The personal computer is about to undergo a transformation the like of which has never been seen before.
During the next 18 months, vendors, led by semiconductor giant Intel, will introduce a series of architectural innovations to help clear performance bottlenecks, drive the development of even more powerful PCs, and herald the introduction of many different forms of device.
These changes will have an effect far beyond the realm of the desktop PC, believes Meta Group analyst Steve Kleynhans. “We believe that by 2006, the majority of what currently would be considered PC
components will be used in other computing devices – personal video recorders, home security systems, digital entertainment – as the PC world’s economies of scale invade the consumer electronics market,” he says.
The driving force behind the forthcoming revolution will be the introduction of a range of interconnect technologies. These will not only offer exponentially higher performance, but will also enable PC makers to build smaller PCs and to use the technologies for a much wider range of applications.
And the revolution is already starting. In the current quarter, PC vendors have started shipping machines that feature the universal serial bus (USB) 2.0, an upgrade to the standard connectivity mechanism used to carry instructions between the PC and external components such as the mouse or the keyboard. That will raise the speed of the USB interface from 12 megabits per second (Mbps) to 480Mbps, enabling a wider range of external devices to be seamlessly connected to the PC, such as camcorders.
At the same time, the means of connecting disk drives, DVD drives and other storage devices internally – fast advanced technology attachment (Fast ATA) or enhanced integrated drive electronics (EIDE) as it is also known – is being replaced by Serial ATA. Again, performance will be radically cranked up, in this case from 33Mbps to 1.5 gigabits per second (Gbps).
Additionally, there is PCI Express, the centrepiece of the coming revolution and the long overdue replacement for the ten-year-old PCI standard for attaching expansion cards and devices to the PCs motherboard.
Development of the current peripheral component interconnect (PCI) interconnect was started in 1992 and introduced with the first Pentium microprocessors in 1995. Back then, a top-of-the-range PC cost £2,000 and offered only the most basic sound and graphics.
As the PC developed during the 1990s, PCI’s typical bandwidth of just 132Mbps rapidly became a bottleneck. It was first ‘out-evolved’ by graphics cards, for example, which adopted the advanced graphics port (AGP) standard in order to accommodate the huge volumes of data required for graphic-intensive games.
But when PCI Express finally emerges in the first half of 2004, it will boast bandwidth that is 20 times faster at around 2.5Gbps, with subsequent generations offering first 5Gbps bandwidth and then 10Gbps thereafter. That will stretch the very limits of copper, says Howard Locker, lead development architect at IBM’s personal computing division.
PCI Express is intended to bring together AGP, PCI and core logic interconnects on the one interconnect technology. Component makers can raise or lower bandwidth by deciding how many PCI Express ‘lanes’ their products offer – 1, 2, 4, 8, 12, 16 or 32, depending on the amount of bandwidth required.
Furthermore, the cable required will be considerably slimmer – resembling a phone connection – which will mean the end of the thick ribbon cables that currently fill the insides of desktop PCs and will also help PC makers to experiment with smaller form factors.
Finally, holding everything together will be universal plug and play (UPnP), a standard connectivity mechanism that will automatically be able to ‘discover’ and connect disparate devices directly to the PC, on Ethernet, over the Internet or on wireless networks that support the 802.11x standard.
But there is also much potential overlap – and confusion – between these various technologies. For example, Meta’s Kleynhans suggests that USB 2.0 could be used for internal as well as external interconnection, supporting disk drives, for example, instead of Serial ATA.
“We expect vendors to begin to use USB not only as an external interconnect, but also as an internal interconnect for less performance-sensitive applications, supplementing the more expensive Serial ATA,” he says. But users are unlikely to be disturbed by such industry wrangling, which ought to be cleared up by the time PCI Express finally appears.
Combined with ever faster microprocessors, partly fuelled by innovations such as Hyperthreading, which enables instructions to be executed in parallel, PCs based on these technologies will be an enticing prospect for some business and engineering users, particularly those that perform data intensive tasks.
Yet the PC industry is not making a big noise about these developments – and with good reason. Fully specified next generation PCs will not be available until mid-2004 at the earliest, and PC vendors certainly do not want users to delay upgrading for that long.