Intel’s embedded roadmap is increasingly being accepted in vertical markets where better performance, lower power, and reduced heat are becoming more critical considerations. That roadmap also is encountering competition in a variety of markets in which Intel has never been a force—as well as in places where companies that previously weren’t considered competitors are looking to expand. Most industry observers expect fireworks over the next several years as all of these companies pursue a collision course.
Over the last few years, Intel has been pushing heavily into the embedded market as part of its long-term growth strategy. Unlike some of its previous attempts to expand its market beyond PCs, which included everything from videoconferencing to communications servers, the embedded market is viewed as a logical extension of its core market. That opens the door for Intel to get involved in everything from consumer devices with its new Intel® Atom™ processor to rugged industrial and military applications.
One of the better-known markets in which the processor giant has seen early success is the Advanced Telecom Computing Architecture (ATCA), which is the carrier-grade equipment market, as well as its scaleddown offshoot, MicroTCA. Both have been hot markets for Intel’s newest chips—in part because the markets themselves are relatively new and because Intel has greatly boosted its commitment to developing technology in this space. That was a big first win for Intel’s chip lineup. But the newest chips are finding their way into a host of other markets as well. Those new opportunities include a variety of industrial applications ranging from medical imaging to the industrial control of everything from trains to tanks.
The newest dual-core, 2.53-GHz Intel® Core™2 duo processor, which is built at the 45-nm process node, is opening even more doors. “As compared to the previous generation, the newest chips are much better,” says David Pursley, field applications engineer at Kontron. “A 65-nm Intel Core 2 Duo processor at 1.6 GHz versus a 45-nm chip at 2.53 GHz is a big difference. Because of the process shrink, we’ve got a lower overall power envelope and better thermals.”
By way of a real-world example, consider the migration of medical-imaging technology. It started out a decade ago in a selfcontained selfcontained room in some hospitals. This technology took up so much space that many hospitals couldn’t add it even if they could afford it. According to Pursley, medical-imaging technology has since shrunk to the point where it can be used in medical-imaging outpatient offices. In addition, the technology is about to go mobile in hospitals. It will be wheeled around on carts so that the imaging can be done at the bedside instead of in a separate room.
In that type of application, heat and a rugged shell are very important. But performance is even more critical. A mobile imaging device is like a single-board supercomputer, which has to run an enormous number of parallel calculations to arrive at a cohesive image.
Intel’s roadmap runs as broad as it does deep. The new Intel Atom processor, for example, is based on the x86 Intel® architecture. The chipmaker is positioning it—among other things—as a replacement for microcontrollers. After all, it’s easier to develop software for the x86 architecture. But the other three pieces of its embedded lineup—Network processors, PXA2XX processors, and the I/O processors—cut a wide swath across the embedded space. They also utilize standard ARM development tools. As a result, developers in this space won’t have to retool their skills to work with Intel’s offerings.
Obviously, Intel will have serious competition in this market. The reason that the company has included ARM-developmenttool compatibility is significant. ARM is the incumbent in the embedded-processor market. In addition, ARM’s customer base is enormous and varied. But Intel’s push into markets dominated by ARM—and, to a lesser extent, MIPS—signals just how important the embedded market has become.
ARM is pushing back as well. The company has been running a campaign showing that in real-world settings, its processor cores are performing better than Intel’s. At the Common Platform conference in Santa Clara, Calif., ARM president Tudor Brown compared his company’s Cortex-A8 and the forthcoming A-9 versus the Intel Atom processorin terms of power consumption and area. He argued that ARM’s chip consumes 25% of the power in 25% of the ARM is pushing back as well. The company has been running a campaign showing that in real-world settings, its processor cores are performing better than Intel’s. At the Common Platform conference in Santa Clara, Calif., ARM president Tudor Brown compared his company’s Cortex-A8 and the forthcoming A-9 versus the Intel Atom processorin terms of power consumption and area. He argued that ARM’s chip consumes 25% of the power in 25% of the
In this war of words, being right may depend upon the individual application, the applications that are being used, and familiarity with the development platform. A single-core processor, for example, can perform better than a lower-clock-speed multicore processor if the application cannot take advantage of all of the cores. If a lower-power chip is needed, however, performance may sometimes be less essential than some of the energy-saving techniques employed in the chip.
“What’s happening here is the PC architecture is moving into the mobile market and ARM is moving into what I would call the larger mobile market,” states Dean Freeman, an analyst at Gartner. “There will be a collision as they hit common ground. ARM has been doing mobile for some time and they’re expanding. Intel is coming from the PC space and now they’re moving in.”
The collision of the two companies and their architectures mirrors the convergence that’s occurring in the consumer and business markets. There, the PC is the one tool that’s connected to more portable devices when users need more mobility. An example is when a user is commuting to work or traveling, but doesn’t want to lug along his or her notebook computer.
Intel also faces an uphill battle if it plans to displace some of the microcontrollers. While the ability to program an x86 processor is simpler than programming a microcontroller, not all applications require that kind of upgradeability. Justin Rattner, Intel’s CTO, explains that automotive companies in particular want to see that kind of programmability to be able to add new features as they become available. But automobile manufacturers often have multiyear design cycles, which make that kind of upgradeability more attractive there than it is in other markets. In the case of the Intel Atom processor replacing microcontrollers in things like motor control, the jury may be out for several years.
