Why Apple hasn't used Intel processors before

When Steve Jobs announced the plan to move Macs to Intel processors, Bruce Chizen, CEO of Adobe, joked, “The only question I have, Steve, is what took you so long?”

Good question! Here's a look at the circumstances that kept Apple and Intel apart until today.

At Apple's founding, Steve Wozniac's processor of choice for the first Apple computer was the MOS 6502. The decision wasn't based on the chip's elegance or technology, but rather because it was cheap. The 6502 also powered the very successful Apple II (and the dismal failure of the Apple III), and outside of Apple, versions of it were used in the Atari 2600, Nintendo's NES, and the very successful Commodore 64, among many others.

As sales of the Apple II line flourished, the company looked ahead to a future platform with far greater power than something like the 6502 could provide. Apple picked Motorola's 16-bit 68000 processor for use in both the Lisa and Macintosh to give the machines unprecedented power for a small computer.

This processor was initially designed with a '32-bit ready' architecture, so while it shipped as 16-bit to save costs, it planned ahead for a future where 32-bit computing would be a reality. Motorola's 68000 was easy to program and offered a very clean memory space, unlike the oddly arcane, but cheap, 6502 Apple had been using. With Motorola's new processor, Apple was ready to change the world.

IBM had no dreams of changing the world when they introduced their own PC. Instead, they were interested in creating a product that would fit into the new, blossoming market for microcomputers, without eroding their existing business in minis and mainframe business machines. IBM's very different motivations were reflected in their choice of CPU.

Intel offered a 16-bit processor, the 8086, but IBM actually picked the 8-bit version, the Intel 8088. This was partly to position the machine at a lower price point, but also to effectively neuter the PC so it would not eat into IBM's other sales. Compared to Apple's futuristic Macintosh, IBM's PC was old technology striving to be sufficient.

For the next decade, Apple used successive generations of Motorola's 68000. In addition to Apple, the 68000 series was used in the Atari ST, Commodore Amiga, Sun workstations, and later the Sega Genesis and Palm Pilot. Its popularity came from its power and elegant design.

With the PC, IBM had established a de facto standard in business computing, but the industry that resulted around copying the PC was stuck with the limitations in Intel's x86 architecture. There was little significant use of the x86 outside of PC clones, mostly because it was an inefficient and antiquated design. Unlike Motorola's flat 32-bit memory addressing, the x86 platform had a clumsy, segmented memory map and other ugly and convoluted workarounds to maintain backwards compatibility with the original IBM PC.

It is important to note that, despite the 8088 not being revolutionary, the limitations of PC-compatibles were not wholly imposed by Intel's lack of innovation or capacity to design processors. The PC industry sprang from IBM's deliberately stunted design of the PC; beyond the cheap 8088 processor, there were a number of other critical flaws and limitations.
Intel was actually stuck with improving upon the 8088 to feed the PC market.

Intel also offered far more modern technology; one example is the i960, a RISC processor Intel pitched to NeXT for the original Cube. It was later used as a high performance graphics accelerator in the Cube's NeXT Dimension video card. But for Intel, x86 was where the money was.


Why Apple hasn't used Intel processors before: Part II
The PowerPC Promise

In order to keep parity with Motorola, who had started with a better design, and yet still maintain compatibility with the 8088, Intel had to design increasingly complex processors. This effort was worth it for Intel, because there was a booming demand for faster PCs, and the money generated from x86 sales far outweighed the development costs of trying to mitigate the flaws in earlier versions.

Intel started facing competition in developing successive versions of the x86. Both Cyrix and AMD began undermining Intel's monopoly of the PC processor industry by offering x86 compatible processors that rivaled Intel's. The x86 world was actually growing so fast that Intel faced more competition at home in developing x86 than they did in keeping pace with Motorola.

The fifth generations of the two processor families resulted in Intel's i586 Pentium and Motorola's 68060. While the Pentium found great reception in the PC world, few of Motorola's customers were interested in the '060. Some, including Apple, held out for the new RISC technology in Motorola's upcoming 88000.

PC sales funded increasing x86 research and development on a scale that began to overshadow the rest of the world, causing Apple to worry that Motorola would not be able to keep up in the technology race alone. Somewhat ironically, Apple decided to partner with IBM in designing a new family of processors, using some of the 88000 technology Motorola and Apple had started work on, and adding in a scaled-down version of IBM's 64-bit POWER architecture.

The resulting PowerPC processor caught the attention of the entire industry. Microsoft announced plans to port Windows NT to run on the new PowerPC, joining Apple's Mac OS and IBM's OS/2. NeXT, which had been manufacturing computers using the same 68040 processors as the Mac, also began efforts to port NeXTSTEP to the PowerPC. The BeBox also used the PowerPC in the development of an entirely new BeOS platform.

PowerPC was a clean, new design with lots of potential. Like Motorola's 68000 a decade earlier, PowerPC was a forward looking (64-bit) design scaled down to fit the present market (32-bits). By comparison, Intel's Pentium was a 32-bit design built on top of 8/16-bit predecessors, and carried a lot of legacy baggage. It required larger dies to manufacture, and it generated more heat.

Intel's next-generation P6 technology was initially released as the Pentium Pro.; it was an even larger processor than the Pentium, more expensive, and had disappointing performance for Windows 95's existing 16-bit software. The PowerPC family performed better on many levels. It matched Pentium performance while running at lower clock cycles and using far fewer transistors, allowing for higher manufacturing yields and therefore cheaper chips that ran cooler and more efficiently.

But Apple didn't just benefit from the PowerPC's initial lead; Apple needed PowerPC to in order to survive. System 7 was closely tied to the old 68000 architecture because much of it was originally written in low level assembly for performance reasons. Fortunately, 68000 code could run well in emulation on the PowerPC, so it was an easy, rewarding move for Apple to migrate from 1980 68000 CISC technology to the RISC PowerPC of the 1990's.

Attempting a move to Intel's processors would not only have been a far more difficult technical effort for Apple (because the Pentium was abysmal at emulating the 68000 fast enough), but it would have been a difficult business case to make as well. Intel's latest designs weren't even performing well in the PC world, and it looked like the x86 architecture was running out of steam.

Why Apple hasn't used Intel processors before: Part III
Two Roads, One Destiny

In 1993, Apple had a large installed hardware base and a firm transition plan to PowerPC. At the same time NeXT, while having made similar plans to move to PowerPC, abruptly decided to leave the hardware market. NeXT lacked both Apple's installed base and any rosy prospects for significant new hardware sales.

Abandoning both their existing 68040 based computers and their work already in progress on new PowerPC hardware, NeXT instead reinvented itself as a software company. They would now exclusively sell their NeXTSTEP operating system to run on the computers they had already sold, as well as a new version of the operating system written for standard PCs.

Unlike Apple, NeXT could move to Intel because NeXTSTEP's higher level frameworks were already processor agnostic. The core OS in NeXTSTEP was a combination of Mach and BSD, and both were already running on the x86. So NeXT had both the technical capacity to move their OS to Intel PCs, and a compelling reason to do so: NeXT's hardware sales were in trouble, and the company hoped that the technical superiority of NeXTSTEP would be an attractive alternative to PC users running Windows.

For the next four years, Apple and NeXT would follow very different paths before eventually merging. Both paths set in place events that would result in Apple current transition to Intel. NeXT's experience in creating a cross platform operating system, and discovery of how difficult it turned out to actually sell it, would later be of enormous benefit to Apple. Apple's own path of transitioning the Mac OS to PowerPC would provide complementary experience that, after merging with NeXT, would enable Apple for the first time to even consider a move to Intel.

This turned out to be critically important because as processor development became increasingly complex, maintaining a competitive, independent architecture for desktop computers next to the PC's x86 became increasingly difficult and impractical. As well, the value of being on different hardware eventually evaporated. Had Apple not been able to move to Intel a decade later, they would have few other options, for reasons that only become obvious later.

So, throughout Apple's existence, there has simply never been both a reason and opportunity to use Intel processors. Intel's earliest chips were inferior to what Apple and the rest of the 16/32 bit computing world was using, and just when PCs began to catch up, an entirely new PowerPC platform promising far more potential emerged. At no point in between could Apple have moved their operating system to new hardware while ensuring third party software would follow.

Examining the details of Apple's PowerPC transition is useful considering in how well Apple can be expected to handle the transition to Intel.