News 

The PCI-SIG consortium features Intel, IBM, Compaq, Microsoft and Hewlett-Packard, among others.

PCI (Peripheral Component Interconnect) has until now been quick enough to handle most of the tasks sent its way, whether they be network traffic, video from graphics cards or data from external storage devices. But increasingly the demands made by the likes of gigabit Ethernet and Ultra 3 SCSI have exceeded PCI's bandwidth. As an interim solution, IBM, Hewlett-Packard and Compaq developed PCI-X with twice the bandwidth, but this will eventually be superseded by the demands of newer and faster peripherals.

However 3GIO, codenamed "Arapahoe", is not expected to be in widespread use until at least 2009, so PCI-X will be with us for some time. Having debuted in Compaq servers, PCI-X is expected to appear in desktop computers by the end of 2001. Most users will be unaffected as it is back-compatible with existing PCI devices, so there will be no need to upgrade or replace. However, its greater

ADVERTISEMENT

fault tolerance - it can reinitialise a faulty card or take it off-line before a system crash occurs - should ease the strain on overworked PCs.

What isn't clear is how Intel's plans conflict with arch-rival AMD's HyperTransport technology, which a number of major players signed up to just last week. Ostensibly HyperTransport deals with communications between chips within computers, while Intel is focusing on I/O transport between computers and peripherals. But some industry figures are concerned that they will end up with two standards that might not be particularly compatible, if AMD decides to push HyperTransport up against 3GIO.

Scott King, a platform engineering manager for Compaq, has described 3GIO as "revolutionary, not evolutionary", requiring that computers be completely rebuilt. At that stage, Compaq, he said, would, "prefer one architecture, so that the whole industry wouldn't have to support two infrastructures."

However, Carl Stork, general manager of Microsoft's Windows Division, warned that, "I don't expect to see what Intel's doing and what AMD's doing come together."

Nonetheless, in the past apparently reconcilable differences have often been resolved. When two sets of people try to solve the same problem within the same environment, you can expect them to reach similar conclusions and it would be churlish for one or the other to claim exclusivity. But with any differences between AMD and Intel apparently irreconcilable at present, the decision by PCI-SIG may be a clear signal of which direction manufacturers prefer.