Recently, I described rumors that Apple would switch from PowerPC to x86 microprocessors as “a bunch of bull.”
OK, I was wrong. I apologize to Don Clark and Nick Wingfield of the Wall Street Journal, David Utter of WebProNews, Paul Thurrott of winsupersite.com, and all the other journalists I called “dupes” for going along with what I thought was a tired, old, often-recycled rumor.
In my defense, I should say that the only thing these reporters got right was the bottom line: Apple is, indeed, switching to x86—Steve Jobs announced the details at Apples Worldwide Developer Conference on Monday—but all of the supporting arguments offered by the Apple-Intel rumormongers were indeed “bull.”
Apple was not looking for “leverage” over IBM. IBM was already giving Apple as much support as it could, with the lowest prices it could offer on the most sophisticated desktop microprocessors it could make.
Apple was not trying to save money on microprocessors. Apple will end up paying more on average for Intel processors than it was paying for PowerPC chips from IBM and Freescale.
It will pay significantly less for support chips, but these are a smaller part of the overall system cost.
Apple was not desperate for access to Intels low-power Pentium M processor, the foundation of the Centrino platform.
Current Centrino systems are nothing to brag about, consuming more power than the original design.
IBMs PowerPC 970FX processor, which Apple calls the G5, overlaps the Pentium M in both performance and power consumption.
A moderate development effort by Apple could have brought the benefits of the 970FXs PowerTune power-management features to the Mac platform.
Apple was not looking for more performance than IBM could offer.
The 970FX is fully competitive with Intels Pentium 4 and Xeon processors, and has a superior system architecture that gives it better performance headroom for the future.
Intels microprocessor roadmap has been in disarray since it cancelled the next-generation Pentium processor code-named Tejas in May 2004; it will be years before Intel recovers from that boondoggle.
The one technical disadvantage of the PowerPC architecture today—lack of variety—is not insurmountable.
Including processors from AMD and VIA, there are a half-dozen different chip designs supporting the x86 PC market. Apple has two.
Although CPUs are expensive to develop, the cost is only a few hundred million dollars each, a small fraction of Apples cash assets.
If all Apple wanted was more variety, it could have paid IBM to develop more processors, just as Sony and Microsoft have done for their videogame consoles.
In short, there was nothing wrong with the Apple-IBM relationship that Apple couldnt have cured with a little money—the same money the company will now spend to buy Intels pricier chips.
Either way, Apples total hardware costs, including research & development, would have been about the same.
Jobs Eye on the
I believe Apple is switching because Steve Jobs has his eye on the big prize—a substantial share of the personal computer marketplace.
In fact, I believe Steve Jobs has been working toward this goal since he returned to Apple in 1997.
Remember, it was Jobs nemesis John Sculley who presided over the switch from Motorolas 68000 processors to the PowerPC architecture, missing the opportunity to gain PC compatibility.
It seemed like a good idea at the time. IBMs Power architecture was part of a wave of RISC (reduced instruction-set computing) technology that promised to sweep x86 away.
Intel couldnt make the switch, but Apple could; Sculley believed RISCs inherent advantages would give Apple a compelling advantage over x86 PCs.
What RISC proponents didnt predict was that Intel (and AMD) could create new x86 processors built around RISC-like cores.
Their higher production volumes more than compensated for the slight inefficiencies of this approach.
The new PowerPC platform also turned out to be less compelling than Sculley expected.
By forcing its customers and software developers to choose between two fundamentally incompatible platforms—the well established PC platform and Apples proprietary designs—Apple forced itself into a small market niche from which it could not escape.
Even with this years remarkable growth in Mac sales—40 percent year over year—Apple still has less than 4 percent of the U.S. personal-computer market.
The transition to x86 will cut into Power Mac sales in the short term, but the new strategy gives Apple an opportunity for growth rates that would be otherwise unimaginable.
How might this strategy play out?
I think the keys to Apples success now lie in the “platformization” strategy of Intels new CEO, Paul Otellini.
Otellini is behind Intels new emphasis on technology beyond simple microprocessor design—the Centrino platform, Intels Virtualization Technology and the security features code-named LaGrande Technology.
Apple, arguably the most platform-oriented computer vendor, will now contribute its considerable skills in hardware/software integration to Intels chipset developers.
Apple will show Intel how to make software-friendly hardware, and Intel will put its unmatched manufacturing muscle into Apples service.
The ideal future x86 Mac will run Mac OS X and Windows, but I think its unlikely that Apple will release a version of Mac OS that runs on non-Apple PCs.
Apple relies heavily on hardware sales to subsidize Mac OS development.
A shrink-wrapped Mac OS that runs on Dell machines, for example, would cut into Mac system sales.
Jobs did not address this question in his speech Monday, but we should learn the answer later this year.
If Apple had adopted the Intel architecture instead of PowerPC, this would have been a difficult problem to solve.
Apple would have been forced to make its systems fundamentally incompatible with the standard PC platform to prevent hackers from making their own Mac clones.
Today, Intel has the answer. LaGrande technology provides an unbreakable cryptographic lock that can keep Mac OS from booting on systems not made by Apple.
The LaGrande solution allows full PC compatibility, so Macs could be able to boot Windows, but dual-boot systems have never been particularly successful. Users dont want to be forced to choose between multiple operating systems when they start their computers.
The ideal solution would offer access to all the software and all the data on the machine at the same time.
Enter another one of Intels platform pieces, VT (Virtualization Technology).
VT makes it possible for one machine to run several different operating systems at once.
Intel has partnered with software virtualization pioneer VMware to implement its own software layer for VT; Microsoft will have another.
VT demos have been fairly primitive so far, forcing users to switch from one virtual desktop to another to run software in different partitions.
Microsoft has the technology to create a more natural windowed environment, but so does Apple—and Apple has proved more agile in developing user-interface technology over the last few years.
Again, Jobs said nothing about this prospect, but I know Apple could make this work, and I doubt theyll overlook the opportunity.
Properly implemented, an x86 Mac wouldnt need to boot Windows to run Windows software.
Mac OS would be the primary operating system, but if the customer wants Windows, Windows could get its own partition.
With Windows running on the same machine, Apple can make Windows applications part of the Mac OS X environment.
Apple could end up with the best of all worlds—simultaneous Mac OS and Windows operation on a wide range of commodity platforms.
Today, it isnt practical for Apple to develop its own tablet computers or eight-way servers because of hardware engineering costs.
With suitable hardware available off the shelf in the PC industry, Apple can create such systems just by doing the necessary software development. Most of this work, in fact, has already been done.
Reaching this promised land will still take a lot of hard work by Apple and its independent software developers.
Apple is targeting the 64-bit mode of Intels x86 processors (the mode originally developed by AMD and dubbed AMD64).
Apple already has 64-bit support in Mac OS X 10.4 (Tiger), but almost none of the Tiger code runs in 64-bit mode. Apple will have to make the transition to x86 and 64-bit operation at the same time.
Its unclear how much of this work has been done. Jobs announced that for the last five years, it has pursued a cross-platform development strategy; Apples operating systems and applications have all been built and tested for x86 and PowerPC compatibility.
But Apple hasnt had access to 64-bit x86 platforms for all this time. I think its likely that the 64-bit transition is still under way in Cupertino.
Its ironic that up in Seattle, Microsoft is moving the other way.
We usually think of Microsoft as a software company, but it sells many more Xbox consoles than Apple sells Macs.
With similar needs for multimedia processing and price/performance, and a large installed base of x86 software, Microsoft selected PowerPC for its next-generation Xbox 360.
For similar reasons, Sony is moving from MIPS processors to PowerPC in PlayStation 3, and Nintendo is sticking with PowerPC for its forthcoming Revolution system.
IBM designed all three of these new PowerPC processors; together, the three consoles will ship almost as many processors as Intel.
Apples future includes less RISC, but more risks.
Faced with a straight-up choice between Windows Longhorn and Mac OS X “Leopard” on the same hardware, some Microsoft customers will switch—but will there be more switchers than Apple would have attracted to the PowerPC platform?
And what about Apples short-term prospects? Pending the arrival of better Intel microprocessors, the first generation of x86-based PCs wont be dramatically better than the new Power Macs Jobs promised us.
Power Macs will also have better software support for years to come, but will Apples existing customers be comfortable buying a platform that is scheduled for cancellation?
Apple is looking at a year or two of combining nervous uncertainty with the hope of fantastic success. Realizing this dream will require a lot of engineering effort from Apple and Intel, and a lot of faith from Apples faithful.
Peter N. Glaskowsky is an analyst with the Envisioneering Group in Seaford, N.Y., a former editor of the Microprocessor Report newsletter and an architect with MemoryLogix, a microprocessor design firm.
Check out eWEEK.coms for the latest news, reviews and analysis on Apple in the enterprise.