Hardware and Software are in a State of Flux

by Alan Zisman (c) 1998. First published in Canadian Computer Wholesaler, March 1998

We?re in a transition period? in this space in past issues, we?ve looked at up and coming gizmos and gadgets designed to overcome the bottlenecks and design limitations that plague current hardware. A collection of 3-letter acronyms that taken together can easily leave the head spinning: USB, DVD, AGP and more (as well as the thankfully acronym-less FireWire).

And we?ve seen how many of these advances have stalled? the hardware?s ready (mostly), but until the next generation of operating system releases, there?s no built-in support. And with few products on the shelves, there?s little consumer demand. And with little consumer demand, few products make it to the shelves.

As well, the products that DO make it out haven?t been entirely overwhelming.

Despite promising blazing video speeds, for example, the first generation of Accelerated Graphics Port (AGP) video cards have been only modestly faster than the PCI cards they aim to replace. Generation 1 Digital Video/Versatile Disk players are unable to read the gold CD disks made by increasingly popular CD-Recordable units (some newer DVD models no longer have that limitation). Universal Serial Bus performance has been as slow as the product?s acceptance.

Even Intel?s next-generation Pentium-II CPUs fail to provide as noticeable performance boost as many expected.

To an extent, product manufacturers blame Microsoft. (A popular refrain these days? though I?ve yet to hear the weather being blamed on Microsoft, it seems like almost everything else is). The interim drivers for AGP, USB, et al are a patchwork sort-of-solution added onto Win95, and not optimized for perky performance. Hopefully, the next generation Windows 98, with future-hardware support built-in will do better.

But while these next generation operating systems provide some answers, they also raise a new set of questions.

Microsoft has promised that Windows 98 and NT 5.0 will share a common hardware device driver model. The Windows Device Model (WDM) will let hardware developers write a single driver usable by both the mass-market Win98 and the high-end NT 5.0 operating systems.

But there?s a catch. WDM drivers should work happily on machines running either of Microsoft?s next generation systems, but will be incompatible with the installed base?100 million + machines running Windows 95 (as well as the several million running NT 4.0). And many if not most of that installed base is not going to be in a rush to upgrade to Win98 (et al)? at least not until they purchase a new machine that comes with the operating system included.

You can develop old-style Win95 drivers? these will continue to work with Win98, but won?t work with NT 4.0 or 5.0.

If you?re producing hardware add-ins (sound cards, video cards, and the like), or are distributing such products, this is a dilemma. For your new products, should you develop WDM drivers, for the new Win98/NT 5.0 users? Should you develop backwards-compatible Win95 drivers? Should you play it safe and develop both (at double the cost?)

Larger companies with deeper pockets will probably play it safe. Creative Labs, for example, is promising to follow that strategy (though Windows 3.1 holdouts will increasingly be out of luck). Universal Serial Bus products from smaller companies may include WDM drivers only, on the reasonable assumption that few Win95 users actually have USB ports. Add-in cards made for the legacy ISA bus may continue to include classic Win95 drivers only, leaving NT users with complaints, just as now.

All in all, not a pretty picture.

Two issues ago, we looked at Microsoft/Intel?s PC98 plans? an outline of where the two giants want to move the industry. Not this year, but ultimately, they?d like to see us all abandon the 1984-era ISA bus? dropping support for those familiar 16-bit cards that, with their limited IRQ numbers and non-intuitive installation make it difficult to implement real PC Plug and Play, and drive up support costs.

Intel has released some specifications aiming to help product manufacturers during the transition. Their Low Pin Count interface (LPC?here?s yet another 3-letter acronym!) is designed to make it easy to redesign motherboards to move traditional components like parallel, serial, and keyboard ports, along with hard disk and floppy disk controllers off the ISA bus. (Even though most motherboards now have these devices built-in, they?re currently actually part of the ISA bus).

National Semiconductor has used the LPC spec to release its PC87360 Super I/O chip, integrating all those devices, resulting in lower cost while taking up less space. Future LPC products will include memory and system management controllers, along with motherboard-based sound circuitry.

Intel envisions a three-phase transition along the way to replacing the ISA bus. First, current motherboard devices will migrate from ISA to LPC, while systems retain the ISA bus for compatibility with legacy peripheral devices (most often sound cards and modems). In the second phase, as USB and FireWire devices become more common, systems will drop the ISA bus entirely, with users adding PCI cards internally, or using USB and FireWire for external devices. Finally, as a wide-range of external peripherals such as printers become available for the new external ports, support for legacy parallel and serial ports will be dropped entirely.

LPC is, according to Intel, a key step along the way. It allows system designers to upgrade existing designs with relatively minor modifications, while upgrading performance? the ISA bus runs at a leisurely 8 MHz, while LPC uses the PCI bus?s 33 MHz clock. Its reduced pin count saves on space and power, and runs cooler, making it especially attractive to notebook designers. Quoted in InfoWorld, Intel?s Platform Component division marketing manager, Jan Camps claims LPC ?will result in lower costs and improved efficiency for hardware OEMs and developers, and make the benefits of higher-performing technologies available more quickly for PC users. Intel has opened the specification so the industry can quickly adopt the technology and integrate it into motherboard legacy I/O peripheral development.?

Technology pushes change. Despite our sense of constant product obsolescence, however, the very success of the personal computer has created a huge installed base that is demanding gradual evolution. These contradictions require careful planning from designers, manufacturers, wholesalers and distributors down to retailers and ultimately end users. Know where the industry is heading, and you?ll be able to manage the coming transitions.