Advocacy
Dojo (HowTo)
Reference
Markets
Museum
News
Other
Myths
Press
General
Hack
Hardware
Interface
Software
Standards
People
Forensics
Web
CodeNames
Easter Eggs
History
Innovation
Sightings
Opinion
Martial Arts
ITIL
Thought
|
IBM had this great RISC processor chip-set (the Power Processor) that was doing well in the workstation market. IBM (Processor Division) had a dream to make this THE processor to use in the entire industry. IBM before this was mostly supplying IBM, but they wanted to take the Power chips to the mainstream. IBM knew they couldn't do this by themselves -- they needed partners.
IBM also new that they needed to add value to OS/2 if it was going to survive, and on way to do that was put that OS on top of a cool chip. But if IBM did it alone, the rest of the industry would not likely follow. Against, they needed to partner with someone to make their Power chips successful -- and maybe save OS/2.
So they approached Apple about an "unholy" alliance. Remember, at the time IBM was seen as the dyeing enemy of the computer industry -- the Microsoft of the previous 30+ years. To shorten the story, Apple was hooked. They didn't like the single-source issues with the Motorola 68000's, but they did have a good working relationship with Motorola, and they didn't fully trust IBM. So Apple brought Motorola into the deal.
The Deal
Basically the deal came down as follows. IBM and Motorola would make a new Processor with a similar instruction set (design) to the Power architecture from IBM. It would use some of the I/O and interface of the 88000, and would be a single chip (instead of a chip set) which would reduce the cost. This chip would be called the PowerPC, and it would have an open design, so that others could implement PowerPC variants if they chose. Exponential did use it later (proof of the openness).
The 88000 was a completed design and so by using parts of it, they could get the chip to market much faster, and Apple already had working motherboards that could be used if the PowerPC used the same electrical interface. Apple helped with the chip design in a very limited way, advising on their interests, and Apple would be able to get chips from either company -- but IBM was going to use a large amount of their chip manufacturing capabilities for their company (IBM has always been the largest buyer of IBM chips).
All three companies went in as partners and built a whole research and design facility (Somerset) in Austin, Texas. Austin was chosen because that was an up-and-coming area, and Apple was already moving some operations there, and Motorola and IBM had facilities there as well.
The idea was that if all companies standardized on a single design, then they could create their own infrastructure, all supporting each other. They were competing, but also complimenting. The results have been a success and AIM lept to the lead in price-performance, performance/watt, performance/MHz, highest speed (MHz), and highest performance. Intel still wins in marketshare -- but for how long?
Intel Responds
To get designers hooked on the new Processor, AIM had to sell them on how superior the PowerPC was (and was going to be). They stated their goals -- to make a processor family that was twice the price/performance as anything Intel had at the time. This meant have a processor that twice as fast on the high end, and to also have another processor that performed the same as the highest intel machine - at half the cost.
AIM made a mistake in that they telegraphed their punches -- they announced, in detail, when products would come to market, and how fast they would be. Intel was not going to sit on its butt and get eaten alive. So Intel poured money (literally) into design and implementation -- they were running scared (for good reason). When AIM came out with the 604 and was ready to double the performance of the Pentiums -- Intel pulled a surprise and released the PentiumPro years ahead of original goals. It was a big, hot, power-eating chip -- but it was fast. Almost as fast as the 604e's, but it was still twice the cost. Intel could market around the rest.
AIM has been able to keep better performance than Intel for the past 4 years, but seldom achieving the 100% faster part of their goal. AIM has been able beat Intel at half the cost (for their respective chips). Furthermore, Intel only achieved their parity by pouring literally 10 times the money into each processors design as AIM has. Intels' strategy will only work for a limited time before the Pentiums design complexity collapses under its own archaic weight. In fact, with the various bugs in the Pentium, PentiumPro's and PentiumII's, their processors are hovering around collapse even as we speak. But Intel only needs to mire the industry in their desceptive glow long enough to come out with a real design that takes them into the RISC camp. This is expected around the turn of the millennium with the Merced. Intel can't really speed THAT up because part of that delay has to do with manufacturing technology being able to produce a chip that has enough switches to be cost effective and is cheap enough to produce to meet their design goals. So all the Pentium Processors are just a way for Intel to stall until they can make the jump to RISC and maybe stand a chance of catching up to the AIM camp.
Intel has been able to market. They convinced people that the PentiumPro was as fast (or faster) than the 604e. It isn't faster -- it is "almost" as fast as the MacOS, if you run Unix on Intels' special Hardware (just for performance benchmarks) using Intels' compiler (that is designed just for benchmarking). Most people run WinNT (which is slower than Unix), running many Win31 apps (which much slower), using regular development tools and compilers (slower again). The results are than in many real world conditions the PowerPC's are sometimes 2 or 3 times as fast as either the Pentiums or PentiumPro's. Sometimes the Pentium itself is faster than the PentiumPro. But marketing (lies) win out over the truth in some cases -- and many are happy in their ignorance.
MMX is another response (stalling effort) to the PPC's superiority. In a certain classes of functions, the PowerPC is just blowing the doors off Intel Processors. These functions are NSP/DSP (Native Signal Processing or Digital Signal Processing) and can be very useful for multimedia. AIM was literally out-performing Intel by 2:1 or 3:1 or more. Intel first responded by trying to convince people that these functions weren't important. For the first few years of the Pentium, Intel was saying that most people don't need that toy "multi-media" stuff. Then Intel came out with "MMX". MMX is a specialty set of instructions, that if Software writers rewrite their applications to use MMX (4 or 5 of whom have actually done so) then they can get performance just a little slower than the current PowerPC's. Now that's not how it is marketed. Intel markets it with dancing engineers injecting "fun" into the processors, because that's far more interesting than the truth. The 4 or 5 apps that actually use MMX do show a marked increase over the older Pentiums, but still have not even caught up to the current PPC's at the same clock rates -- and the PowerPC's see their performance across ALL applications (while it is only a very few Apps that see the advantages of MMX). PPC's are also available at much higher clock rates -- and the PPC's are getting faster still, while the Pentium families performance growth has stalled. In fact the PentiumII is slower than the PentiumPro which it is supposed to replace. So while MMX is a marketing success, it is a technological failure.
The Markets
The funny thing about this industry is that most analysts aren't looking at the entire market. Many processor companies make their real money on embedded controllers -- not just mainstream processors. Imagine all the chips (controllers) that go into cars, microwaves, instruments, and other equipment. Literally hundreds of millions of processors (controllers). Companies like Motorola, literally create their big processors (68000 and PowerPC) as a test bed for high performance versions of their controllers, then they move that technology down into the controller chips. The R&D money is spent on the big chips -- but often lots of money is made on the smaller ones.
Motorola has moved almost all their controller business over to PowerPC based designs. This is huge business. Motorola could be losing money on every Desktop PowerPC chip made (which they aren't) are still make money -- because the Mind-share in the big chips, guarantees market share for the smaller "micro-controllers". IBM is also in the controller business, and the same rules apply to them -- but not for Intel. The Pentium is a big old bloated design that uses too much area (and power) to be a good controller. Intel used to market lots of controllers but they have been lost this market to others (both AIM and the companies cloning x86 design).
IBM is an interesting situation as well. They were already using the Power chips (RS/6000's) before the PowerPC's. Remember, the PowerPC is just a less expensive version of the Power Architecture. IBM uses these RISC chips for their mini-computers (AS/400) business which is huge. Most of IBM's real business comes from these larger machines. These AS/400's often have dozens or hundreds of the PowerPC's in them. Some for I/O cards, or controllers, and some for the main processors. Some of the PowerPC's used in the AS/400's are specialty versions (designed just for those big computers), many are not. IBM is using the PowerPC to subsidize the costs of R&D for all their mini-computers (and visa versa). So IBM can lose money on Every PowerPC they sell to the outside world (which they don't), and still make the money back on the rest of their business. Intel can't do that.
Deeper-Blue (the computer that beat Chess Champion Kasparov) was built around PowerPC's. So the PowerPC is the only processor in the world to beat a world-class chess champion. Of course it was 32 PowerPC's that did it -- but it does show the Processing Power of the PowerPC. The first variant of Deep-Blue, that lost to Kasparov, was built around the Pentiums.Super-computers (massively Parallel Computers) are being built around the PowerPC. Intel took on a project to build a superduper computer with lots of Pentiums that would break 1 Terraflop (1 Trillion Floating Point Operations per second). IBM has been working at the same time to build a super-computer based on PowerPC's that breaks 4 Terraflops. Even when Intel was hyping their 1-Terraflop super-computer, I believe the partially completed IBM machine was already beating that (it was at about 1.5) -- but was still not completed and up to their 4-Terraflop design goal -- so IBM did not publicize it.
Conclusion
PowerPC has been a raging success. AIM has been producing chips that out perform the competition often by 2:1. The design advantages of RISC have paid off in the past with much less design money producing chips that are faster, use less power, and are cheaper to manufacture than Intels' chips. But the bigger issue is that the superior design is allowing much more money to be put into future designs, which will allow the superiority and advantages of the PowerPC processors to increase.
Motorola is huge, and most of their chip business is being built around the PowerPC and its continued success. Motorola can not, and will not, allow the PowerPC to fail. It is not only important as a main-stream computer processor, but it is important to their large controller business as well.
IBM is still the largest computer manufacturer in the world. Their mainframe business, controller business, and much of their chip manufacturing business is built around the PowerPC. IBM can not let the PowerPC fail. They are committed to the PowerPC, more than any other company.
So to believe that the PowerPC will remain anything other than a raging success, you must believe that IBM and Motorola (as well as Apple) will disappear or give up huge amounts of their business. Size matters, and IBM is much bigger than Intel. Motorola is also nearly as large. The irony is that the industry analysts see Intel as "big" when they are dwarfed by the competition that is the PowerPC and AIM alliance -- and the PowerPC is riding on a superior design that is years ahead of Intel. Yet some naively think that PowerPC is the underdog, and Intel is the favorite.
|
