers (no ecosystem of I/O cards), although interoperability has been greatly improved by the work of competitors cooperating within a trade group. SBC form factors with slot card expansion or no bus expansion don’t need trade groups (consider Mini-ITX, ATX). But historically where off-board interfaces to other boards are involved, trade groups have been essential for safeguarding these coveted interfaces. If not defined clearly and protected, interoperability goes out the window. 

 

It will be interesting to watch the various SBC expansion inventions quietly discussed and debated at Design West. Will any of these become mainstream someday? Maybe blaze a trail without the overhead of a trade group? Or will some go “straight to DVD” or “straight to Netflix” like a bad summer movie?