Chips with everything

Intel is getting increasingly aggressive with the Pentium 4, both with its pricing of the main processor, and the release of the new Celeron.

This processor is based on Pentium 4 technology and offers extremely good price/performance. It is between 10 and 15 per cent slower than the Pentium 4 at the same clock speed, but much cheaper. You can't help but wonder whether rival AMD will be able to withstand this onslaught.

The differences in strategies between the two companies have mattered less than perhaps I would have expected a few years ago. Sometime around 1996, AMD was telling me of its future plans, and how it was optimising its designs for office-type applications in our Winstone performance benchtest.

I urged them not to do this, making the point that performance mattered little for such applications. Of course, I couldn't tell them what I knew of Intel's plans at the time, which were to try and work out what future applications would be like, and design for them. That seemed to me to be the obvious direction to take.

Naturally, when Intel came out with the Pentium 4, AMD made as much as it could of the fact that the Pentium 4 showed no improvement for Winstone-like office applications, and Intel's marketing at the time left much to be desired. Partly as a result of this, AMD has had a good run in the past couple of years.

Mind you, both companies' marketing suffers from the same problem: the relative speed of modern processors depends on what applications you run on them, and these differences are only going to get greater.

If you could compare modern systems with, say an old 486, office type applications would obviously run a lot faster. However, the speed gain for computer intensive workstation type applications such as CAD would be even greater.

So AMD has optimised its processors more for the office-type application, but has also taken on board the new instruction sets. Intel has decided to optimise its Pentium 4 for its view of modern and future applications, and not worry about office suites. This spread of performance gain applies to both, but is somewhat greater for the Pentium 4 than for AMD's processors.

As a result, each company tries to distinguish its processors in different ways by choosing the benchmarks that show them in the best light. AMD, for example, makes much of not judging processor performance by megahertz alone, but doesn't make a clear case for its strange rating system.

There is one area, however, where there's a clear distinction between the two companies' plans. This is for future 64-bit computing. Intel has already shipped its 64-bit processor, the Itanium, with a completely new architecture and instruction set. A radical break from the past.

AMD will later this year bring out its Hammer processor based on the x86 architecture, but with 64-bit extensions. An evolutionary change.

Intel maintains that the x86 architecture is running out of steam and needs to be replaced, and AMD is insisting that compatibility with the huge legacy of x86 development of the past 20 years outweighs any advantage from a move to a completely new architecture.

These competing processors will find their first roles in servers and workstations. AMD has developed a decent market share in the workstation market and I know that workstation manufacturers are already taking a strong interest in Hammer.

The Pentium 4, expected to be able to go up to 10GHz in about three years time, has certainly shown that there's a lot of life yet in the x86 architecture.

In this respect, I think AMD has made the right decision with Hammer. So where does that leave Intel, and the Itanium? Unless everybody in Santa Clara has taken complete leave of their senses, there must be a parallel team working on a 64-bit x86 processor inside Intel.

Maybe it doesn't expect to ship this processor and is developing it 'just in case'. But I think they will have to offer it.

Ironically, software that makes use of the 64-bit extensions in Hammer should run on an Intel 64-bit x86 processor immediately so, if it appears that Hammer is doing better than the Itanium, and this looks quite possible, it will not be very difficult for Intel to ship its 64-bit device.

What will it be called? Something like a Xeon-64, I expect, and it will probably surface next year.

Like this story? Spread the news by clicking below:

 del.icio.us     Digg this     reddit!