Intel and AMD - countdown to the desktop shootout

WHILE WE BELIEVE Intel will have a very clear advantage over AMD for CPUs in notebook machines in 2003, and while the smaller chip company has a way to go in the server market, the world+dog will be watching to see how both firms will slog it out in the desktop market.

AMD has, on the face of it, a fighting chance to dig into Intel's Pentium 4 market share although there's a lot of ifs and buts.

One is AMD's precarious financial position. Hector Ruiz, the new CEO, is not only fighting what the old Jerry Sanders III used to call the "enemy without", but also is rejigging the firm within because of various collywobbles which we'd love to know more about but suspect are being hidden under a darker veil than usual.

AMD has to produce its Athlon64 (Clawhammer) and Opteron (Sledgehammer) chips on time, in sufficient quantity, at the right price, and in the right channels.

These are all formidable challenges - if AMD stumbles on manufacture, on supply, or on pricing, Intel will not hesitate to swoop.

No one should underestimate Intel's resolve. It has billions of dollars in the bank, it's de facto head, Paul Otellini, is an aggressive and competent long timer, and it's founded on a financially stable model that AMD simply can't match.

The Pentium 4 - what of it
Anyone remember this Russian rendition of AMD's AMD's roadmap, the original of which is at the AMD site and based on a Sanders' presentation, here (pages 14 and 18).

This should be compared with the Opteron roadmaps we've recently seen, and which use GHz rather than PR+Hertz speeds -- here.

Then have a look at where AMD hopes to be by this time next year with the Opteron processors, according to recent roadmaps we've seen.

Sledgehammer up to eight way DDR 333, 1MB L2 3 HT links .13µ SOI	Athens up to eight way DDR II 400, 1MB L2 3 HT links .09µ SOI
Sledgehammer MP up to two way DDR 333, 1MB L2 3 HT links .13µ SOI	Troy up to two way DDR II 400, 1MB L2 3 HT links .09µ SOI
Sledgehammer MP uniprocessor DDR 333, 1MB L2 3 HT links .13µ SOI	Venus uniprocessor DDR II 400, 1MB L2 3 HT links .09µ SOI
Barton one to two way 266FSB 512K L2 ,13µ
First Half 2003	First Half 2004

On the face of it, and provided that AMD gets all of its ducks in a row, its 64-bit chips will compete successfully with anything Intel can throw into the Pentium 4, such as 800MHz system buses, DDR-400 memories, hyperthreaded P4s at every level and cheap, cheap Celerons at high speeds.

Early benchmarks, such as those published by c't, indicate that the Hammer family will perform well. It appears it has several OEMs lined up to dabble in these waters, and it also seems to have adequate 64-bit software support from Oracle, from the Linux camp and from Microsoft.

The Battle for Process Technology
Most observers agree that when Intel is performing at its best, there is no match for its ability at the factory level to produce solid process technology at the right price and the right quantity. It stumbles too - the introduction of Pentium III technology was effectively delayed for months because it couldn't manufacture enough chips in volume.

Intel appears confident that it can move its processors from the current .13µ (micron) technology to the next generation 90 nanometer technology with the introduction of its "Prescott" microprocessor in the third to fourth quarters.

It has clear advantages over AMD because of the sheer size of the organisation and its team of engineers worldwide. AMD is promising its technology will also migrate to 90 nanometers, but that won't happen until "the first half" of 2004.

If Intel gets its move to 90 nanometers right, that means, in lay terms, that it will be able to churn out more chips at higher frequencies less expensively, and apply the pricing screws on AMD.

Perception, Perception, Perception
We coined the term "marchitecture" at another place. This, in brief, is the extraordinary ability of semiconductor firms to fly in the face of reality by, for example, promoting a CPU and its clock frequency as the "brain" of a computer, and plunging hundreds of millions of dollars into advertising, marketing and "below the line" campaigns to get its message across.

Other good examples of marchitecture are claims that a microprocessor somehow miraculously "improves your Internet experience" or will turn you from a dabbler in digicam technology into a 21st century Eisenstein.

There's no doubt that there's a lot of pent up demand for AMD's Hammer technology. Many of our readers have told us that they've postponed buying PC technology because they're waiting for something new to come along, rather than just an Athlon XP or a Pentium 4 running incrementally higher.

This favours Hammer, which has been long in the making. Back in 1992, Jerry Sanders III reckoned that the K9 architecture would be out by the end of that decade, and that means that the K8 is now several years delayed. Intel's big challenge in 2003 will be to position its Pentium 4 as sufficiently exciting to match the 64-bit architecture AMD promises.

And it may just have to gulp and swallow when benchmarketeers start comparing its Itanium platform to AMD's 64-bit server version of its Hammer family. µ