Stardate 20021019.0621

(On Screen): I don't understand the psychology of people who nervously clutch religious icons and hold them out to ward off evil. Long ago I remember seeing pictures of villagers in Mexico whose village was threatened by lava from a volcano, carrying their local statue of the Madonna out to look at, and stop, the threat. I recall thinking that educated Americans, of whatever faith, would never believe that such a thing would work. Certainly well-educated and sophisticated Americans would never think that icons of their religion could ward off evil, could they?

This was before the creation of the Macintosh, however, and before it fell behind in technical performance and in market performance, and before Steve Jobs returned from the dead to lead the Mac faithful to the promised land. His presentations at various MacWorld's are virtually indistinguishable in style and substance from a revivalist minister, and his disciples routinely wave various icons towards the x86/Windows volcano to stave off extinction.

Sometimes the icons look like jelly beans.

Chief among those icons is the holy Altivec, the true source of computing virtue and wonder, capable of making the blind see and the lame run, unless you're a PC user blinded by the reality that the Mac is slow, or unless the lame is Apple's new OSX which limps on the fastest Mac available.

For the one thing that no zealot for any religion can ever admit to is the possibility that his religion is actually inferior to another. Wars have been started for less.

A few months ago, it was leaked that IBM was going to work on a cut-down version of the Power 4, which would be intended for use in high-end workstations and hmmm..., maybe, you don't think, possibly in Macintoshes too? Could be?

Now it's more or less confirmed that such a thing is coming, sampling next summer, with volume production in the second half of 2003. It should probably be noted that this kind of thing is notorious for missing schedules, not to mention the fact that there will be a lag time from when IBM has parts available in quantity until when they're actually available in consumer products. Quantity shipment in Macs almost certainly won't begin until early in 2004.

Apple has a long tradition of solving problems that don't exist until after a solution has been found, and so it is here: The Macolytes are convinced that this new chip, tentatively named the PowerPC 970, will once again catapult the Mac into the performance lead over the heathen PC – despite the fact that the Mac is already faster, since the Mac is invariably faster. You see, there's this thing called the "Megahertz Myth" which says that no matter how much faster the PC gets, the performance of the Mac improves enough to stay in the lead without actually changing its hardware in any way. It's a miracle!

It must be a miracle, because I've never seen any other technology whose specifications change so rapidly without actually being upgraded. Older Macs of certain clock rates seem to improve in speed over time. It's remarkable indeed.

You see, the one true path to computing salvation is Photoshop filters. No one ever does anything besides running Photoshop filters. It is the first thing they do in the morning, the last thing at night, and everything in between. The entire computing universe is obsessed with algorithmically distorting photographs.

And there is no faster way of attaining computing salvation than by wielding the magic sword of Altivec, which cuts through these filters in a way which leaves all other computers behind. Surely Altivec is the most holiest ultimatest computing engine that can exist? After all, whenever the holy Jobs proves to us that the Mac has yet again increased in speed without actually increasing substantively in clock rate, it is Altivec which yet again leads the charge against the infidel.

One wonders why it wasn't named Altivorpal, given its amazing ability to slay even the most complex of problems in a single clock cycle, or at most two.

In its presentation, IBM provided some tentative numbers for the expected performance of the PowerPC 970, and acolyte Waymen Askey compares them against the competition to see how it will land. And he is reassured to find that the 970 is among the fastest processors in the land – or at least, it would be if it were available today rather than a year from now. So if Intel and AMD would just promise to stop improving their products for the next 16 months, then everything will be wonderful. Alas, Intel is unlikely to cooperate in this rosey scenario:

The SPECfp results indicate that the PowerPC 970 could outperform today's Pentium IV. But once again by next year the Pentium IV will be nearing 4 GHz which could push its SPECfp results to over 1300 (more likely closer to 1200). Thus in floating point performance a 4 GHz Pentium IV should still outperform the fastest PowerPC 970 (although I would expect the results to be closer than under SPECint).

Unacknowledged by Askey is the fact that the performance figures for the 970 are highly speculative, given that not even IBM has working silicon yet, and the fact that the numbers provided were for the 970 running in its full 64-bit mode. If the 970 is incorporated into a Mac, i