| The old vision that disparate computer and medium
technologies would converge on a single platform is fading
fast. Features are converging, but existing platforms
can't stand the strain. They're splintering into new
platforms and subplatforms to satisfy an exploding
computer market that's extending its reach from TV
Internet terminals and $300 Windows PCs to 64-bit
workstations and enterprise servers. One-size-fits-all PCs
can't cut it anymore. The dominant standard that we've taken for granted for a dozen years, the x86-based PC clone, is showing clear signs of these fractures. It's becoming a victim of corporate rivalries and a strange technological paradox — hobbled by too much old technology to keep up with new technology, and swamped by too much new technology to justify large investments in current technology. In its place, a looser, more stratified PC "standard" will maintain a semblance of software compatibility while addressing a much broader range of users. And users who are overwhelmed by all these changes are beginning to revolt by spending less on computers that they know will be instantly obsolete. Some market watchers think this trend is the inevitable culmination of the PC's success. "What you're calling platform fragmentation is what I call market segmentation," says Steve Tobak, vice president of corporate marketing at Cyrix. "Did anybody really think that when PC volumes climbed to 100 million that there would be only one kind of PC?" Four Layers Here's how the new world order will look: • IA-64 will be the next-generation CPU architecture for high-end Windows workstations and servers. And the first implementation will be Intel's 64-bit Merced processor, scheduled to debut in late 1999. But IA-64 won't replace the 32-bit x86 architecture for a long while. (See "Beyond Pentium II" in the December 1997 BYTE.) • The x86-based PCs will continue to thrive well into the first decade of the twenty-first century as the mainstream architecture for desktop PCs, notebook computers, and department-level servers. Intel and other x86 vendors will continue to introduce new designs for years to come, even after the debut of IA-64. • Super-integrated CPUs built around x86-compatible cores will address the low end of the market for Windows PCs in homes and businesses. The leading example today is Cyrix's MediaGX, which integrates the functions of a CPU and several motherboard devices on a pair of low-cost chips (see the sidebar "System On a Chip"). The MediaGX powers a new breed of affordable PCs priced from $499 to $999. Later this year, prices will drop to an unbelievable $299. • TV Internet terminals will anchor the bottom end of the market for consumers who don't need or can't fathom a full-blown Windows PC. This segment got a big boost when Microsoft reversed its opposition to network computers (NCs) and bought WebTV Networks. Although it will take a while to establish these novel devices in the technology-resistant consumer market, they will eventually be a popular alternative to overcomplicated PCs. The new stratification doesn't end there. It will span multiple technical standards as well as market segments. Whether you call it fragmentation or segmentation, it's not like yesterday's comfortable PC environment, in which all PC clones were basically the same except for their degrees of megahertz and megabytes. The fault lines separating the new layers will run as deep as the OSes, the motherboards, and the microprocessor chips. Splintering Platforms By the end of this year, for example, there will be seven not-fully-compatible versions of Windows in wide use, each with an installed base in the millions: Windows CE, Windows 3.1, Windows 95, Windows 98, Windows NT 3.51, Windows NT 4.0, and Windows NT 5.0. Each upgrade maroons another island of users who don't have the hardware or the desire to keep up. Meanwhile, Microsoft struggles to unify the driver models, user interfaces, feature sets, and APIs for all those variations of Windows. Microsoft's well-publicized plan is to eventually converge the PC platform on Windows NT, with Windows CE as the scaled-down solution for palmtop computers, WebTV boxes, and other alternative devices. But to paraphrase an old Army joke, that would leave Microsoft with only two sizes: too large and too small. The increasing segmentation of PC hardware will probably force Microsoft to abandon its grand unification theory and continue to offer one or possibly more medium-size choices. Unless Microsoft manages that process uncommonly well, it will fracture the PC platform still further. Someday, you might need Java just to write a program that runs on every version of Windows. Meanwhile, outside the world of Windows, there are still tens of millions of users who think the idea of a one-size-fits-all OS is as quaint as a one-size-fits-all PC. Contrary to shortsighted beliefs, there will always be alternative platforms such as Java, Linux, Unix, the Mac OS, OS/2, BeOS, and Rhapsody. If Microsoft stubbornly pushes its plan for convergence on Windows NT, one of these upstarts will fill the gap. Drilling deeper, down to the motherboard and microprocessor levels, even more cracks are appearing in the PC platform. The leading vendors of x86 processors are introducing mutually incompatible CPU interfaces to replace today's standard Socket 7 (see the sidebar "The Slot Thickens"). To accelerate multimedia performance, they are also promoting mutually incompatible extensions to the standard x86 architecture. (See "Beyond MMX" in the December 1997 BYTE.) Both of these fundamental changes will lead to more diversity in processors, motherboards, and chip sets — and more hazards to compatibility. As if all that weren't enough, there's also an alphabet soup of new technologies intended to completely overhaul today's PCs: 100-MHz system buses, the Accelerated Graphics Port (AGP), AGP-2x, AGP-4x, synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR-SDRAM), Rambus DRAM (RDRAM), DVD-ROM (digital versatile disc), DVD-RAM, the Device Bay interface, MMX2, the slow eclipse of ISA, faster/wider PCI, ISO-1394 FireWire, universal serial bus (USB), Fast Ethernet, the Advanced Configuration and Power Interface (ACPI), Ultra DMA, Ultra 2 SCSI...and more...and more. Users Revolt There's growing evidence that computer buyers, both consumer and corporate, are reacting to the uncertainty over these changes. Although they're buying more computers than ever before, they're spending less money on each computer — often, much less. According to market researchers at Computer Intelligence (La Jolla, CA), about 70 percent of the PCs sold through U.S. retail stores in recent months cost less than $1500. Even more remarkable, 30 percent to 40 percent cost less than $1000. A year ago, only about 5 percent of PCs cost less than $1000. What was once an insignificant crumb has suddenly become the whole cake. In the dealer channels that serve mainly corporate buyers, average prices are a little higher — but still, a third of these PCs sold in recent months cost less than $1500. In December, Hewlett-Packard introduced a Vectra-series PC for the corporate market at a base price of $999. Of course, everyone likes a bargain. But in the past, users resisted low-priced PCs because they feared the machines would become obsolete too quickly. Now they're resisting high-priced PCs because they fear those machines will become obsolete too quickly. "People have fought this fear of obsolescence for a long time, and they've given up," says Matt Sargent, an analyst at Computer Intelligence. "No matter what kind of PC they buy, it's going to be obsolete in two years anyway. So they'd rather spend less on the PC and replace it more often." This signals a major upheaval in PC purchasing habits. For years, the standard computer-magazine advice has been to buy the most powerful PC you can afford as protection against early obsolescence. That advice is now obsolete. So much new stuff is coming that early obsolescence -- indeed, instant obsolescence — is inevitable. No amount of money you spend today will protect you against it. Obsolescence is a relative term, of course. If you don't install any software that was written more than a year after the machine was made, it will last virtually forever. However, few users are so disciplined. Each new software "upgrade" gradually turns their big, fast computer into a small, slow computer. Eventually, the performance becomes so intolerable that the only solution seems to be the purchase of another big, fast computer. Another vicious cycle starts. The new philosophy calmly accepts the inevitability of this cycle — and, in fact, speeds it up. However, because each replacement machine costs less, users spend less money per year to stay current. Prices are plunging so fast that one computer-store chain is selling a custom-designed Windows PC for only $499, and it's not as stripped-down as you might think. By the end of the year, similar systems will retail for $299. (See the sidebar "How Low Can You Go?".) Welcome to the disposable computer. Even if you get only one year out of the machine before handing it down to somebody or donating it to a school, the amortized purchase cost is only half as much as a $2000 system that lasts two years or a $3000 system that lasts three years. And for many common tasks, a $499 PC is fast enough. Indeed, cheap PCs have the same attributes that made the Volkswagen Beetle popular in the 1960s: basic transportation, no complicated frills, and freedom from fashion anxiety. You never have to worry about it becoming obsolete because it's already obsolete. Buying Smart However, it's equally obvious that a VolksPC isn't the ideal machine for everyone. Some users need more power to do their jobs efficiently, and anything less than the fastest available system isn't a worthwhile trade-off. Another huge consideration — especially for businesses — is the less-visible cost of setting up a new PC. Low prices save money up front but do little to address the total cost of ownership. An often-quoted study by the Gartner Group estimated the total business cost of owning a PC with Windows 95 at $38,900 over five years. That includes administrative expenses for technical support, requisitions, purchasing, legal reviews of contracts, security, policy enforcement, and asset management. Gartner assumed a new PC every three years. If a company replaces its PCs every year, the administrative overhead could consume the purchase savings. That's why cheap PCs don't necessarily threaten NCs, which reduce the total cost of ownership on the administrative end. So here's the amended computer-magazine advice: Minimize your investment in transitory technology and administrative costs while carefully matching the computer to the user. That's hardly radical advice, but it does open more options. In some cases, the best solution will be a 333-MHz scream machine loaded with all the latest trimmings. In other cases, it will be a $500 disposable PC or an NC. You might recommend a $300 WebTV to a technophobic friend, even if you wouldn't buy one yourself. The goal is to keep up with traffic, not with the Joneses. By applying this template to the stream of new technologies, you can filter out the ones that aren't relevant. For instance, users who aren't multimedia developers, graphics designers, or game players probably don't need things such as MMX, MMX2, or AGP. Writers and accountants probably don't need a 100-MHz system bus, a 300-MHz CPU, a FireWire interface, or the latest DRAM technology — at least, not within the life expectancy of the next system they purchase. Workers who spend all their time interacting with a centralized database might get by with an NC. Not only can buyers save money, but they can also acquire new technology sooner than they would have otherwise. This happens when the technology churns so fast that users get new features essentially for free merely by purchasing a new system, whether they sought those features or not. For example, Intel and other x86 chip vendors have stopped making non-MMX processors. All new PCs this year will have MMX at the same prices (or even lower prices) than last year's non-MMX PCs. Someone who bought an expensive pre-MMX system in 1996 might still be amortizing the high cost over a longer period, while the bottom-feeders are snapping up sub-$1000 boxes with MMX. Technology churn will accelerate the trend toward the cheaper PCs. Why should users lock themselves into costly systems that will not have next year's features? Paradoxically, this motivates the industry to churn the technology even faster. Vendors make less profit on low-end systems; new whizbang features justify higher prices. But when the technology churns so fast that users can't amortize a high-end system before it becomes obsolete, they tend to buy less expensive systems. It's a feedback loop! So if you don't like the idea of planned obsolescence, don't get mad — get even. Buy a new PC every year. Tomorrow's Tributaries "There's obviously a battle shaping up over what the mainstream PC is going to be," says Don Clegg, vice president of marketing at Tyan Computer, a major motherboard manufacturer. Time was, you could easily recognize a mainstream PC on sight: It was a big, ugly beige thing that took 3 minutes to turn on and sounded like a vacuum cleaner. Soon, that concept will be obsolete. Already, some PC vendors are trying to change that image to further the illusion that PCs are user-friendly consumer appliances. You can buy PCs with charcoal-gray cases, built-in stereo speakers, and "instant-on" power buttons (which, to be more accurate, are really "never-off"). More significantly, leading vendors such as Compaq and Gateway 2000 are experimenting with wildly different types of PCs that are hybrid TV-PCs, DVD-PCs, or home-theater PCs. "Culturally, the industry is very limited in what it thinks is a PC," says Paul Pascarelli, Cyrix marketing manager for the MediaGX chip. "MediaGX is an architecture for devices that need x86 intelligence but don't necessarily have to resemble ordinary PCs or run the same kind of software as traditional PCs." It's not just an issue for the home market. In the business world, the need for more alternatives has spurred the evolution of NCs, NetPCs, Windows CE palmtops, the PalmPilot, and a wider variety of notebook computers. In nerd-speak, segmentation lets users match the computing device to the task with a greater degree of granularity. Don't be fooled into thinking that everything is converging on a single platform: x86 and Windows. Superficially, it looks that way, especially with the decline of alternative platforms such as the Mac. However, no platform can be all things to all people. The apparent winner of the platform war is unmistakably splintering into subplatforms. In terms of system architecture, for instance, a MediaGX PC requires a specially designed motherboard that won't work with any other CPU or standard core-logic chips. It doesn't need a sound card, a graphics card, or an L2 cache. It runs Windows and works with standard peripherals, but internally it's an architecture unto itself. By next year, it won't even have an Intel-standard x86 instruction set. Then there are completely different architectures, like Microsoft's WebTV. For the first time, Microsoft has direct control (as opposed to de facto control) over a system architecture. What will Microsoft do with it? Future WebTVs could always remain WebTVs, or they could grow upward to become more like regular PCs — an evolution that's already apparent in the latest WebTV Plus models, which come with hard drives and printer ports. The same opportunity is open for video game consoles such as the Sony PlayStation and Nintendo 64. These devices are more powerful than many PCs and are made by companies that have vast experience in consumer marketing. In the past, rogue architectures found it difficult to survive, even when they offered breakthrough features. (Remember the Amiga?) However, as the computer market continues to broaden and grow into the hundreds of millions of units, there will be further room for alternatives — as long as they don't challenge the status quo too radically. Sidebars:
Where to Find
Advanced Micro Devices
The New World Order Tom R. Halfhill is a BYTE senior editor in San Mateo, California. You can contact him at tom.halfhill@byte.com. Copyright 1994-1998 BYTE |