EDITORIAL

Rebel Alliance

Believing they needed a competitive edge to combat a common foe, a small band of large companies formed an alliance more than four years ago. The mission of the alliance was to create a new computer capable of running each vendor's software, while providing advantages over that of the "Wintel" (Windows running on Intel-based platforms) empire. The alliance decided that, if it could base its approach on a RISC architecture that offered price and performance advantages over the empire, it might have a fighting chance. And while Intel was protecting its chip's architecture through patents, the alliance agreed to create a common instruction set architecture (ISA), allowing other chip makers to create processors that will run the same code.

Four years later, both Apple and IBM have launched hardware platforms based on the Motorola/IBM PowerPC chip. Now, several operating systems are either in development or already on the market, including AIX, Solaris, Linux, OS/2, Windows NT, and Apple's Copland. Additionally, developer tools are beginning to show up in numbers. The big question now is whether corporate developers will embrace the new technology.

Developers must consider a dizzying array of options. First, there was the original PowerPC 601 microprocessor, manufactured by IBM but sold by IBM and Motorola. The 601 was designed to bridge the gap between PowerPC and the POWER chip used in IBM's RS/6000 workstations. Thus, it uses the older POWER instruction set, which has since been eliminated from the specification. The MPC603 and MPC603e included power-management functionality for notebook computers. The 64-bit MPC620 chip is still in development. However, features and options such as instruction and data caches all vary with the different implementations. For instance, the MPC602 (which is targeted at consumer electronics and embedded applications) has dual 4-KB instruction and data caches. The MPC604, on the other hand, contains dual 16-KB caches, while the MPC620 will have separate 32-KB instruction and data caches.

The alliance had originally intended to deliver on its performance promises by now. Although Motorola contends that its MPC604 chip outperforms the Pentium, the estimated 15-30 percent improvement in performance still falls short of the 2:1 increase originally promised. Moreover, Intel's upcoming P6 microprocessor is expected to show performance comparable to that of the MPC604 chip.

IBM finally began shipping its PowerPC-based machines in June of this year. However, the entry-level models running Windows NT cost consumers some $3700, well above the $2500 for similar Intel-based machines running the same operating system. Further, IBM announced nearly a year ago that it was delaying the launch of its PowerPC computers so that it could make ready its OS/2 for PowerPC. Apparently, the port of the operating system is taking longer than expected, and the company could wait no longer to deliver its Power Series computers.

Meanwhile, the first Linux kernel for PowerPC is up and running on Motorola's PowerPC VME 1603. But when asked about a similar port to PowerMac, project coordinator Joseph Brothers indicates that Apple cannot come up with the necessary programming specifications for the PowerMac's NuBus, nor can it provide necessary information on devices, memory maps, or interrupt hardware. Motorola has tried for more than a year to obtain the necessary specifications from Apple. (Incidentally, the Linux kernel is available via anonymous ftp from liber.stanford.edu/pub/linuxppc.)

Despite the short-term glitches in cost, performance, and roll-outs, the PowerPC is impressive, and most major operating systems will likely run on PowerPC-based platforms in the future.

Still, at this stage, the fate of the PowerPC alliance is in the hands of developers such as yourself. May the source be with you.

Michael Floyd

Executive Editor


Copyright © 1995, Dr. Dobb's Journal