By Jeff Child, Contributing Editor
MARCH 15--The dismal national economy wasn't enough to keep exhibitors and attendees from trekking to Long Beach, CA, last January for the Bus&Board/2002 Conference www.busandboard.com). Indeed, attendance totaled 368, some 30% more than last year's count, making for a "standing-room-only" show. Late registrants had to be turned away. Meanwhile, the number of exhibiting companies rose to 46, more than 20% over last year's tally.
Codeveloped by the VMEbus International Trade Association (VITA; www.vita.com) and the PCI Industrial Manufacturers Group (PICMG; www.picmg.org), this small, but important, conference has become the key information-sharing event of the embedded modular computer board industry. "I think Bus&Board was a watershed event this year," says Ray Alderman, organizer of the conference and executive director of VITA. "With all that happened last year, I think everyone felt a strong need to come and see the new technologies and to see each other."
Sixteen presentations spread over two days spelled out the major activities and trends in markets, technologies, and specification development. The conference indicated signs that the embedded bus-board industry is on track to shift more emphasis to markets that have an image-processing or machine-vision twist, such as industrial automation, medical electronics, military applications, and others. Behind that shift was the dramatic bursting of the great telecommunications and dot-com bubble over the last year.
During the past several years, several board vendors have invested tremendous resources in telecom/datacom products and technologies, primarily CompactPCI-based, while several others had completely abandoned noncommunication-type markets. That said, market analysts presenting at the show indicated opportunities in all market segments and forecast overall growth in business.
This upward market shift was enumerated during the market-research portion of the presentations. For his part, Warren Andrews at Tech Trends Research (Bonita Springs, FL) forecasted steady growth ahead for the overall high-end embedded modular computer market. His forecast showed last year's dip recovering this year and then doubling from $4 billion this year to $8 billion by 2005. The percentage of that market going into communications applications got knocked down significantly from 40% in 2000 to a mere 18% this year, according to Andrews. Moreover, communications, while expected to enjoy continued growth, isn't expected to regain any kind of dominance in the embedded board market for the foreseeable future, with a predicted climb to 25% of market share by 2005.
According to Andrews, it's not the military/aerospace segment of the market that will fill the gap left from the communications crash. Following its high of 35% market share last year, the Mil-Aero portion of the embedded board market is forecasted to slowly decline over the next few years. Rather it's the 'everything else' category, comprising medical, industrial-control, transportation, environment, test/measurement, scientific, and security segments, that will capture the largest share of business, at around 50% share. As part of that category, security, says Andrews, is beginning to look like a major growth area driven by post-Sept. 11 conditions.
Perhaps the most dramatic technology news at the conference was the new initiative by the Motorola Computer Group (Tempe, AZ) to reinvigorate the VMEbus board arena. In his presentation, Jeffrey Harris, Motorola director of research and system architecture, detailed the company's plan for a so-called VME Renaissance strategy. In many ways, the move represents a shift within Motorola, as much as a shift in the market. As one of the co-inventors of VME 20 years ago and a long-time market leader in the VME market, Motorola is an influential player in VME. But like many other major embedded board vendors, Motorola put the bulk of its efforts in the last few years on CompactPCI board products and technology and a focus on communications market opportunities. In fact, Motorola hasn't introduced any VME product in more than three years.
The first step in the VME renaissance is Motorola's initiative to launch a PCI-X to 2eSST VMEbus bridge code-named "Tempe." The Tempe chip will implement the 2eSST protocol, which was established as an industry standard by VITA. The protocol allows the VMEbus to run at a bandwidth of 320 Mbytes/s, giving the bus an 8X performance increase over VME64's practical speed.
Supporting existing VMEbus protocols, the chip is designed to be backward-compatible with existing VMEbus cards, enabling existing cards and new Tempe-enabled cards to work together in the same system. The Tempe chip has a PCI-X bus host-side interface running at up to 133 MHz, which provides transfer rates of up to 1 Gbyte/s. This is a 2X improvement over a 64-bit/66-MHz PCI interface.
Accompanying the Tempe chip is a new set of bus transceivers from Texas Instruments (Dallas, TX). These new transceivers, coupled with Motorola's Tempe chip, will allow Tempe-enabled boards to achieve 2eSST speeds in existing VMEbus backplanes. Motorola doesn't plan to make and sell the Tempe chip. The chip is expected to be available to all industry players through a third-party reseller in the fourth quarter of 2002. Motorola offers a white paper on the Tempe chip on the Web at www.motorola.com/computer/tempe.
Motorola's unveiling of its "VME renaissance" plan didn't go without criticism from within the VME industry. Ben Sharfi, president and CEO of VME board-vendor General Micro Systems (Rancho Cucamonga, CA), criticized Motorola for not telling the whole story about other 2eSST protocol implementations. Two years ago at Bus&Board 2000, General Micro demonstrated its own OmniVME system, offering 720-Mbyte/s throughput over 21 slots using 2eSST chips from Arizona Digital (Scottsdale, AZ).
While the VME renaissance bodes well for high-end VME image-processing applications, particularly those in the defense market, the vast majority of vision-system designers are more interested in the future of bus architectures beyond PCI. As of now, PCI-based frame grabbers and image processors comprise the majority of the mid- and low-vision market. Like everyone in the PCI board-business, makers of such boards must base their future product plans on the question: What interconnect technology will become the next dominant interconnect scheme beyond PCI?
Attempting to address that question were a set of presentations at the conference on chip-to-chip interconnect alternatives and switch-fabric technologies. Chaired by Rick O'Connor, CTO at Tundra Semiconductor (Ottawa, Canada), this section of the conference explored the various alternatives including InfiniBand, RapidIO, HyperTransport, 3GIO, StarFabric, Packet Switched Ethernet, and others.
The interconnect presentations made it clear that it's too soon to tell which, if any, of the vying technologies will be the successor to PCI. With the PC market no longer the key technology driver for semiconductors, it's not necessarily a given that the next I/O interconnect to be used in PCs will migrate into the embedded board realm as PCI did. What's more likely is that the embedded board market will leverage technologies that gain acceptance in several markets, including servers, network equipment, and communications gear.