Vendor rivalry puts progress on hold

It was hard to avoid Gordon Moore at the Intel Developer Forum (IDF) earlier this month. Not that the great man was speaking at the San Francisco event but because the whole conference was a love-in for the Intel co-founder and the law that helped induct him into the IT virtual hall of fame.

Outgoing Intel chief executive Craig Barrett rejoiced that despite the naysayers, Moore's Law was still going strong. His juniors, almost without fail, followed suit.

In a 1965 paper submitted to Electronics magazine, Moore, then at Fairchild Semiconductor, stated that the number of transistors it was possible to pack onto an integrated circuit doubles every two years. Since then, the curve has largely followed that line. In the article, he suggested that "by 1975 economics may dictate squeezing as many as 65,000 components on a single silicon chip". A dual-core Itanium this year will host about 1.7 billion transistors.

There's plenty of fun to be had re-reading Moore today, especially when he says that, "Integrated circuits will lead to such wonders as home computers ... automatic controls for automobiles, and personal portable communications equipment."

Moore was prescient indeed to state that, "The future of integrated electronics is the future of electronics itself". But while there is much truth in that statement, it is also fair to say that Moore's Law is an increasingly imperfect benchmark of overall computing benefit.

Today, we have stopped expecting the death of silicon: nobody demurred when Intel suggested at IDF that silicon will go on until 2020 at least, or that next-generation chip fabrication technologies will see Moore's Law live after the CMOS process is dead.

But as business software runs happily under older processors and vast hardware power has become as cheap as, well, chips, the roadblocks ahead lie outside silicon and often are political.

At IDF, for example, Intel described an architecture called Active Management Technology (AMT) that provides computer and network components with interfaces so that security and other management routines can be applied on client devices even when no operating system is available. Also, Intel went deeper into describing Virtualisation Technology (VT), its hardware support for software virtualisation.

What have AMT and VT got in common? They are technology platforms that Intel does not want to share with rivals. So AMT won't work, for example, with proliferating AMD-based, or other non-Intel, network clients.

Sure enough, AMD is preparing Pacifica, its version of VT. Don't be surprised to also see an AMD flavour of AMT - and don't be surprised if there are hurdles for software developers and end-users to leap. Intel is not alone in this. It is an indictment of the IT industry that gains made in one part of the loop are often largely lost to another bottleneck because big vendors don't want to play ball with each other.

Intel is a great industry bellwether; it's just a shame that it does not distribute some of its broader technologies to rivals that are a fraction of its size. Another Intel-founder, Andy Grove, was famous for believing that "only the paranoid survive". His legacy, like that of Gordon Moore, clearly remains intact.