The PowerPC Triumph

Credits: Insanely Great: The Life and Times of Macintosh, Upgrading and Repairing Your Mac, The Mac Bathroom Reader

There are sections in this site where I describe the period in Apple's life where they sold super-fast and super-expensive machines such as the Mac II and Quadra series. This was around the late 80s and early 90s, a time where Apple's stock was at an all time high, just like the criticism. Several people criticized the high price tags of the models Apple was developing, to which a former Apple VP replied "We make Hondas, not Yugos". Though the quote became popular among Mac enthusiasts, projects at Apple also started creating more reasonably priced computers that were still very innovative, like the Mac LC. Even so, it became clear that Apple was traveling down a crowded path and again needed to create a new breakthrough technology to secure its position as a leader in innovative products. Like the Apple II and Macintosh projects before it, it would be one man's vision that would make this happen for Apple. Jack McHenry was a veteran Apple engineer who worked on such influential products like the first Mac hard drive, and was the father of Apple's the Mac IIfx, Apple's most powerful machine at the time.

The end is near for the 68k

The Mac IIfx: a symbol of bloated speed
The Mac IIfx was the epitome of blazing speed, and had a ridiculous price tag to go with it. Not since the Lisa had their been such an expensive speed demon. But despite its seeming immortality, the IIfx was easily surpassed by even faster, bulkier, and more expensive machines that went from the Motorola 68020 processor, to the 68030, ultimately to the 68040 (called the 68k family). Essentially, Apple was doing what Intel is doing now, they were trying to sqeeze as much processing power into old technology as they possibly could. Fortunately, there were those at Apple that saw that the 68k family of microprocessors would be an eventual dead end. A fact that did seemingly did not occur to the Wintel community until much later. Anyway, also at about this time there was a new revolution in microprossessor technology called RISC, which stands for Reduced Intstruction Set Computing. It was a radical new architecture for processors that were designed to run much fewer instructions per cycle, thereby drastically increasing the speed of the processor. The 68k family as well as the entire Intel family were based on CISC, or Complex Instruction Set Computing. This is the old way of thinking; using large instruction sets for the processor to crunch. What does this mean? Well basically, when RISC and CISC processors are tested side by side, RISC processors are consistently found to be much faster at performing the same tasks as CISC processors. Though exotic, unusual, and unlike anything previously thought up in the industry, RISC seemed like the future of computing.

This was a fact not lost by Apple, already seeking a way to create even faster Macs. The "Jaguar" project was created to develop a new type of RISC-based machine that would use this technology to boost Apple ahead of its competition. The only problem the project saw was compatiblity. The one problem with switching to any kind of processor is that all software written for that platform suddenly becomes obsolete. You can't run software for a processor that it is not designed for. This meant that the new RISC-based Apple machine would be totally incompatible with current Mac software. Like the Lisa before it, the machine would be useless without anything to run on it. Of course, several programmers gladly lent their expertise to the project. They were all awed by the power of RISC computing, and the things it would be capable of bringing to the market. By 1990, only a year after the introduction of the Mac IIfx, there were two to three dozen engineers working on projects for Jaguar.

RISCy Business: The fate of 68k software

Enter again Jack McHenry. He didn't see the Jaguar project as the answer to Apple's needs. Backward compatibility was a must for a project that would succeed the 68k Mac. How can 6 years of software development to a company whose software base is already miniscule be scrapped? It is why on a March, 1990 ski trip to Reno, McHenry and a group of fellow engineers decided they would embark on a project that would implement RISC technology but still retain 68k compatibility. The project was later name "Cognac", after a pioneering engineer of the RISC processor named Jack Hennessy who had a surname similar to the liqueur. The group included colleagues that worked on the IIfx project, liked Bob Hollyer and Jon Finch. Again similar to the Mac/Lisa development, there was much competition between the two projects. Cognac beleived Jaguar was doomed to failure, for their philosophyt was flawed. Though some applications would inevitably run on the Jaguar machine, incompatibility with 68k software made the large amount of smaller commercial, shareware, and freeware programs unusable to the new system. Surely enough, and luckily for Apple, the Jaguar project would never release a machine.

In the meantime, the Cognac team had to figure out a way to incorporate 68k compatibility in their machine. They had two logical options, each with their pros and cons. Unfortunately, both of them were mostly cons. One choice was two include two processors on the motherboard. The dual-processor idea would guarantee that software written for both chips were run with optimum speed . The problem with it was that dual-processor machines had never been successful in the market up to this point. Consumers were not willing to pay that much money for a computer, despite its features. It was a novel idea, but Apple could not justify yet another costly beast to roll of their production line. The second option was emulation. You can emulate hardware, in which you use the processor to trick the machine into thinking it is running hardware that it is not. They could use the super-charged RISC chip to run 68k instruction sets using a built-in emulator. The problem with this method is the fact that you greatly sacrifice speed. If you have ever used an emulator that mimics another processor (like SoftWindows, or VirtualPC) then you know that emulation of this sort can be a headache. It seemed this would be another dead end. At least until later that year. This is when the team figured out a way to create a much faster emulator exploiting what is called the 90/10 rule. The 90/10 rule is just a principle that states that when you write a large computer program, like a word processor, you will be running only 10% of the software's code 90% of the time. Using this principle, the Cognac team could create an emulator that would exploit this principle to create a much faster emulator. All they needed now was a RISC processor and a machine to put it in, and they could start working on their theory.

The man that would make this happen was Gary Davidian, an expert in the field in programming microprocessors and ROMs. He wrote the 68k emulator for every RISC chip Cognac was using currently, which changed often. When Cognac used a new chip, Gary wrote a new amulator. It was an unimaginable feat making not just one but several processors mimic the instruction set of the 68k processor family. Both RISC projects worked on their projects to reach the first major step in developing their new machine, point is called "gray screen". Thi is when for the first time an image appears on the monitor of the machine, usually a checkerboard pattern of white an black pixels, hence the name. Cognac worked at a smooth pace, and in late 1990 they achieved gray screen status. Finally, in 1991 Apple decided on a RISC chip. Despite consistent deadline problems Apple and Motorola had maintained a good working relationship until now, as Apple had used Motorola processors to power their machines for roughly a decade. So it was no surprise that Apple decided to use Motorola's RISC chip. An excellent proposition to Motorola it seemed since even though the company would use their RISC chips in their radios, pagers, and so on, recruitment by Apple would increase their business beyond their wildest dreams. Gary rewrote his emulator using the tricks he had learned along the way. Meanwhile, the Jaguar team changed their codename to "Tesseract", but had no new progress to show off with their new name.

RLC: The 1st RISC-based Mac
 
The Cognac team was making great progress, just a couple months after achieving gray screen fo rthe first time they created their first working model of a 68k compatible RISC-based Mac. It was dubbed the RLC because it was a Mac LC case with a RISC processor. The machine was incredible, it ran 68k Mac software at speeds that rivaled the Mac II. But the speed of the 68k emulator wasn't the most impressive feature, since it would be much tweaked in the time to come. The thing that left Apple engineers in awe, especially Tesseract project members, was how compatbiel the chip was with old software. It literally ran everything. Project members ran everything from current software to programs that had been obsolete for years. The RLC kept running 68k code as happy as can be. It was at this time that the Cognac team proved that such a project could be done. Until now they hadno proof that a 68k-compatible RISC machine would be a logical buy for the consumer. If either speed or compatiblity of crrent software was compromised, the project would be a failure. This is why the RLC was such an important event in the development of the RISC-based Mac. Then of course there was the unparalleled power of the RISC chip itself. Though no software was yet written for the new chip, several graphics and other benchmark routines were run on the machine with incredible, unparalelled, unprecendented speed. Cognac was able to develop more prototypes in confidence of their success.

Apple and IBM...allies?!?

If you ask an avid Mac user about the origins of the PowerPC, they will usually start with what happens next. At the same time Cognac was working on the RLC, Apple had been collaborating on another revolutionary project. On April 12, 1991 Jack Sculley, then CEO of Apple, gave a surprising demonstration. IBM engineers presented an IBM PS/2 Model 70 running a prototype OS that looked remarkably like the newly released System 7. Now you must remember that at this time IBM was still the archnemesis, the "dark side" to Mac users. With this in mind imagine the site of an IBM computer running an Intel-savvy version of the Mac OS running faster on a 486 than a 68040. This was a sign of things to come. The OS that was running on the PS/2 was a protype of "Pink", an object-oriented version of the Mac OS that would run much faster and more efficient than System 7. IBM was interested in Apple's RISC project and on July 3 sent a letter of intent saying that it would help create Pink if it will use their yet-to-be-developed RISC chip. After some negotiations, Apple announced to a shocked computer industry of their future plans on October 2.

It was called the Deal of the Century. Among the many agreements, Apple and IBM will create RISC-based machines and produce two companies, Taligent and Kaleida. The latter company would work on Pink, which would run on the RISC processor, dubbed the PowerPC. Motorola would be work in conjunction with the PowerPC project and its distribution. Kaleida would develop multimedia tools. The main deal here was the PowerPC, as it was planned to be the centerpiece of both Mac and IBM computers. This was highly unusual, and many thought it was a vain attempt that would surely fail. "Whoever thinks PowerPC will work is smoking dope" commented one journalist. Frankly, you couldn't blame them. Two formerly rival companies that are exact opposites designing a new operating system for a processor that they will both use but that doesn't exist? It seemed downright ridiculous to Mac and IBM users alike. But that was politics to Apple and IBM engineers, who were more concerned with how they were going to work together to make this happen. Cognac now had to rewrite their emulator to conform to a vaporous processor. Moreover, they now have to collaborate with IBM, who had totally different development methods. The complexity of the project had increased one hundred fold. Execution would have to be near perfect if they wanted this to work.

Fortunately, near perfect it was. It was apparent that everybody wanted this to work. IBM officials often tried to accomodate Apple and vice versa so that both could comfortably work in unison. One popular anecdote talks of one of the first meetings between the two computer giants. IBM engineers dressed in blue jeans for the meeting, while Apple engineers dressed in suits. They both tried to make each other feel compfortable by conforming to what they thought the other was like. After that the Austin, Texas PowerPC design center, called Somerset, was a lot more relaxed. It was now the end of 1991, and by that November Apple and IBM had finished the details in the transaction. The first PowerPC processors were to be designed at Somerset and would arrive at Apple in one year and Apple would introduce the PowerPC-based Power PowerMacintosh series on the Mac's 10th anniversary, January 24, 1994. Skepticism was high after this announcement. Computer users know that deadlines in the computer industry are more like unlikely targets, and this was even more so with the chip industry. As I said before Morotola had several problems with making deadlines, with two other companies working together simultaneously, depending on each other to make everything work made it seem like an impossible target.

The PowerPC processor

 
I'd like to take some time out here to describe exactly what the first generation PowerPC processors were to include. The first and most important processor at the moment is the PowerPC 601. It is the basis of the first generation of Power Macintoshes and its design is taken from the IBM POWER architecture that was used in IBM's RS/6000 workstations. What makes the PowerPC so incredible is that it brought the power of the RS/6000 to desktop machines like the PowerMac. The PowerPC would later go on to power the RS/6000 series (including Deep Blue, the chess champion). It has three execution units, which are the parts of the processor which process the instruction sets. The units are the Integer Unit which performs mathematical calculations on integers, a Floating Point Unit (FPU) that does calculations on real numbers, and the Branch Processing Unit that controls the flow of the program the processor is running and gets the instructions from memory. The 601 also has 32k of cache, where it can hold frequently used data and instructions.

Another thing that makes the PowerPC cool is that it supports Superscalar processing. What the hell does that mean? Well, it can perform tasks using the three units described above simultaneously per cycle. It can do all three things at once. This is the reason why even though some PowerPC processors have a slower clock speed, it can actually run faster than those with a higher clock speed. The last ingenius feature of the PowerPC is how small, cheap, and energy efficient it is. PPCs are half the size of Pentiums, cost half as much, and use much, much less power.

Race to the finish: Tesseract, Piltdown, and the first Power Macs

PowerMac 6100/60, aka Piltdown Man
 
This is why I think the accomplishment of this feat is the greatest one in this decade for Apple, IBM, or Motorola. Because in less than one year, Paul Nixon and his team at Somerset designed, built, and shipped the PowerPC 601 to Apple on September 2, 1992. Cognac and Tesseract didn't just wait around for Somerset to mail them a chip either. They worked their butts off to design computers that would meet the requirements of the PowerPC 601 before the chip came. Cognac began work on a Mac that would serve as the bridge betweem the 68k Macs and Tesseract's high-end machine. Cognac's new Mac would be the "missing link" between 68k and RISC, incorporating features of both. This is why the project was renamed Piltdown Man (PDM) after the missing link hoax. Anyway, Apple finally received the 601 (in Christmas wrapping no less) and that very week PDM had gray screen, ROM boot up, and Mac software running of the machine. The next month, October 3, Apple had a working PowerPC-based Macintosh was running current software and most importantly, the Finder.

PowerMac 7100/66, aka Cold Fusion
 
Tesseract was a different story. The troubles that Tesseract had before the Apple/IBM alliance worsened afterwards. Tesseract was still the major RISC project at Apple, and PDM was still a side project. Apple began to worry that Tesseract would not meet their deadline. Sure enough that December, months after the working PDM prototype, Tesseract informed Apple that it would not be able to meet its January 1994 deadline. By March of the following year, Tesseract was dead.

PowerMac 8100/80
 
Apple engineers were running in panic, they had no high-end version of their PowerPC machines, and they had less than a year before their deadline. All eyes now turned to McHenry, and his side-project turned into Apple's only hope for saviour. Luckily, the PDM team had designed their machine to be flexible. Probably knowing the shakiness of the Tesseract project, the motherboard was developed so that it could support more advanced versions of the PowerPC 601 so as to create the exact machine Tesseract was developing. PDM now had the to recreate the Tesseract project-with 68k compatibility! PDM members wasted no time, and using faster versions of the 601 developed by IBM they create two other mutated forms of PDM. The first was the original, using the Quadra 660av style casing and a 60 MHz PowerPC 601. Next came the the mid-range version, a slightly suped-up PDM in a new case. Keeping with the hoax tradition, it was called Cold Fusion and used a 66 MHz PowerPC 601 in a new case design. The high-end model, Carl Sagan (I won't get into the troubles caused with this naming scheme except that it almost cost Apple a lawsuit) used a blazing 80 MHz PowerPC 601 in the Quadra 840av-style case. With a new March 1994 deadline, McHenry and his team was done with the hardware. There was no time to spare though, since they now needed to work on the software.

This is where the PowerMac project ran into problems. PDM, Cold Fusion, and Carl Sagan all ran versions of the Mac OS and Mac software just fine. The problem was that it didn't run much in native code. The immense power of the PowerPC was being wasted on 68k instruction sets instead of blazing through native PowerPC instructions. It was in part due to prejudice towards either IBM or the PowerPC project all together. Mac heads didn't want to work on a new System that would eventually do away with the 68k. It turned out that a new System wasn't going to happen. Pink was way behind schedule, and would soon be cancelled. Taligent would soon close. Apple engineers found that Power Macintoshes should run the exact same thing as normal Macintoshes anyway. So that's what they did. System 7 was rewritten with some native code, but essentially was exactly the same as 68k versions of System 7. Even today Mac OS 8 still had some 68k code and is not completely "native", though the Finder and many components are. It won't be until the 1999 release of Sonata that the true power of the PPC is used. Anyway, engineers finally released the PowerMacs, now called the 6100/60 (PDM), 7100/66 (Cold Fusion), and 8100/80 (Carl Sagan, BHA, LAW). In March 14, 1994, right on schedule, the models were introduced at the Lincoln Center in New York City.

The future of the PowerPC

The 1.1 GHz PowerPC G3
 
In conclusion, the project went through many twists, turns, and create the birth of new allies. Even so, it accomplished its mission. The PowerMac again set Apple apart from the rest, while still maintaining perfect backward compatibility. Ever since 1994 Apple has been years ahead of Intel in the production of the fastest desktop computers available. The recent introduction of the PowerPC 740 and 750, the third generation of PowerPC processors, have blown the computer industry away. Apple computers are now so far ahead of their competition that they are up to twice as fast. Also, using new copper-wiring technology, IBM demo'ed a 1.1 GHz (1100 MHz) PowerPC G3 which will be in prodution by 1999. This is so far ahead of anything imaginable by Intel, Digital, or any other company that its laughable. The PowerPC also has made headlines in other circles. Deep Blue, IBM's chess playing computer (which is really an RS/6000) was made famous when it beat world champion Gary Kasparov. Of course, the processors inside Deep Blue were PowerPC 604s. Even more interesting is that PowerPCs are now extraterrestrial. The Carl Sagan Memorial Station (aka Pathfinder) uses several PowerPCs to carry out its mission on Mars. It is apparent that the PowerPC is the future of computing and will lead us into the 21st century.

 

 



Copyright ©1997-98 Andy F. Mesa. Reproduction of this information should be done for educational purposes with proper credit given where needed.

Apple, Macintosh, Mac OS, Quadra, and Power Macintosh are registered trademarks of Apple Computer, Inc.
PowerPC is a registered trademark of International Business Machines Corporation.