Having read the CNet News.com story about Apple's supposed impending switch to x86, let me propose an excellent code name for this forthcoming system: “Nessie.” Like Nessie's namesake Loch Ness Monster, the rumor of Mac OS on x86 rings of the stuff of tabloids, not something that people take seriously. Of course, that leaves us to ask what we are to make of it when one of the most respectable online computer news sources, News.com, reports as virtual fact that Apple will be switching to Intel, and the story apparently seems credible enough to get Reuters to pick it up.
That's a question probably most analysts are wondering this weekend as we wait for Steve Jobs' annual WWDC keynote on Monday. Has Jobs fallen off his rocker or is this simply another rumor on par with the perennial Apple PDA announcement that has never occurred? To be sure, that it is not reported as mere rumor, indicates that News.com thinks they are on to something; any journalist knows that as tempting as the “big scoop” may be, a reputation is built slowly but can be crushed by one under researched and over reported story. Just ask Dan Rather.
Unless Jobs is trying to dry up the demand for Mac hardware, so as to be able to concentrate more fully on everything iPod, I will submit four points as to why Apple will not be switching to Intel.
1. The PowerPC Revolution
We are less than twelve months away from the biggest thing to ever happen to the PowerPC architecture. By Christmas, the item that every kid is going to want will run PowerPC (yes, I am referring to the Xbox 360), and Christmas 2006, will be the same, as Nintendo and Sony join the foray into the modern Power-based architecture. This means, in the most simple terms, that tens of millions of Power-based chips will begin shipping for use in devices not all that unlike PC's. This will mean thousands of developers will become familiar with the architecture and how to make it work best in graphically intense applications, a very desirable situation for a company like Apple.
Moreover, with Xbox 360 game developers using PowerMac G5's presently (and Microsoft using the same to demo the games of the new system), Apple presently has the opportunity to offer the ultimate gaming PC — porting a game to another OS on the same PowerPC G5 architecture would probably be easier than porting it back to Windows, thereby placing Apple in the desirable spot of another segment of the high end PC market.
With both IBM and Freescale moving toward powerful, yet less power hungry dual core PowerPC processors, this seems like an unlikely time to abandon the platform Apple has spent over a decade promoting. Unlike when Apple made the jump to PowerPC in the 90's, the company does not find itself on a dying platform presently.
2. Leaving a Legacy
Any platform producer knows the necessity of supporting legacy software. This has become a major hassle for Microsoft and its aging DOS/Windows API that it has been forced to keep pulling forward, but it has been an issue Apple has had to deal with as well. Had Apple not managed to create a fairly decent legacy solution in the form of Classic within Mac OS X, it seems a near certainty that Mac OS X migration would have been much slower, and it didn't start off all that fast as it was. Unlike a processor switch, however, Apple was able to bundle Mac OS X along side Mac OS 9, quietly building up market share until the time when developers and users were ready to make the jump. Short of making a system with two different types of CPU's in it, such a smooth migration simply is not possible with a PowerPC to Intel switch.
Apple would face a difficult decision if it made the jump. Do you support old PowerPC programs through emulation? It would be virtually necessary to do so, if it wishes to retain any respectability, but doing so would almost definitely provide less than satisfactory results. Short of getting Intel or AMD to produce an x86 with some kind of Transmeta “Code Morphing”-like technology, emulation speed would almost certainly be dismal. Although Microsoft Virtual PC for Mac does a decent job on non-graphics accelerated software (read that carefully while remembering that Mac OS X thrives on graphics acceleration thanks to Quartz Extreme), it is still not a speed demon after years of development, and even if Apple could achieve similar results going in the opposite direction, who would trade in their speedy PowerMac G5 for a system that runs their software in a slower emulation mode? Intel has been learning this the hard way with respect to their Itanium processor line and its poor 32-bit x86 emulation.
This is a very different type of emulation from the kind that Apple did over a decade ago when it switched to PowerPC from the M68k processor series. Like Palm's similar switch from M68k to ARM processors on its PDA's, Apple's switch was from a rapidly aging platform to a powerful new one that can easily make up for the emulation overhead with the major speed gains. Even if we presume that the best x86 processor is every bit as good as the best PowerPC processor, nay, that it is even better than the best PowerPC processor, no x86 processor is better by a large enough margin to make up for emulation overhead, especially since this would be RISC-to-CISC emulation.
Moreover, such drastic emulation is hardly ever perfect. Many have speculated that Microsoft purchased Virtual PC to use for the purpose of Xbox compatibility in the new Xbox 360. While it is not entirely clear how Microsoft will achieve backwards compatibility as it transitions from x86 to PowerPC in its gaming machine, it has already made it clear that it is not aiming to be compatible with all first generation games. If Microsoft cannot achieve compatibility with a just hundreds of games officially released under its own licensing blessing, how will Apple achieve compatibility with thousands of Mac OS X and OS Classic applications that were developed under much looser controls?
3. A Need for Stability
Many computer companies have tried to survive the switch from one OS to another or from one platform to another, all but Apple have failed to smaller or greater extents. Be and NeXT, two companies that had ties to Apple in various ways, are two interesting examples. Even Microsoft failed in its attempts, which included making Windows NT available on PowerPC and Alpha architectures.
Apple has succeeded in two major transitions, first the switch from M68k to PowerPC and secondly the switch from OS Classic to Mac OS X, but each switch has had its costs and the latter transition is still in the process right now. The former, as I have already noted, was eased greatly by the fact that M68k applications could be run on the PowerPC via decent emulation, and for that matter, can still be run on Mac OS X under Classic. The latter was eased by Classic itself, again, as has already be noted.
This stable transition cannot be repeated on a switch to x86. The M68k phase out was made much easier because developers could make “fat binaries” that could run on both M68k and PowerPC natively, similar to the way Cocoa allows for applications that can run on both OS 9 and OS X. This definitely eases the transition for developers who are loath to have to support separate code bases, or even binaries, for different platforms. Creating PPC/x86 fat binaries seems out of the question, thus leaving developers to contemplate whether to develop for the new x86 Mac platform, for which there are no users, or the old PPC Mac platform for which there are millions of users. Guess which one they are going to pick? Likewise, when users contemplate whether to buy an x86 Mac, are they really going to jump when their favorite applications are only available for PowerPC? You can see a vicious cycle, a catch-22, here.
Even putting that aside, Apple's third party developers have spent years optimizing their code for PowerPC specific technologies such as the AltiVec vector processor and now the G5 64-bit processor. If anyone thinks that Adobe and other high-end multimedia software producers are going to like the idea of having to rework their Mac code bases to use MMX/SSE/SSE2, which arguably are not even going to begin to match AltiVec, and x86-64, contact me. I have some premium swampland for sale, and I know you will find it worth purchasing. Another destabilizing change on the Mac platform so close to the last one, which relative to the stable PC platform, was not that far off from the one before that, is probably more than either users or developers are willing to bear.
For those who have not run off to write me about my real estate holdings, I present the real threat of Apple making itself vulnerable in this way. GNU/Linux as a desktop OS has been growing for several years, and some distributions, such as Linspire, appear poised to go mainstream. If Apple opens itself up to a major shake up of its platform again right now, and one that promises nothing more than to make Mac OS X run on the same platform that Windows and GNU/Linux already run on, GNU/Linux distributions will have a major opening to go on the attack and grab Apple users who are trying to decide what to do about this disruptive migration. Microsoft is scared of GNU/Linux; if Apple is not, then their Reality Distortion Field has turned against their execs.
4. The Microsoft Factor
Speculation has gone on for years as to whether Apple could make a go of offering Mac OS X for the general PC populous similar to the way Microsoft has done with Windows. This is impractical for a company that aims for a “just works” experience, since not even Microsoft can offer drivers for every possible combination of PC hardware. It is only through Apple's vertical solution stack that the company has been able to insure a reliable, easy to use experience. GNU/Linux, while far more of a native to x86 than Mac OS X is, has struggled for years to support all of the hardware combinations on the x86 platform. Even though it offers probably a larger array of drivers out of the box than Windows, it still supports less hardware than Windows. The why is obvious: Microsoft has managed to get most equipment manufacturers to produce their own Windows drivers; the likelihood that Apple could pull the same thing off is unrealistic.
At the same time, even if Apple ends up building its own x86 hardware, perhaps with a few proprietary hooks to keep people from installing Mac OS X on a Dell or HP system, Microsoft is still going to be more than aware that at any time Apple could choose to go the other route and try to make deals with PC OEM's. Thus expect any release of Mac OS X for x86 to endure Microsoft's rule against offering any software for alternative x86 OSes. This could be summed up as follows: if you come onto Microsoft's processor turf, don't expect Microsoft's help.
For those unfamiliar with this policy, I would point to the most clear-cut example of this policy in the form of Internet Explorer. During the height of the browser wars, Microsoft saw it to its advantage to support a version of Internet Explorer for UNIX. The product was released for free, and made available for two particular UNIX operating systems: HP-UX and Solaris SPARC. Despite the desire of users, and perhaps what would have served to further expedite Microsoft's war against Netscape, Microsoft never released IE for UNIX on an x86 UNIX or Unix-like system, such as FreeBSD, GNU/Linux, UnixWare, OpenServer or even Solaris x86.
Microsoft wisely sees the benefit of getting its product onto computers that will never run Windows. Microsoft Office:mac came off the endangered species list about the time Microsoft killed its Windows NT PPC initiative, for instance. Macintosh users want Office, so as long as their systems cannot run Windows, why not sell it to them? Yet, if Macs are essentially only glorified Intel machines, and at any time, Apple could make OS X run on any Intel machine, it would be naïve to expect the Redmond-based Office developer to be keen on continuing Office:mac development, as doing so would position Apple in a place that would allow it to perhaps come out of its niche and right into Microsoft's well entrenched x86 desktop market.
On the Other Hand
In the above few paragraphs, I believe I have shown why it would be suicidal for Apple to make a jump to x86, and therefore why Jobs will not announce the switch to Intel processors. At the same time, one news event from last week may prove the most interesting hint that Apple will insanely (but not greatly) choose to go with x86. According to the Wall Street Journal, Intel CEO Paul S. Otellini acknowledged, albeit grudgingly, that Mac OS X may be worth considering for those who want a short-term fix to Windows' security woes. Why would the Intel CEO, even when pressed, do anything but deny the usefulness of switching to a non-Intel alternative? It sounded peculiar at the time, now it sounds intriguing.
Alternately, perhaps Apple has inked a deal with Intel, but not to switch its computer line over to Intel processors. The ever present Apple PDA rumor mentioned above, if true, could run on Intel StrongARM processors, or perhaps another favorite Apple rumor, a media center, might do the same. Many major Apple rumors have turned out to have substantial truth to them, while not being exactly on the mark; this supposed leak of Monday's keynote announcement could prove the same.
I can see no reason why Apple would essentially set itself on such a risky (but not RISCy) course when sales are up far beyond similar same-quarter sales, for other first tier PC makers, and the PowerPC's future looks far brighter than ever before. Until Jobs says otherwise on Monday, people would be far better off researching more reliable rumors. The present location of a sasquatch, for instance.