Embedded Processor Watch

Editor: Cary D. Snyder

Contributors to this issue: Michael Slater, Peter N. Glaskowsky

and Markus Levy

In This Issue:

• The Future of Computing Platforms
• ARM Embraces SIMD Support
• **** EDITORS NOTE
• Athlon Edges Out Pentium 4 (Summary Version)

The Future of Computing Platforms

By Michael Slater {1/9/01-02}

In the past few years, a tremendous amount of attention has been focused on the impending shift to the "post-PC" era, in which the PC is replaced by myriad information appliances. This vision has many merits, but its realization is further in the future than its proponents have hoped--and it understates the role that general- purpose platforms will play.

It is easy to hate the PC: many things are wrong with it. Architecturally, PC hardware design has reached an entirely acceptable level. Much of the legacy hardware (such as the ISA bus) has been purged from the current generation of systems, and 1394 and USB ports eliminate most of the drawbacks of the ancient serial and parallel ports. The PC's biggest hardware drawback today is one that the industry should be able to fix--if it can break the inertia: the predominance of ugly, bulky, noisy systems.

When we get to software, there's a lot more to hate about the PC. Twenty years of incremental evolution have left us with a system that is far more complex and unreliable than it should be. Windows 2000 improves reliability significantly, and in time this technology will trickle down to the consumer market, but the complexity remains. And in PC application software, featuritis has run rampant, while quality design that has the real needs of users in mind remains rare.

And then there's the business element: Microsoft and Intel earn a great deal of the profit made in the PC business, and their dominance can't help but generate a lot of resentment.

As the Web rose to prominence and computing power became cheap enough to distribute into less-expensive devices, the multitude of reasons to hate the PC fueled enthusiasm for a new wave of application-specific appliance-like devices. As I discussed in my September column, the idea of Web appliances, whether counter-top boxes or mobile tablets, is very appealing, but these devices have largely failed in the market so far. The reasons range from the high cost of displays and the challenges of supporting plug-ins to the difficulty of finding early adopters for products essentially aimed at laggard markets (e.g., the Web appliance for the masses).

This model has also run into trouble with the idea of a Web browser as a do-all environment. Thirty years ago, computing was done mostly on large shared computers: mainframes. Then PCs came on the scene, and by the time they had been around 10 years or so, nearly all applications had been moved off the mainframe and onto the PC. A few years later, with the spectacular rise of the Web, the pendulum--at least in terms of the design point for future products--had swung too far in the other direction. In the future, many thought, applications would run on the Web, data would be stored on the Web, and all we'd need on the PC would be a browser running Java. No more administration of local systems, no more overly complex PC software, no more local state.

This was a nice theory, and its appeal was strong in light of all the good reasons to hate the PC and the PC business. But it won't succeed as a mainstream replacement for consumer computing for one simple reason: it unnecessarily compromises the user experience.

To fit into the "everything is on the Web" model, users must give up too much. The PC model has lots of drawbacks, but it also has a lot going for it, not the least of which is an installed base that now covers essentially every business desk and around 60% of homes in the U.S. In theory, starting over with Java-based programs would free us all from the PC standard and stimulate a new wave of innovation. In reality, however, Java programs tend to be annoyingly slow, and the sheer effort of recreating the essential base of applications would be massive. For a software developer, it is hard to argue against building your program to make the most of the hardware--the vast majority of potential customers already have and that, for better or worse, is the Windows PC. (The full version of this article is available online to Microprocessor Report subscribers at http://www.mpronline.com/mpr/h/2001/0102/150102.html )

ARM Embraces SIMD Support

By Markus Levy {01/02/01-03}

MCore, move over; ARM is here. Motorola has always considered ARM cores as competition to its own MCore cores, so it was interesting, but not surprising, news that Motorola has licensed the ARM processor families. The company plans to rely on its ARM-based products to supply customers that are creating the next generation of PDAs, smart phones, and Internet appliances, particularly devices that are based on the Palm OS. This licensing deal doesn't imply that MCore will be disappearing--at least for the time being; Motorola has a healthy roadmap and many design wins for MCore-based cores and devices. The advantages of the homegrown core are many, including not having to pay royalty fees.

In response to customer requests, Motorola plans to use its ARM license to expand its DragonBall family, the extremely successful processor line that powers Palm's handheld computers. The 68VZ328 and 68EZ328, both 68K-based processors, currently comprise the DragonBall family. Although these processors continue to serve their purpose, they top out at 33MHz. The ARM processors will certainly provide increased performance capabilities.

Motorola hasn't released details of any core or silicon implementations, but it is clear that its license will allow it to make architectural enhancements. Until now, Intel has been the only other licensee allowed to make such enhancements. Motorola will be making ARM-based product announcements in 1Q01 and be in production by 4Q01. MDR expects that one of the first products will marry an ARM core to a StarCore or other Motorola DSP for wireless communications functions. This product may be similar to Intel's combination of its XScale processor with its new Frio DSP and Texas Instruments' union of its DSPs with ARM10 cores. (The full version of this article is available online to Microprocessor Report subscribers at http://www.mpronline.com/mpr/h/2001/0102/150103.html )

**** EDITORS NOTE ****

At times readers may view MPR as biased toward one vendor or another - as analysts at MPR we form our own opinions supported by technical data often provided by vendors and associates. I received large amounts of data and support from Intel on the first XScale Chip implementation and was favorably impressed. Granted, being the relatively new editor of the Embedded Watch my enthusiasm could be a bit zealous at times, which some might misread as "marketing hype." However, I am fully committed to providing strong technical content. I certainly welcome additional feedback on these types of issues. The following summary is being included in the EPW to provide some direction to what I predict will find its way into future Embedded Applications.

Athlon Edges Out Pentium 4

AMD Slightly Faster, But Intel Chip Shows Great Potential

By Peter N. Glaskowsky {1/08/01-01}

Our first hands-on tests of Intel's 1.5GHz Pentium 4 and AMD's 1.2GHz Athlon have borne out the predictions we made four months ago based on Intel's preview of the P4's microarchitecture. The P4's higher clock speed gives it impressive peak throughput on selected tests, but overall, on a variety of tests, it comes in slightly behind its primary competitor. (The full version of this article is available online to Microprocessor Report subscribers at http://www.mpronline.com/mpr/h/2001/0108/150201.html )