Mimsy Were the Borogoves

Editorials: Where I rant to the wall about politics. And sometimes the wall rants back.

Apple switches to Intel processors

Jerry Stratton, June 15, 2005

So, Apple has switched from PowerPC to Intel for their central processors. It’s been big news, but I don’t think it’s a big deal. Apple has been planning this for five years. When they bought NeXT, it was cross-platform, and they have ensured that it has remained cross-platform.

For most people, this will be an invisible migration. They will buy a new computer, and their software will continue to work just as it always did. All of their Apple software will be recompiled to work with Intel; all of their commercial software that they buy will be a universal binary that works on both Intel and PPC; and the rest will be invisibly converted from PowerPC code to Intel code every time they start it up.

Note that this is not the first time that Apple and Intel have worked together. Apple was the driving force in popularizing Intel’s USB technology when Apple introduced the USB port on their first iMacs.

Why stop using the PowerPC?

The real question isn’t “why has Apple stopped using PowerPC,”, but “why has Apple stopped using custom chips”. For over a decade, Apple has been a partner innovating new chip technologies with Motorola and IBM. This brought us the PowerPC, the G3, Altivec on the G4, and the G5.

It’s been great for processor technology. But in those ten years, Motorola spun off their processor division, and IBM is now making game chips. Because Apple does not have its own chip fabrication facilities, the actual manufacture of the chips the group designed have always been made by someone other than Apple: first Motorola, then Motorola and IBM. Because of this, Apple has always been extremely dependent on those other companies. It’s a running joke that whoever Apple teams with ends up having fabrication problems. First Motorola with G4 stagnation, and now IBM with G5 stagnation.

They’ve been, effectively, custom-designing CPUs. This makes for great leaps, but it has become unsustainable in the long-term. While the new designs have, when they were created, been very promising and better than the comparable Intel chips at the time they were designed, they have never been able to keep up with later speed increases by Intel. Apple’s custom chips have been clearly better when they were first introduced, but over time have, if not fallen behind, lost their lead.

With the G5, the development stagnation really hurt Apple in their portable line. Over two years after it was introduced, the G5 still uses too much power and generates too much heat to be used in a portable. This has left Apple in the bind of having to introduce inexpensive consumer-level iMacs that are significantly more powerful than their pro-level PowerBooks.

Going with Intel processors means not having to spend time innovating new processors to compete with Intel. If Apple’s Intel processors are no faster than the Intel processors of their Windows-based competitors, they won’t be any slower either. They’ll be the same or virtually the same chips.

Apple will be able to focus their time and money on their software and on the design of their hardware. And they’ll have to, to distinguish themselves from their competition.

Why now?

We can easily see why Apple is choosing to move now by looking at IBM’s response to the switch:

IBM is aggressively moving the Power Architecture beyond the PC, as shown by our recent successes with the next-generation gaming systems announced by Sony, Microsoft and Nintendo.... IBM is focused on the highest value opportunities in each marketplace, and our direction with the Power Architecture is consistent with that strategy.

IBM is moving “beyond the personal computer”. They sold their own personal computer division. It makes sense, then, for IBM to drop the personal computer market. The game system market and the server market are very different from the personal computer market. Game systems come out once; they last several years, and then a completely new system, usually with backward compatibility, comes out. The personal computer market currently requires continual speed updates, but game systems do not.

Intel is not likely to ever decide that they don’t want to make CPUs for personal computers. It is their core business. And if they do, Apple has places to go that do not require any migration.

Why the low-end machines first?

In the past, Apple has introduced new chip designs on their high-end machines first. But this is a very different migration. When Apple moved from the 68000-series to the PowerPC, and from the PowerPC to the G3, and the G3 to the G4, and the G4 to the G5, they were moving to more advanced, technologically better solutions. With the Intel migration, they are moving to a safer solution, but one that is not (yet) technologically more advanced.

This means that their high-end customers will not necessarily see a major speed increase with the transition. And since their high-end customers often use non-Apple software, Apple has no control over how fast those software makers come out with Intel-based versions of their software. Apple is going to ensure that PowerPC software runs on Intel using “Rosetta” translation software, but Rosetta will not necessarily be reliable, functional, or fast for all software. I’m sure they’ll optimize Rosetta for Microsoft Office, but there will be problems elsewhere.

Apple is introducing it on the Mac Mini (and probably the eMac) first because those customers use Apple software. The consumers who buy these low-end machines are the kind who don’t buy extra software. Why should they? The Macs include just about everything people need: all of the iLife apps as well as AppleWorks. The only software they’re likely to use that Apple doesn’t make is Office, and Apple can ensure that Rosetta works with that one piece of software if necessary.

The other major market for the Mini is Unix geeks who want a cheap Mac OS X box. They compile their own software or use scripting languages; their software will work as soon as they recompile it.

And, finally, these lower-end machines are also more likely to get a major speed upgrade by moving to Intel and they are not a technological step backwards. The Pentiums are still 32-bit (with 64-bit extensions). The G5 in Apple’s high-end offerings is 64-bit. While Intel may have introduced consumer-grade 64-bit chips by next year, there is no guarantee, so Apple is doing the safe thing and replacing their current 32-bit computers with Intels first; even if the Intel chips are still 32-bit when the transition occurs, this won’t be a downgrade.

Does this mean I should wait to buy a Mac?

My advice, as always, is to wait until you need a new computer, and then don’t wait any longer. This transition does not affect that. Unless you’re a developer, there is no need to wait because of this. The applications you buy today will continue to run on the computer you buy today. Upgrades will continue to run on the computer you buy today until at least the end of 2007 and probably until the end of 2009 or later (or, if they don’t, it will be for different reasons).

If you wait to buy anything, it should be software. Any software you buy today will need to be updated to run on future Macs. Unless an application is a “fat binary” (universal binary), it can’t run on both versions of the Mac.

I don’t see anybody needing to wait, but I do see some people needing to buy earlier. The only people whose purchase plans are likely to be affected by this change are those who need the PowerPC. If their upgrade cycle would otherwise put them past the Intel changeover, they’ll want to buy their next computer a little early.

Anyone who continues to rely on software that runs only in “Classic” mode, under Mac OS 9, may wish to purchase one of the last PowerPC Macs to come off the assembly line. Rosetta doesn’t run Classic applications, and this will give them that much more time to find replacements. But most Classic users don’t need an upgrade cycle at all: they aren’t buying new stuff.

There are possibly a few scientific and technical users who are making use of the Altivec extensions in the G4 and the G5. They may wish to get the last PowerPC so as to put off purchasing a non-Altivec Intel-based Macintosh. But eventually, even the benefits of Altivec will fail to overcome the natural speed increases of newer processors.

Will I be able to run Mac OS X on any Windows PC?

No. If you want OS X, you’ll have to buy a Macintosh. This might be a first step for Apple to produce an off-the-shelf version of Mac OS X for third-party computers, but that is a lot of work with little pay-off for Apple. It is more profitable for Apple and more reliable for the consumer if Apple manufactures “the whole widget”.

What about my software?

Most software will just work. Apple has made it very easy for software makers to recompile their software as a universal binary that runs on both PowerPC and Intel. If you have one of those applications, it will run on your new Intel-based Macintosh natively. For PowerPC-only software, Rosetta will translate it to Intel code when you double-click it. However, Rosetta does have some limitations:

  • Classic applications--those written for Mac OS 9 or earlier--will not translate;
  • Applications that require AltiVec extensions, or those that require a G4 or G5, will not translate;
  • Some system-level enhancements, such as kernel extensions, and applications that depend on them will not translate;
  • Java applications that are not cross-platform will not translate.

For most people this means that if you’re using any software that will not run on a G3 or that still requires Classic mode, you’ll want to upgrade that software before you buy an Intel-based Macintosh. If the software already runs on any Macintosh, it will continue to run on any Macintosh when the first Intel-based Macintosh comes off the line.

Will I be able to run Windows applications on Mac OS X?

Currently, emulating Windows on the Macintosh requires three things: emulating the chip that the software runs on; emulating the hardware; and emulating or providing the operating system. One-third of that just disappeared.

It is possible for an enterprising third-party to create an X-Windows like system that recreates Windows system calls on Mac OS X. This is likely to be difficult. It will, however, be easier for software such as VirtualPC to run a licensed copy of Windows--but you’ll still have to buy a copy of Windows.

As far as installing Windows on a Macintosh and rebooting when you need to use Windows, this is also theoretically possible, but also requires enterprising third-parties. Apple has stated that they will not support it, but will also not block it. Normally, the manufacturer of Windows hardware provides special drivers that Windows needs. If Apple is not going to write these drivers (and they shouldn’t--they’re writing Mac OS X), someone else will have to.

Given how annoying dual-booting is--you’d have to reboot your computer every time you went from one operating system to another--I’m not sure how popular such software would be.

Shifting the focus of competition

While the geek in me is very pleased that Apple has created an elegantly CPU-independent operating system, I am also disappointed that we will no longer see an architecture-based competition between CPU designers. This competition has brought us some fascinating, useful, and powerful advances in chip design. It’s been exciting, for computer geeks, to wake up Christmas morning (in late June) and see the G5 specs waiting under the Apple tree.

The problem is that while there was exciting competition between chip designers, there was no such competition between PowerPC manufacturers. So the designs have been powerful but advancement has stagnated. For Apple, every major increase in chip speed and design came because Apple got together with their partner, and the team designed a new chip to Apple’s needs. The PowerPC, the G3, the G4, and the G5 all promised to make Apple computers faster than anything else out there; and they all failed to live up to that promise. The initial design was powerful, but then a lack of competition caused speed increases to stagnate.

Meanwhile, in the Intel-compatible world, the competition between Intel-style chip manufacturers ensures that new Intel designs continually speed up. If Apple ever decides that Intel is slowing down, they just go to whoever is faster.

While teaming with other companies to design innovative computer chips meant that Apple could do great things, it also meant that they were reliant upon companies that did not necessarily share Apple’s vision. That Intel-style manufacturers don’t necessarily share Apple’s vision either doesn’t matter: as long as Windows computers still use Intel and as long as Windows is Apple’s competition, Apple’s chips, while they may not be faster than the competition, will also not be slower. And if Intel, AMD, and every other Intel-style manufacturer stop innovating, then Windows will slow just as much as the Mac OS will.

But this does likely mean that CPU innovation will slow down. Innovation in new chip designs has come from cross-CPU competition.

  1. <- British Police State
  2. DEA Needs Drugs ->