Intel-mania

Moore’s Law states that every 18-24 months, CPU processing power doubles.

Here then is another law: every 18-24 months, the tech industry will be captivated by unsubstantiated rumors that Apple is switching the Macintosh to Intel (or Intel-compatible) processors. Cf. “Qwerty”, published here two years ago in the face of very similar rumors.

The current round of Apple-Intel rumors stems from this report earlier this week in The Wall Street Journal’s “Heard on the Street” column — the Journal’s equivalent of a rumor column. That link requires a paid Journal subscription, but Paul Thurrott has a long excerpt on his weblog. Here’s how the Journal sources the information:

Two industry executives with knowledge of recent discussions between the companies said Apple will agree to use Intel chips … Talks between Apple and Intel could founder, as they have before, or Apple could be engaging in negotiations with Intel to gain leverage over IBM.

Note that the Journal does not imply that either “industry executive” works at either Apple or Intel.

This rumor has gained traction because it’s fairly obvious that Apple is not happy with IBM’s G5 production. When the PowerMac G5 was introduced at WWDC 2003, Steve Jobs famously predicted they’d have systems running at 3 GHz within a year. It’s now two years later and Apple’s fastest system runs at only 2.7 GHz. That’s not to say a top-of-the-line PowerMac G5 isn’t a nice computer, but you certainly don’t hear any talk about them being the fastest PCs in the world anymore.

So, yes, there’s a motive for Apple to consider such a switch. But that doesn’t make it plausible. None of this week’s Apple-Intel rumor reports seriously address the enormous hurdles Apple would face if they made such a switch.

The biggest of which is, simply, software compatibility. All existing Mac OS X software would need to be recompiled for an Intel processor architecture. A decade ago, when Apple switched the Mac from Motorola’s 680 × 0 family of processors to the PowerPC, the transition was nearly seamless because the PowerPC was capable of emulating the 680 × 0 at very reasonable speeds. But emulation is out of the question for a switch now — Intel chips may be faster than current PowerPC G5s, but they are nowhere near fast enough to emulate them at an acceptable speed.

The only plausible scenario I can imagine would be for Apple to pre-announce the move to x86 (say, at WWDC) to get developers on board a year or more in advance. The idea is that by the time Apple released the new Intel-powered Macs, developers would have had time to develop, test, and release Intel-compatible software updates.

The problem with this scenario is not technical. It’d be a piece of cake for Apple to roll out an update of Xcode that generates such dual-binary apps — the compiler at the heart of Xcode is GCC, and if anything, GCC is better at generating x86 code than it is PowerPC code. Darwin already officially supports x86 processors, and it seems quite plausible that Apple secretly keeps the rest of Mac OS X’s source code compilable on x86 processors. (NextStep supported multiple processor architectures.)

No, the obvious problem with this idea is marketing: the minute Apple announces they’re moving to x86 processors, sales of current hardware dry up. Who’s going to spend $3000 for a deprecated CPU architecture?

But they’d have to pre-announce the move in order to give developers time to recompile — and in some cases re-write portions of — their software. Apple couldn’t just spring the new machines unannounced; who’d buy a Mac that ran no existing third-party Mac software?

My advice is to pay no heed whatsoever to any Macintosh-to-Intel rumors that don’t address this issue. The fact that it’s technically possible doesn’t mean there’s a good business case for such a move. It’s wise for Apple to keep such a move available as an option, in case something drastic happens to the PowerPC processor family. But the current performance gap, while serious, is far from drastic.

Also note that it is entirely possible that Apple is planning to use Intel chips, but for something other than Mac CPUs. Perhaps a next-generation iPod, or a new iPod-like consumer electronics media gadget. Or maybe a next-generation AirPort system, with higher bandwidth and range, based on WiMAX. Such a deal would make perfect sense — Intel makes great chips, and Apple has been making great new products other than Macs. But that’s not what the Journal and others have reported; this week’s rumors are that Apple is moving the Mac to Intel.

(Paul Thurrott even managed to attach the Mac-to-Intel rumor to the other great recurring Mac rumor fantasy — the fabled “tablet” Mac. Just remember that a tablet-shaped gadget is not necessarily a tablet-shaped Mac.)

Reading between the lines, I think this is less about whether Apple actually intends to switch processors, and more about planted leaks intended to spur IBM. (For what it’s worth, the Journal article mentions this as a possibility; but few of the sites that breathlessly linked to the story mentioned anything more than the “The Wall Street Journal says Apple is moving the Mac to Intel!” part.) More than just the money IBM makes from Apple under their current arrangement, there’s also the pride/publicity factor: the underlying theme of this rumor is that Apple might turn to Intel because IBM can’t compete against Intel’s technology. Whether it’s a fair assessment or not, it’s not the sort of idea IBM wants in the public’s conventional wisdom.

My prediction: I’ll be writing about this again in 2007.