Apple’s Clock is Losing Time
The recent nose-to-nose horse race between Intel and AMD to push the clock rate of x86 processors to 1 GHz stands in striking contrast to the dismal situation Apple Computer finds itself in with the snail-like pace of innovation of the PowerPC processors used in its Macintosh computers. Motorola and IBM’s PowerPC RISC processor line barely reach half the clock rate of the fastest x86 processors, despite the use of semiconductor process technologies that are in some ways more advanced than those employed by AMD and Intel. Although clock rate is not equivalent to performance the two are closely correlated and industry standard CPU benchmarks show a large performance gap has opened between Wintel and Macintosh computers. More importantly, the majority of the computer buying public perceives clock rate as the primary indicator of performance. This dire situation for Apple arose mainly due to the greatly increased competition between AMD and Intel that began last year after the K7 Athlon was introduced.
In many ways Apple is the author of its own misfortune. Years of work and billions of dollars of investments are required to design, manufacture and maintain the competitiveness of a family of microprocessors for the desktop computer market. Time and time again Apple has changed business strategies abruptly, only to reverse itself again a short while later in ineffective attempts to stem its gradual but consistent losses in market share. The PowerPC semiconductor partners, Motorola in particular, has written off hundreds of millions of dollars in losses caused directly by the erratic actions of Apple Computer, such as encouraging and later crushing a nascent market for Macintosh clones. The mercurial nature of its primary customer, combined with its minuscule and generally diminishing share of the desktop computer market, have meant that at least the last two generations of PowerPC processors have been designed primarily with embedded control, and more recently, digital signal processing applications in mind. This has left Apple in the position of only being able to differentiate itself on the basis of curved system form factors and translucent plastic. But it wasn’t always this way.
The Day Apple Took a RISC
Six years ago, Apple Computer brought the exotic world of RISC-based computing to the masses with the introduction of the Power Macintosh 6100, 7100, and 8100. Based around the 32 bit PowerPC 601 processor, clocked at 60, 66, and 80 MHz respectively, these systems were up to 50% cheaper than the least expensive RISC based workstations of the time, such as the Sun SPARC Classic, and provided compatibility with a large software base of personal productivity applications for the original 68K/680×0 based Macintosh family. The original Power Macs also compared favorably in price and performance to IBM compatible PCs, the fastest of which were based on the 90 MHz Pentium. By the end of 1994 PowerPC had taken the lead in RISC processor-based computer sales away from Sun Microsystem’s SPARC family.
The nearly seamless transition of the Macintosh architecture from a CISC-based processor family to a RISC-based processor was an impressive software achievement. It was helped by the fact that the 680X0 family had fallen sharply off the performance curve relative to Intel’s x86 line, and the much faster Power PC 601 could run original Macintosh applications under emulation with no apparent loss of performance to the user. As native Power Mac applications became available users experienced a large performance increase, which brought them to parity, if not ahead of, the best x86 based PCs, especially on floating point intensive applications..
The PowerPC family evolved under the AIM alliance (Apple, IBM, and Motorola) from the original IBM hybrid Power/PowerPC 601 design to new processor cores, such as the 603 and 604 originating from the Somerset design center in Austin, Texas jointly operated by IBM and Motorola. The high point of the PowerPC and Power Macintosh was the introduction of the 4- way-issue superscalar PowerPC 604. Even Microprocessor Report analyst Linley Gwennap, usually an irrepressible Intel booster, proclaimed the 604 with the front page headline "PPC 604 Powers Past Pentium, PowerPC Chip Will Open Performance Gap, Possibly Permanently". Somerset engineers projected that the 100 MHz 604 would achieve 160 SPECint92 which was more than 50% higher than an equivalently clocked Pentium. The 604 fell embarrassingly short of 160 SPECint92 at 100 MHz but still comfortably outran Pentiums, especially on floating point intensive code.
Discuss (18 comments)