2000: the chip odyssey

This year really saw the X86 war heat up, as AMD kicked in with its Athlon processor and gave Intel a run for its money, at least in the desktop space.

As usual, the year started with a vengeance, with AMD determined to take the price war right to Intel's ramparts, and slashed and burnt the Athlon prices in January. It seemed AMD caught Intel with its knickers down right up until May, when the leviathan finally realised it had to respond. All during those months, both firms, rather like Tweedledee and Tweedledum, attempted to boast that anything one could do, the other could do better -- certainly in terms of megahurts.

The trouble is, this megahurts stuff was all marketing flim-flam, and behind the scenes, Intel was struggling to make enough CuMine .18 micron processors to satisfy its customers - a hangover from their launch the previous October. One of the bigger European distributors complained in February that Intel was costing it big bucks as it was unable to satisfy its dealers.

AMD got there first to the 1GHz post but Intel had its CuMine 1GHz processor at the ready.

January also saw Transmeta hype itself in good old US of A fashion -- there wasn't much real news there -- we'd covered the essentials in 1998. But the world's financial press saw this as a good excuse to beat up on the Aunt Sally of the industry, the Intel Corporation. It would return to this theme later in the year.

At the Intel Developer Forum in Palm Springs in February, we saw Intel preview its Willamette platform, and we also saw people who should really have been in their offices in Wall Street sitting in the press office on the blower, and a-buying and a-selling Rambus stock like there was no tomorrow, even while the Intel suits were still talking.

The RMBS share price veered like only a crazy veering thing can veer throughout the year, at one point reaching the dizzying heights of $500 a share. The jury is still out on the Shorty Long stock of the year.

Rambus was in the news all the way through 2000 - mostly because it was either legalling other memory firms or because it was itself being legalled. It will be in the news next year too, no doubt about that at all.

The Computex show in Taipei saw the local manufacturers jumping onto the Information Appliance (IA) wagon. We were privileged to view the FIC wooden Transmeta appliance. FIC deserve our thanks for reviving the good old IBM tradition of making mock ups that may never make it to market.

It was also obvious that Intel had messed with the minds of the mainboard manufacturers, who were still trying to cope with the ramifications of the 820 Caminogate debacle, never mind the swift change that they'd have to make between a 423 and a 478 pin version of Willamette.

Other manufacturers pledged their support for Transmeta too but the economics of the processor for an IA simply didn't make much sense. Why buy an IA with a Transmeta chip when the bill of materials (BOM) suggests you should bung it in an X86 style notebook instead? We reported then that Sony and others had such cunning plans.

The second half of the year saw little let up in the continuing Megahurts war between AMD and Intel. The latter, still suffering from "tightness" (shortages) in the second half, decided to give the Megahurts Madness a miss at the second Intel Developer Forum (IDF) of the year, and instead focused on a Napster style cunning scheme to share data across corporate networks.

Everyone was now waiting for Willamette (the Pentium 4) to arrive, and Intel did leak a little info about the platform at that IDF. But it left everyone confused. Screaming Sindy II complicated the position while suits from Intel baffled us mere hacks with semiconductor science.

When the Pentium 4 arrived to hysterical excitement from the press, the szechuan wasn't that much clearer. We were in New Delhi at launch time - the Times of India opened its story about the Pentium 4 with one word - Yawn.

The hardware sites huffed and puffed attempting to compare AMD's microprocessor with the Intel offering but seemingly to little avail. The testing was complicated by the fact that there was little software that took advantage of Screaming Sindy II, and also by conflicting claims about Rambus memory and the like.

In all this welter of benchmarketing and graphs, a small silent voice (can't be our conscience, we don't have one), reminded us that when Intel intro'd MMX for its microprocessors with little software support, it didn't make a damn bit of difference. Intel MMX won the day. Marketing always seems to beat benchmarketing when the chips are down.

On the chipset front in 2000, plucky not-so-little Taiwanese vendor Via made solid progress, aided and abetted by Intel's seeming inability to get any of its chipset platforms right, and AMD's apparent reluctance to commit itself too heavily to additional projects. It seemed to us, when we visited Taiwan again in September, that Marmosetzilla was pursuing its own agenda - and we were left with the definite feeling that if it was forced to plump between INTCzilla and AMDzilla, it would choose the former, rather than the latter.

Behind the scenes, Via is quietly pushing its own microprocessors, usually codenamed after Old Testament characters, in important markets such as China, India and Latin America.

It was in Taiwan in September that we learnt the horrible truth that we'd predicted from day zero of the Coppermine platform -- Intel would not make the fat profits it usually did in the Q, mostly because of us pesky Europeans and nothing to do with its inability to make enough microprocessors.

Throughout the second half of this year, AMD was quietly cleaning up in the European market, although its grasp on the corporate sector remained slender.

But it appears now that the world+dog had got a little tired of being expected to buy ever faster microprocessors, whether from the AMD or the INTC stable. The Megahurts Madness, as we predicted here in January, had taken its toll and consumers were not voting with their feet and buying shiny bright new PCs for stacks of extra bucks just to get a few extra mips to play Quake III or whatever.

Still, both AMD and Intel (and for that matter Via) are now firmly on the 1GHz bandwagon and pretty soon your wordprocessor will work ever-so-much-faster, whether you like it or not. The advantage to us plebs is that we get more powerful PCs, the disadvantage to margins for the big semiconductor firms is clear enough. When every other PC component is dirt cheap, why should the "brains", whether it be processor or memory, cost so much?

The question for AMD, Intel and Via, as the new year, century, and millennium dawns on 1 January 2001, is whether their strategies are diverse enough to tough out the price bloodbath in the coming 365.225 days. ®