Daniel Eran Dilger
Random header image... Refresh for more!

Leopard and the History and Future of Mac OS X on PowerPC

PowerPC Future
Daniel Eran Dilger

How long will Apple continue to support existing models of Macs in the latest version of Mac OS X? Previous versions of Apple’s OS have drawn the line for officially supported Macs based on practical considerations, rather than just being arbitrary or artificial. Here’s what the past suggests for Mac OS X Leopard and the version that comes after it.

The Post-Copland Crisis.
Apple carried along official support for the 1986 Mac Plus through Mac System 7.5.5 in 1996. That established an expectation for Mac users that any new Mac System Software would be able to run across a decade long generation of old hardware. Further, Apple had only begun officially selling System 7 as a retail product a few years earlier; many Mac users continued to think of the Mac operating system as something that was available for free, as it had been in the past.

That unreasonable support expectation combined with the sense of entitlement held by Mac users had helped to complicate Apple’s mid-90s failure to deliver Copland as a successor to System 7 between 1993 and 1995, and would continue to dog the company in its plans to provide a significant system software update after Copland was mothballed.

Faced with the task of maintaining full backward compatibility for both existing applications and a wide range of hardware–but without any assurance that a significant number of Mac users would actually pay for the upgrade–it’s no wonder why Apple was stuck at System 7 for over a decade (Mac OS 8 and Mac OS 9 were only retoolings of the System 7 operating system released in 1991), and why plans to completely overhaul System 7 with Copland and Gershwin failed.

If Apple had the luxury of operating outside of a real market economy and could simply rely on guaranteed future sales at high retail prices, it could have plowed along for twice as long and eventually released something, as Microsoft did a decade later with Windows Vista. As Windows Enthusiasts like to point out, Vista will eventually get deployed no matter how bad it is.

Vista Copland

The Secrets of Pink, Taligent and Copland]
Has Leopard Fallen into a Copland-Vista Conundrum?]
SCO, Linux, and Microsoft in the History of OS: 1990s]

Spindler’s Complications.
By 1990, Apple CEO John Sculley had recognized that Apple needed to set a reasonable minimum hardware threshold for its operating system releases and to figure out a way to get Mac users to fund the expensive operating system development the company was doing. At that time, both Microsoft and IBM were charging PC users around $100 for retail copies of Windows or OS/2, neither of which were even really usable.

Figuring out how to actually accomplish those goals never got done at Apple. Instead, Sculley’s successor Michael Spindler attempted to imitate Sony by releasing ranges of Mac hardware under a variety of vaguely Latin sounding names–Quadra, Centris, and Performa–and a series of confusing, nondescript model numbers.

Starting in early 1994, Apple also underwent a complex transition from its original 680×0 Macs to PowerPC hardware. Since much of the original Mac software was written in assembly language, the transition relied on emulation of the existing Mac System Software, which further complicated efforts to deliver significant new features without breaking existing software or prematurely cutting off support for existing machines. Non-PowerPC Macs continued to be sold into 1996.

Spindler’s Apple also began plans to license the Mac software to other hardware makers in late 1994, including APS, Bandai, DayStar, Motorola, Pioneer, Power Computing, Radius, and UMAX.

That effort skimmed off the cream of Apple’s profitability and handed it to the cloners, leaving Apple to service the low end of the market at Sears with its Performas while also funding the development of nearly profitless Mac System Software to support an increasingly wide range of hardware.

why apple failed

[Why Apple Failed]

Simplifying the Mac Hardware Lineup Around the G3.
When Steve Jobs returned to Apple in 1997, the company’s product line was all over the place, although efforts were already underway to simplify things. Apple had only just discontinued the last of its 680×0-based Macs a year earlier.

Under Spindler replacement Gil Ameilo, Apple had also scraped together a “Unity” release of System 7, newly rebranded as Mac OS 7.6. That release officially extended support back to all “32-bit clean” Macs, which included the eight year old Mac IIci from 1989.

The installed base of Mac hardware not only spanned across two hardware platforms–680×0 and Power PC–but nearly each individual Mac model from Apple had also used its own highly customized and often uniquely quirky hardware design. The cloners were also introducing subtle differences in their own machines, too.

Despite using the very modern PowerPC processors and Open Firmware, Macs in 1997 still incorporated old Mac ROMs to maintain software compatibility with the existing Mac OS. After taking control of Apple in the middle of that year, Jobs announced the release of a highly simplified product line using the new G3 processor.

The G3 was such a significant leap over earlier PowerPC processors that even the entry level G3s were faster than the top of the line models Apple had been selling. So while Apple had a confusing array of eight different major PowerMac models at the beginning of 1997, by the end of the year it only had two: a desktop G3 and a tower G3. It also shipped a G3 PowerBook.

[How CPR Saved Apple]

Mac OS X 10.0 – 10.2: G3 Only.
The introduction of the G3 processor created a clean line between it and the wide array of odd PowerPC hardware designed prior to 1997. The G3 also signaled the end of the line for the various models built by Mac cloners, who all refused to license new versions of the Mac OS at terms Jobs deemed reasonable. The G3 was also the first PowerPC processor optimized to run Mac software. That made it an easy minimum target for Mac OS X, which remained in development through 2001.

In the meantime, Jobs bought out Power Computing–the largest Mac cloner–for $100 million, and terminated other clone agreements by releasing Mac OS 7.7 as “Mac OS 8” in mid 1997. That revision also became the first edition of the Mac OS to really be successfully sold at retail; over 1.2 million copies were sold within the first two weeks.

The next fall in 1998, Apple released Mac OS 8.5, which was the first version to be PowerPC-only, limiting support to Macs sold over the last five years. In 1999, Apple shipped Mac OS 9. The new Apple had proved it could plan, ship, and sell regular releases of an operation system. The next task would be shipping Mac OS X as a major new leap past the classic System 7.

Between 2001 and 2002, the 10.0 to 10.2 versions of Mac OS X limited support to the G3 desktop Macs, including those first introduced in late 1997. It did not support the original PowerBook G3 unveiled alongside the G3 desktops however. The first supported PowerBook was the “WallStreet” revision introduced in May of 1998.

That maintained the roughly five year support window for machines to be updated with new versions of Mac operating system software.

Vista, Leopard, iPhone

Apple Sells 1.2 Million Copies of Mac OS 8 – Apple]
Leopard, Vista and the iPhone OS X Architecture]

Mac OS X 10.3: New World Macs Only.
After moving its hardware line to the G3, Apple next delivered a revised “New World” platform which modernized the Mac’s hardware and removed its old hardware ROMs, replacing them with “ROM in RAM” software loaded from disk.

The first New World model was the first iMac in 1998. The beige G3 Macs from 1997 were replaced with a single new “blue and white” G3 in early 1999, which used the same translucent plastic as the iMac. Apple shipped its first “New World” laptop in the Lombard PowerBook G3, distinguished by its translucent bronze keyboard. In July 1999, Apple released the iBook.

The release of Mac OS X Panther 10.3 in 2003 extended support back to Macs with G3 processors and built-in support for USB. This wasn’t due to an actual requirement for USB, but rather a shorthand way to describe a cutoff for the support of the significantly different architecture of “Old World” Macs designed prior to the iMac, as all New World Macs also provided support for USB. Panther retained a roughly five year support window for existing Mac models.

Mac OS X 10.4: Modern New World Macs Only.
In 2004, Mac OS X Tiger 10.4 retained support for most New World Macs using G3 processors, but required support for built-in FireWire. Again, this wasn’t related to a need for FireWire ports, but rather a way to exclude support for the earliest of the now five year old New World Macs, which Apple decided would not run Tiger acceptably, including:

  • the original 1998 iMac.
  • the original 1999 iBook.
  • the 1999 “Lombard” PowerBook G3.

These five year old machines can still run Tiger using XPostFacto, a third party enabler designed to force Mac OS X to run on earlier systems. However, significant differences in their hardware–coupled with their limited performance–prevented Apple from officially supporting them.

In the case of the Lombard PowerBook, its DVD drive was never supported for movie playback under Mac OS X because the system did not have the power to decode DVD video in software; under Mac OS 9, it relied on a hardware decoder. Rather than holding up Mac OS X to develop custom support for the obsolete hardware decoder in the now half-decade old Lombard PowerBooks, Apple told its users to continue using the playback software it came with.

[XPostFacto: OS X for Legacy Macs – Other World Computing]

Mac OS X 10.5: 867 MHz Processor Required.
For Leopard, Apple is specifying an 867 MHz G4. That excludes support for the now functionally obsolete G3s, and draws a line down the middle of the 2001 “Quicksilver” PowerMac G4s, excluding support for the 2001 G4 Cube and the first three generations of the Titanium PowerBook G4 up to late 2002. This again maintains official support for five to six years of Mac models.

This break roughly corresponds to the arrival of the G4+, a revised version of the G4 with support for L3 cache and improvements to AltiVec. It is also near the line for supporting Quartz Extreme and the higher end Core Image, both of which are technologies used to delegate graphics work to the video card.

However, Core Image is not a requirement for using Leopard; such a requirement would exclude support for all G4 desktops and laptops prior to 2003.

Leopard Looms Large.
That indicates Apple is being fairly liberal in officially supporting older models in Leopard. The obvious reason for this is that Apple wants to sell Leopard to as many Mac users as possible, even more than it wants to use Leopard to sell new Macs.

Between 2001 and 2002, Apple sold just over 6 million Macs. From 2003 to the present, Apple has sold about 23 million Macs. Apple wants to target the broadest possible market for Leopard, so excluding support for older machines is done with some hesitation. By extending support back into 2001, Apple is selling to an audience of nearly 30 million versus 23 million.

At the same time however, the likelihood of selling retail copies of Leopard to users of older Macs begins to drop as six year old machines go out of service or are no longer viewed by their owners as needing brand new software. This spring, analysts estimated an installed base of around 22 million active Mac users, an increase of 6 million over their figures from 2005.

[Mac install base estimated at 22 million pre-Leopard – AppleInsider]
Market Share vs Installed Base: iPod vs Zune, Mac vs PC]

Is Leopard the Last Hurrah for Power PC Macs?
The reports of PowerPC’s obsolescence have been greatly exaggerated. Last year, the rumor was that Leopard would be released only for Intel Macs. This year, with Leopard looming on the horizon, the new rumor is that Mac OS X 10.6–possibly named Lynx or Cougar–will be Intel-only. However this is only uninformed speculation.

When this rumor came up earlier about Leopard, I posted the article, “Unraveling The PowerPC Obsolescence Myth.” It pointed out that Apple would not release an Intel-only Leopard for an audience of the roughly 3 million new Intel Macs sold in 2006 when it could reach an installed base of around 20 million Macs with a Universal Leopard.

It noted, “If Apple continues to sell new Macs at current rates, it will be 2008 before Intel Macs begin to outnumber PowerPCs, and that assumes that every year, 4 million old PowerPC Macs will be destroyed. There will be a significant proportion of PowerPC Macs still buying software well into 2010, and the market will accommodate them.”

[Unraveling The PowerPC Obsolescence Myth]

Why the Mac OS X Backward Compatibility Window May Increase.
Apple’s Mac OS support troubles back in 1996 related to the support of multiple platforms, a wide variety of different models, and an inability to effectively market the Mac OS. Those issues are no longer factors today. Despite Apple’s maintenance of dual platforms since the transition to Intel began in 2006, technology has erased the barrier as a real problem.

The majority of the installed base of around 22 million Macs is PowerPC; less than 10 million are Intel Macs. Apple has started to sell dramatically more new Macs at a faster rate over the last couple years–displacing the PowerPC majority more rapidly–but there will still be a lot of PowerPC Macs well into 2010. Worrying about 10.6 or even 10.7 being Intel-only shouldn’t be among anyone’s greatest concerns.

By 2009, the likely ballpark release date of Leopard’s successor, the trailing end of officially supported Macs would include over 8 million PowerPC Macs sold since 2004, even more machines–and more recent models–than Apple is targeting now by reaching back into 2001 to support QuickSilver G4s in Leopard.

Further, supporting machines from 2003–including the first G5s–will be no difficult stretch, because the Mac architecture didn’t change dramatically between 2003 and 2005 in the way that it rapidly did between 1997 and 2000. In addition, Mac OS X hardware dependancies have been designed to degrade gracefully. For example, the acceleration framework and Core Graphics libraries make use of specialized hardware if available, or simply run on the general purpose CPU if it isn’t.

It’s also interesting to note that prior to 2000, Macs weren’t sold with Mac OS X because it didn’t yet exist. That means earlier versions of Mac OS X supported years of Macs that were never really designed to run it, while Leopard still supports the vast majority of the machines anyone ever bought with the expectation to actually use Mac OS X. New generations of Mac OS X will have fewer reasons to exclude support for existing hardware, leaving the support line tied to practical performance.

[Why Apple hasn’t used Intel processors before]

Intel-Only Not Necessary.
Around 11 million Macs were sold between 2003 and 2005, and all of them were PowerPC. It would be foolish for Apple to simply exclude that audience in the next revision of Mac OS X without good reason. As it works out, there really isn’t any good reason for Apple to ditch PowerPC.

Apple’s Universal Binaries architecture makes it relatively easy to maintain support across multiple platforms. It’s not like the move from Motorola 680×0 classic Macs to PowerPC, where old 680×0 software was emulated at significant cost on PowerPC, and new PowerPC code couldn’t run at all on 680×0 Macs. That situation left developers to wonder which they should invest their support in and for how long. Universal Binaries means there isn’t any tough choice to make.

Universal Binaries not only support PowerPC and Intel, but also make supporting 32 and 64 bit architectures easy. Leopard supports all four Mac platforms in the same software release:

  • 32 bit PowerPC G4
  • 64 bit PowerPC G5
  • 32 bit Intel
  • 64 bit Intel

Microsoft faces big problems in migrating its users to 64 bits, because it has no seamless architecture to waltz its 32 bit Intel users onto 64 bit hardware. Instead, Windows users have to obtain a separate 64 bit edition of their operating system, new 64 bit drivers, and new 64 bit applications. Supporting both is problematic, and deploying software across both is also trouble. Even Microsoft hasn’t delivered its portfolio of applications for its 64 bit versions of Windows.

Microsoft faces enough troubles selling Vista, let alone its deferred plan to deal with 64 bits and EFI at some point in the future. Apple already has both issues covered, allowing it to concentrate on more interesting tasks.


[How Apple’s Firmware Leapfrogs BIOS PCs]

Applications that are Intel-Only.
For Apple and third party developers using Apple’s Xcode tools, supporting both Intel and PowerPC architectures is really no more difficult than supporting just Intel Macs. In fact, Apple has also ported Mac OS X to the ARM architecture for use in the iPhone and the iPod Touch, demonstrating that it can flex its multi-platform muscle in several directions, not just as one-time, disposable transition plan. Universal Binaries isn’t a crutch, its a powerful deployment technology.

There are only three types of developers that will have any reason to deliver Intel-only Mac apps:

  • Companies like Adobe, which base their applications on their own custom, internal cross platform architecture. Since Adobe maintains its own system that is based on Intel-centric development, its new apps such as Soundbooth aren’t ever going to appear for PowerPC. If it used Xcode’s Universal Binaries, this would not be a problem. Xcode doesn’t target Windows though (at least not in a way Adobe can use!), so Adobe rolled its own system.

  • Software designed for Windows and ported to Intel Macs using a WINE-like engine. This is how EA is porting its new games to the Mac. They are actually Windows games running on a thin portability layer that emulates the Windows APIs. Since games don’t integrate into the desktop UI, a full Mac port isn’t very valuable for users or worth doing for the developer, particularly since the Mac gaming market is still pretty small. Porting over Windows games is far faster and keeps new releases in sync so that Mac gamers will have access to new titles sooner, and won’t miss features such as network play. 

  • Environment emulators and other software tied directly to the x86 architecture, including Parallels. These can’t be ported to PowerPC for the same reason that it makes no sense to port Virtual PC to Intel. Apart from running Windows–which is tightly bound to the 32-bit x86 architecture–there is really very little software that needs to run on a specific processor.

For most other software, including the vast majority of what makes up Mac OS X, it really isn’t difficult to deliver both PowerPC and Intel versions, so as long as there are PowerPC Macs around, there’ll be PowerPC software. That makes it extremely unlikely that Apple would drop support for PowerPC in the next generation of Mac OS X after Leopard.

Maintaining and Growing the Mac Installed Base.
What about the argument that Apple would prefer to “force” users to buy a new Mac to get the latest system rather than simply upgrade their existing hardware? Consider that Apple’s Mac profit margins are around 20% or less, while its Mac OS X margins are closer to Microsoft’s 80% Windows margins.


[Office Wars 2 – Microsoft’s Outrageous Office Profits]

Mac users paying to upgrade to Mac OS X are likely to buy a new Mac eventually as a replacement, so Apple’s delaying that hardware sale for a profitable software upgrade makes more sense than forcing existing Mac users to go out and buy new hardware, which might involve comparing a new Mac against a PC running Windows.

The more Macs that can run the most recent version of Mac OS X, the more attractive the target is for third party developers. Apple wants to maintain the majority of Mac users on the latest version of its software. In contrast, Windows Vista is competing against Windows XP, and the fact that Microsoft only earns 20% of its revenues from (the much more expensive) retail box sales indicates that most PC users upgrade when buying a new PC. More Mac users pay to upgrade their software.

That fact contributes toward making the Mac platform far more valuable than Windows; despite having only 3% market share of the entire world’s production of PCs, Apple makes more money on hardware sales than Dell with its 15% share of the market, and–after including Microsoft’s tremendous losses from its non-monopoly businesses–made half as much money in software as Microsoft did with its 98% share. Windows Enthusiast prefer not to think about this.

Even stripping Apple of its iPod revenues, which PC pundits love to do, the company still earned $4.4 billion on its Macintosh business last year, over a third as much Microsoft brought in from its entire Windows, Office, and server operations combined. Apple’s 2% of the PC market doesn’t seem so small anymore.


Can Apple Take Microsoft in the Battle for the Desktop?]
Market Share vs Installed Base: iPod vs Zune, Mac vs PC]

What do you think? I really like to hear from readers. Comment in the Forum or email me with your ideas.

Like reading RoughlyDrafted? Share articles with your friends, link from your blog, and subscribe to my podcast! Submit to Reddit or Slashdot, or consider making a small donation supporting this site. Thanks!

Technorati Tags: , , ,