This article documents a current event. Information may change rapidly and initial news reports may be unreliable. The latest updates to this article may not reflect the most current information. |
At the 2005 Worldwide Developers Conference (WWDC), Apple Computer CEO Steve Jobs made the historic announcement that the company was beginning a transition from the use of PowerPC microprocessors supplied by Motorola and IBM in their Macintosh computers, to processors designed and manufactured by Intel, a chief supplier for most of Apple's competitors.
Precedents
The Macintosh line underwent a similar transition in the 1990s, when Apple switched from the use of Motorola's 68K series of chips to their PowerPC processors, developed jointly with Apple and IBM. This took several years, and was accomplished by Apple producing versions of the Mac OS which could run on either platform, fairly low-level emulation of the 68K architure by the PowerPC models, and third party developers releasing "fat binaries" that could run natively on either architecture.
More recently Apple has brought the Macintosh line from the earlier Mac OS family to Mac OS X, a Unix-like operating system with a different user interface. This transition also took a number of years (a small percentage of older Macintoshes still run the earlier operating system), and was facilitated by the inclusion of Classic, an evironment in which an instance of Mac OS 9 could be run, permitting the execution of programs which had not been ported to Mac OS X.
A long-rumoured internal project within Apple, known as "Marklar" was designed to ensure that builds of Mac OS X were sufficiently portable as to compile for both PowerPC and x86-class processors. Jobs confirmed this, stating that every version of OS X had in fact been compiled for Intel processors as well as PowerPC. It is not known what other processors, if any, Apple maintains current builds for.
Reasons
Jobs stated that Apple's primary motivation for the transition was their disappointment with the progress of IBM's development of PowerPC technology, and their greater faith in Intel to meet Apple's needs. In particular, he cited the performance per watt (that is, the speed per unit of heat generated) projections in the roadmap provided by Intel. This is an especially important consideration in laptop design.
In 2003, Jobs had introduced Macintoshes based on the PowerPC G5 processor and promised that within a year the clock speed of the part would be up to 3 GHz. Two years later, 3 GHz G5s were still not available, and rumours continued that IBM's low yields on the POWER4-derived chip were to blame. Further, the heat produced by the chip proved an obstacle to deploying it in a laptop computer, which had become the fastest growing segment of the personal computer industry.
Some observers were surprised that Apple had not made a deal with AMD, which has in recent years become a strong competitor to Intel, sometimes introducing technologies more quickly than the traditional industry leader. AMD's shorter track record and smaller production capacity, and Intel's significant brand awareness among the consumers and ability to also provide Apple with complete motherboard designs, have been offered as possible reasons for the choice of Intel.
Benefits of the Move
Advocates of the transition point out software benefits. Technical users will appreciate the ability of Apple systems to run all four classes of software at near native speeds; OS X binaries, Java applications, GNU/x86 applications and potentially now Win32/.NET/x86 applications. No other hardware vendor can offer more than three of these. Virtual PC, a Windows emulation solution for Apple PowerPC sold by Microsoft, could now enjoy much more success with performance improved through virtualisation rather than emulation. For those customers wishing to achieve a more conventional environment, a dual, triple, or even quadruple boot solution (with OpenSolaris say), would be possible on an x86 Apple device. Apple have already indicated they do not intend to take steps to prevent other operating systems being deployed on their new machines.
Although most games are constrained through the use of DirectX API's not available for the Apple architecture (on either processor type), reductions in the time required to port these from Windows nevertheless might be observed if developers are able to ignore endian issues associated with moving from x86 to PowerPC.
Hurdles Associated with the Move
Not all the outcomes are positive however, and the Macintosh community has voiced its fear and uncertainty since the announcement was made.
Psychological
Apple has benefited greatly among its user community from the psychology of "thinking different." Many Apple users have enjoyed the ready availability of a consumer desktop that was completely separated from the "Wintel" alliance, and ads claiming the PowerPC architecture as superior had been a key part of Apple advertising for many years. With that separation gone, many longtime Mac users have expressed fear about Apple's future, and whether Apple's brand identity will be preserved. Some of the fears expressed include
- forced usage of the Intel Inside marketing campaign, including the decals and jingle;
- concerns about the x86 instruction set, and whether it will affect system stability and application quality;
- the possibility of Windows running natively on Macs; and
- the early announcement of the change causing an Osborne effect.
The PowerPC's benefits, both real and perceived, have been at the heart of the debate. Advocates of this architecture suggest that x86 has succeeded in the marketplace by sheer strength of research and development cash, rather than from an inherently better design. Moreover, they suggest that a large amount of historical baggage is carried around by x86 for backwards compatibility reasons that isn't needed by Apple. Others counter that whilst the demonstration machine was a Pentium 4, at no point did Jobs say that x86 would be the final deployed architecture. Still others say that the debate is moot because of the cross-pollination of RISC and CISC concepts over the past decade, and that while x86 has its downsides, few people will ever see them because almost all machine code is now generated by compilers.
Hardware oriented
The most obvious problem Apple has to deal with in the short term, even assuming that they are able to carry users with them to the new processor, is the so-called "Osbourne effect." This is named after Adam Osbourne who was so successful at marketing his upcoming new devices that customers stopped buying the current offerings in anticipation of their arrival. The company went bankrupt before they could be completed and shipped.
Clearly Apple needed to involve developers at an early stage so that software would be available for the new machines when they begin retail. But the reason the announcement has been relatively low-key is so as not to alienate users who might otherwise have bought a PowerPC Macintosh but will now delay purchasing until the new Intel variants have become available. Apple has more cash available than did the Osbourne company, but no company will wish to sit on stockpiles of unsold products for long.
There are questions over the extent to which Apple will retain control over the non-processor components of the system design. The interior of a current Apple G5 is as much a work of art as the exterior. Apple is traditionally a systems builder and if it is simply purchasing whole, or nearly whole, motherboards and chipsets from Intel then it is not apparent how much industrial design differentiation can be expected. On the other hand, Apple is a very agile vendor with little historical baggage to carry. Intel may treat Apple rather as Ford does Aston Martin - a way to test the latest and greatest technology in a premium product hand-crafted for maximum effect before some months later the technology filters down to cheaper systems. At the very least, purchasing most components from Intel ought to guarantee cost savings at the wholesale level.
Apple has indicated that the new Intel PCs will not use their traditional Open Firmware. Some users swear by certain features in this technology (particularly "Firewire Target Disk Mode") and the loss would be keenly felt. A new Intel technology for firmware, Extensible Firmware Interface (EFI), promises more functionality and removes the traditional PC reliance on the BIOS, seen as non-standard and dated.
Intel itself is seen among the Macintosh community as being a purveyor of hot-running chips. Indeed, this unfortunate feature of the Pentium range was the subject of a mickey-taking "Burning Bunnies" advertisement by Apple. If Intel can indeed produce cooler chips (and the widely acclaimed Pentium-M series demonstrates that should be possible) then only the lingering marketing message need be overcome. However versus the current G5, floating point performance is seen as limited and the number of registers available to applications is rather fewer than in a PowerPC alternative. Moreover it is not clear whether the next generation of Intel chips will be able to match AltiVec functionality and is 32-bit only, at a point when Apple has made 64-bit a cornerstone of its marketing.
Finally it has been rumoured that Apple was backed into this position by content producers eager for Apple to take advantage of Intel's otherwise roundly condemned on-chip digital rights management (DRM). This seems unlikely since Apple would not be able to take advantage of it on PowerPC boxes and users would simply continue using their current equipment for tasks requiring DRM. Nevertheless there is every chance that this technology will be used to prevent OS X from working on non-Apple "white box" PCs, a position Apple is determined to maintain despite this limiting the potential take-up of the OS.
For Apple to allow otherwise would cannibalise the hardware sales which still form a very large percentage of their revenue. Whilst Apple devotees anticipate that a "win" for the cracking community of such scale as OS X on an unsupported but highly-desirable top-end Opteron for example, means that it will be inevitably tried, they see the demographic of people willing to accept the consequences of this (such as not being able to use Software Update for example) as being relatively small. This is an unproved hypothesis, however.
Finally the use of x86 means that software performance will be much more transparent than when features of the PowerPC enabled benchmarkers to hide behind the "MHz myth." This was a claim that clock speed hid the true story of a computer's performance as it didn't take account of differences between architectures. While ostensibly true, it allowed machines deficient in specification to be sold long after upgrades were due. Now, identical applications placed side-by-side on OS X and a competing operating system will be comparable based purely on the speed of the software.
Existing applications
Ironically, because of the NeXTStep and FreeBSD cross-platform heritage it is likely that application-related disruption will be minimal. The evolution depends somewhat on the as-yet unresolved question of whether the transition is to Intel, or to a dual-platform strategy that includes Intel.
Firstly, Java applications that don't rely on JNI, Dashboard Widgets, and scripts that execute inside an interpreter all work immediately on both processors and are immune from changes. OS X applications that can't be migrated run inside a PowerPC dynamic translator on Intel called "Rosetta." This has limitations, most particularly in that it can't run AltiVec code, but most applications that use AltiVec fall back to a G3 instruction set when AltiVec is not found and will still run in that configuration under Rosetta. Rosetta itself is broadly a better solution than Classic was for OS 9, as it doesn't require a whole OS to be loaded as a sub-system before the application can work and translations are cached for maximum performance on the second and thereafter executions.
AltiVec itself has been encapsulated since OS 10.3 by a vectorisation library that should enable vector-aware applications to be ported readily. It is understood that Intel's SSE3 is being extended to make it "better for games" but this may in fact include certain AltiVec-related changes as well.
A simple recompilation step that generates a Universal (previously called "Fat") Binary is expected for Cocoa applications. Carbon applications may require some additional tuning but not of the complexity of the transition from OS 9. It is worth noting that this is only likely to happen when the target wasn't specifically x86. For example, games ported from x86 or Virtual PC applications are unlikely to get significant PowerPC optimisations if the strategy being pursued is an eventual migration entirely away from the minority architecture.
The one facet of current Mac OS X that will be lost with the switch is support for Classic, the 'Blue Box" compatibility layer that allows OS 9 applications to function under OS X.
Future
It seems as of this writing that the transition is fully away from PowerPC although internal builds of these may proceed as have Intel builds for the last five years. Apple has shown that it likes to have processor options and will act on them if the business case seems appropriate. Within the same department, builds should also include those for AMD, as well as SPARC and even Cell.
AMD was the obvious and in many opinions preferable choice for Apple. There are good reasons for even the most ardent detractor to grudgingly admit the benefits of x86 but those same ardent individuals are unlikely to accept Intel at all. AMD has the performance crown and with new production facilities can meet Apple's volume requirements. Moreover since they do not supply the entire platform, there would be less anxiety among the Apple customer base about the loss of identity. AMD were selected by Sun Microsystems to power their recent workstation range and Sun's technical demographic is not dissimilar to Apple's.
However, Sun is not a laptop vendor and AMD does not appear to offer laptop specification chips of the same calibre as Intel. They are also still bound to IBM's apparently failing process technology. Apple may pursue a dual-supplier strategy with AMD in the future to enable those Wintel detractors to once again adopt the platform.
Cell does sport a general-purpose PowerPC core as well as the multiple SIMD cores and while is not intended as a workstation chip has been recently demonstrated powering a Linux environment. This required significant effort to establish however. Finally Itanium has not been a market success and would arguably not improve the Xserve line's penetration if adopted.
Not counting Itanium, the roadmap for Intel in the immediate future suggested by Ars Technica includes "Yonah" (a dual-core Pentium-M successor with a 65nm process), "Sossaman" (a desktop version of Yonah) and "Conroe" (another desktop Pentium-M successor, and 64-bit). None of these processors have been confirmed for Apple at this stage. It is likely that laptop and lower-end machines will be revised first as they are most in need of refresh in comparison to the market as a whole. High-end machines will be revised last as the current G5 is still very competitive and an Intel alternative would not appear to add a great deal of performance.
Less realistically, it is possible that Apple is using this hardware change in order to rekindle previous opportunities to license the operating system. Other than Apple no major vendor sold a desktop PowerPC based system. Now it is possible that Apple will select certain hardware suppliers to increase the roll-out of OS X without simply unbundling the operating system and rekindling the "clone wars." As an existing hardware partner for the iPod, Hewlett Packard seems a plausible choice.
Alternatively Apple could license Sun Microsystems to bring the OS to AMD processors. Apple already has a deep relationship with Sun through Java. Apple has licensed this and related technologies for some time and Sun have incorporated certain desktop Java recommendations from Apple into the source base. Most recently Jonathan Schwartz of Sun has even dared suggest OpenSolaris as an alternative kernel for Mac OS X instead of XNU, a move that for many long-time operating system watchers, would bring together the best of all worlds.