Dojo (HowTo)







  Easter Eggs




  Martial Arts

PowerPC chip hits fork in the road
Old news is no news!

By:David K. Every
©Copyright 1999

Another article (or three) have started polluting the Internet with errors and misinformation. I think this is the PC crowds last gasp at making Apple or the Macs seem irrelevant -- or a desperate attempt to scare people out of making reasonable choices. Remember, Apple's sales are up, and people are using their brains more and more -- wouldn't that scare you if you depended on people's ignorance to come to you for information and decisions?

Months ago there were some half-informed observations, that were over sensationalized into, "IBM and Motorola Split" and how IBM hates Apple, and the PowerPC was doomed, and so on. It was all over generated hype, meant to get attention and scare the Mac customers.

Along that line, coming in 6 months late, and with nothing new, Brooke Crothers offers his article (via C|Net) ripe with too little wit or wisdom. The author does its best to scare up fears and make people concerned for the PowerPC and drive questions in their heart -- of course it would be nice if he had a clue as to what he was talking about.

Even his intro tells the whole desire of the article:

Motorola and IBM have very different plans for the PowerPC chip, raising questions about its future as the processor powering the Macintosh.

Is there any doubt where this article is going to go? Scare the masses, make the PC apologists feel secure. The article creates its own reality from the title on.

IBM and Motorola both produce PowerPC chips and currently supply them to Apple and other hardware makers. But the continuing fallout from the dissolution of a seven-year Somerset partnership indicates that IBM is now focusing its PowerPC chips, more than ever, on internal server computers, likely leaving Motorola as the sole major supplier to Apple, according to analysts. IBM will continue to develop 64-bit PowerPC chips for its AS/400 and RS/6000 server and workstation lines.

OK. There is a lot of politics when two (or more) companies try to mix goals. IBM was doing 64 bit versions of the Power Architecture (the forerunner to PowerPC) long before Somerset -- of COURSE they are going to do it long after. That was what the breakup was about. IBM felt that they were contributing more to Somerset than the others, and they wanted more high end PowerPCs (which wasn't Somerset's focus) -- combined with the usual politics and whose in charge games and so on. But the break wasn't hostile. It was that each wanted to be more free to evolve the PowerPC lines in different ways -- which means choice and GOOD things!

Motorola is also looking elsewhere. Though it will continue to produce 32-bit chips for the Macintosh, the company is eyeing communications equipment and other areas for PowerPC uses which may prove just as lucrative for the chipmaker and would allow it to defray significant Macintosh-specific development costs.

Something like 98.5% of communication switches (TelCo type) being sold in the U.S. are PowerPCs. "Eyeing" isn't exactly the correct word. There are also big wins in many areas besides that.

This fork in the PowerPC road has its starting point in June of last year when IBM and Motorola dissolved a partnership at the vaunted Somerset design center in Austin, Texas.

The starting point was in 1994 when Somerset was started. It has continued through every product definition and implementation phase. It was slowing down the decision making processes, and causing turf wars. But none of that is new to companies. Despite this they continued to develop great processors. Now that they are freed from infighting, the results will likely be MORE progress - not less.

Now, recent movements suggest the beginning of a cooling off of chip development for the Macintosh as Motorola alone tackles development of the next generation of "G4" PowerPC processors, according to analysts. Apple's Macintosh was ranked No. 7 worldwide in computer shipments in the fourth quarter of last year, according to International Data Corp., shipping over one million units in that period.

Cooling off? They are in the middle of development for the G4, heavy into the design phases of the G5 (MultiCore), and in the requirements and early design phases of the G6 (ISA and architectural enhancements). A company 1/100th the size of Intel (in engineering) is out producing Intel, releasing more architectural improvements, and bigger improvements, and the PowerPCs superiority in portable markets is unquestioned -- and competitive with the best Intel can offer in the high end (with a fraction the Power requirements, cost, size, and manufacturing costs). Geeze! Poor, poor Apple and Motorola -- their doomed!

"Having only Motorola do PowerPC development for Apple isn't as good as having IBM help too," said Linely Gwennap, publisher and editorial director of the Microprocessor Report. Motorola has typically trailed IBM in so-called "process" technology, the ultra-fine circuit etching techniques used to make chips, and thereby lagged behind IBM in bringing out the fastest chips.

Microprocessor Report has been re-architecting themselves from a technology journal, into a x86 and PC technology journal. When I went to Microprocessor Forum, every one I spoke to from any of the RISC companies (IBM, Motorola, HP, DEC/Compaq, TI, etc.) all were griping and grumbling about how all things RISC were being brushed aside because Microprocessor report's audience is mostly x86, and they are tailoring information accordingly. Taking one of Linely's quotes out of context can make that seem even worse.

IBM does have better process technology -- but Motorola knows that, and is acting accordingly. They're making alliances and realize their mistakes in the past. Also while IBM has good process technology, they also don't have as much volume in many of their higher end processes. So there are tradeoffs there too. Both Moto and IBM seem ahead of Intel, by a good amount, when it comes to process technology -- but I guess none of this was relevant.

IBM revealed aggressive plans for the PowerPC processor at a major chip conference in February. The company has been touting its advanced copper and SOI chip technology that will lead to chips running as fast as 1.1 GHz, almost three times the raw speed of today's processors.

Poor, poor Apple and PowerPC. Doomed to such speeds.

Of course the truth is that by the time those things hit the market, there will likely be 700-800 MHz (or more) by other manufacturers -- so the difference will not be three times the competition. But IBM is going to shoot for high speed (MHz). Not because that is always the fastest performing Microprocessors (in true performance), but because it is easier to sell fast MHz (and slow performing) rather than the other way around. IBM has learned this lesson the hard way as they have 200 MHz Power3's that out perform some chips (like 21264 Alphas) that are running 2 or 3 times the MHz. Despite the performance of the Power3's it is a hard sell because many can't see beyond the MHz. So IBM is trending back to higher speed (but lower performing at a given speed) processors. Where's the big deal here? Actually, Motorola is likely to do the same thing, as is AMD and others.

But the emphasis at IBM these days is on 64-bit chips that are used in its in-house computer lines--not the Macintosh. "64-bit is not a roadmap for the Mac," said Dennis Cox, a senior technical staff member with IBM's AS/400 group.

Guffaw. You ask an AS/400 guy, likely making some political play, what the directions of his company are, and what the hell do you expect him to say? He'll tell you that the AS/400 is the future. Of course the RS/6000 guy will say the RS/6000 is the future. The embedded guy will tell you it is the low-end PowerPCs like the 40x or 80x series. And so on.

Apple has little need for 64 Bit right now -- it would drive up cost more than it would return performance. But does that really matter? If the future of the PowerPC was 64 bit, Apple could get there in a hurry. But all this is quite ironic when it is right after IBM seems to have decided (internally) that G4s and AltiVec are going to offer serious advantages and they do want to be a second source for Apple. Motorola is getting so much interest with AltiVec that IBM realized the stupidity of ignoring it.

Though IBM has 32-bit chips--running as fast as 580 MHz--that could be used on the Macintosh, it is not clear whether Apple Computer will use them, according to Cox and others at IBM.

The processor code names Glacier (running at 580 MHz) is a G3, and will be coming out around the time as the G4 (let's say July). There is no doubt at all that if IBM can produce them that Apple will eventually use them (or something like them). While the future of Macs is definitely G4, it is going to take a while for that to permeate the product line -- especially low end and portables! So if IBM can produce them in volume, and produce them cheaply, then of course Apple will use them. If IBM can't do those things, then the failing is IBMs, and has nothing to do with Apple.

Meanwhile, Motorola is moving in a direction with a different emphasis, based on its AltiVec software technology for improving communications and multimedia performance.

Actually, since IBM is doing work on AltiVec as well (and owns rights to some earlier versions of it), it is not that different a direction -- just a slightly different focus. What will the numbers of G4s versus G3s be for each company and so on. But the embedded and communications markets are thrilled with the AltiVec unit -- and that too is forcing IBM to take notice. So the point is that both companies are doing the same thing -- they are making the fastest and cheapest processors they can (at different price points) and both are competing in how they are going to do that. This is bad for Apple how?

While Motorola is still committed to the Macintosh, it is also targeting the technology at the communications market which analysts said could prove to be quite lucrative.

Sure. And since high speed communications market like things like high speed processors that have streaming characteristics (like AltiVec) this helps Apple. More money for Motorola (or IBM) means more R&D and more innovation, and better, cheaper, faster processors.

Motorola is aiming the AltiVec instructions for hardware which, for example, controls pools of modems at large companies or as a processor for cat scanners, according to Will Swearingen, an executive at Motorola's semiconductor operations. There is no official word yet when Apple will sign on to AltiVec. "Apple is committed but they haven't said when," said Swearingen. Apple would not comment.

Apple doesn't comment on future machines because it can hurt current sales. DUH! Apple has stated clearly, many times that they are committed to AltiVec -- they just won't put a date on the release of future machines. If I was a betting man, I'd guess probably July MacWorld -- but there are a lot of variables in the equation, like inventory of other machines and so on. But Apple is being smart -- they are selling the great products they have -- instead of hyping everyone on the excellent products they will have in the future. (I also have no fear that there will be G4 upgrades for older Macs).

Other sources claim, however, that Apple is fully committed to AltiVec and will reveal systems later this year that exploit the technology. An IBM spokesperson said it could use AltiVec technology in chips if it chooses to do so but would not comment beyond this.

In other words -- companies don't like to pre-announce their strategies. Big surprise.

The reasons for IBM's slow retreat is tied to a shrinking market, said analysts. Had Apple continued to allow Macintosh clones to be produced, the outlook might have been different. "Looking backwards, Apple's halfhearted approach to cloning killed the PowerPC's Macintosh opportunity," said Nathan Brookwood, an independent chip analyst in Saratoga, California.

I am underwhelmed at Nathan's comment -- but it may have been taken out of context. Certainly one must look at the source. He looks at "independent chips" -- what is his bias? Take a guess -- he wants more people producing independent chips. Since Apple "killed cloning", their sales have gone up, their support costs have gone down, and more people are buying their products than before. This is not a bad trend. This whole paragraph is an oversimplification of things that are very complex -- like it is just as true that IBM and Motorola halfhearted attempts to advertise and market the PowerPC chip hurt Apple's ability and motivation to clone. But why try to explain something when you can oversimplify something into a useless caricature devoid of meaning?

Others see the PowerPC processor falling behind because of lack of incentives. "The problem is whether or not Apple will pay Motorola enough to get the processors that they desire; my suspicion is that they may have made that decision already, and that the PowerPC will fall behind," said Martin Reynolds, a chip analyst with Gartner Group Dataquest.

Garnter Group makes its money where? Doing analysis for the PC market maybe? Of course I believe it was many analysts like this one that said Apple was doomed years ago and that they were in a spiral they couldn't pull out of, and so on. So I don't put much faith in most analysts, particularly when their analysis has no depth or insight.

As for lack of incentives, Billions of dollars in communications controllers, embedded controllers, mainstream processors and so on, seem pretty damn lucrative to me. If there is not enough incentives to make cutting edge processors, then Motorola will just curl up and die, because it is too hard to compete. Most of the concepts in the mainstream processors do find their way down the line. Of course one could say that Intel has no motivation to compete because anyone will buy their crap, so there is a lack of incentives there -- but that would be as inane as Reynolds comment, and Brooke Crothers quoting it.

"On the other hand, Apple is no longer a performance story and the issue may well be irrelevant," he adds.

Meaning what? That because Mac Portables out perform PC portables by a factor of 2:1 that Apple no longer cares about performance? Or that because the highest end Macs still out perform the gross majority of PCs (even highest end ones) that Apple doesn't care about performance there? If he means that because Apple's real value add is their superior usability, simplicity and reliability -- and performance is just one valuable factor among the many advantages of Macs, then I agree with him.

Gwennap says Motorola won't be as focused on Apple as it has been in the past. "The problem is that Motorola is now pushing PowerPC into [other] markets, so it may not be as responsive to Apple's needs," he said, mentioning the "embedded market" as a primary target for Motorola with its new architecture segment which covers everything from communications devices to industrial products and handheld computers.

Motorola has always been focused on their embedded market. Motorola has also had some pretty bad management at times. But they are getting a clue, and realizing more and more that process technology matters across the board, and that the key to communications and most of their products has been making superior high end processors. Mindshare also leads to marketshare -- so coming out with hot processors is the key to keeping advances in their "cooler" ones as well. If Motorola doesn't realize that, then that is their failing, not Apple's.

"Motorola sees the PowerPC as part of its own opportunities in communications. Therefore, they will drive it upwards but with technologies such as AltiVec that suit their needs," said Dataquest's Reynolds.

Let me get this straight. Keith Dieffendorf was one of the principle designers of AltiVec and worked for Apple. Apple pushed Motorola and IBM in this direction, despite both of their resistance (early on). This has enabled Motorola to get huge amounts of interest in whole new markets. So because of that, Apple is doomed because Motorola will never listen to them again? Does this make sense to you?

Many (most) of the technologies that will empower Apple, will empower Motorola, and vise versa. Why these guys think it is mutually exclusive is beyond me. If nothing else, Apple can take their designs to fabs elsewhere -- but I seriously doubt there will be any need for that.

Meanwhile, IBM looks inward. "IBM can easily justify continued development Of non-merchant PowerPC components for its own system lines, based on the much larger revenues those lines generate. AS/400 is a $4 billion or $5 billion business, and S/390 is slightly larger," said Brookwood.

UH, S/390 is IBMs CISC mainframe (though it does use some PowerPCs as I/O controllers).

Yes, IBM has always been incestuous and concerned with fulfilling their own needs. But they also sell to the outside. Are they going to walk away from free money? Not only that, despite their arrogance, they are learning that some of the things Apple and Motorola were pushing for weren't such bad ideas. Second sourcing PowerPCs for Apple is easy -- and Apple can exploit the 64 Bit PowerPCs into server lines at some later date. If nothing else, IBM can become Apple's fab if Motorola and IBM drop the ball -- and Apple owns the rights to designs, and has many talented people of it's own. Things just don't seem that bad or extreme.

"Given the tremendous importance high-performance Power chips have for IBM's system products, and the relatively small size of the general-purpose Power PC [Macintosh] merchant market, it's not surprising to see IBM leave this opportunity to Motorola," Brookwood said.

Interesting since the political trend I hear about in IBM is the exact opposite -- that they have been looking more seriously at 32 bit versions and AltiVec... and all these points seem to be based on 9 month old reports.

The current state of the PowerPC, which was originally envisioned as a major competitor to Intel's x86 chip family, contrasts with the furious pace of the Intel-based market, driven by fierce competition between Intel, Advanced Micro Devices, National Semiconductor's Cyrix arm, and other Intel-architecture chip competitors.

What contrast? Intel just added Floating Point SIMD -- Motorola has been sampling them for a while, and the AltiVec SIMD is a 2 - 8 TIMES better performer (with spikes higher) than Intel's SSE or AMDs 3D-Now. While there is little or no architectural innovation coming from Intel (they are using the same basic P6 core that they used in the Pentium Pros in 1995), PowerPCs have made 2 different architectural jumps since then, and are about to make a 3rd, with a 4th due next year. AMD is due to make the K7 -- which should cause some competition for Intel, but no sane Mac would give up their superior performance and usability for a K7 based PC, even if it was faster (and it does not look like it will be).

So there is competition by 4 or 5 companies in the PC market, to all make the same exact thing (x86, power hungry processors). There are 3 companies on the PowerPC side of the fence, all making divergent and complementary (and competitive) solutions -- and some how the PowerPC is at a loss? Look at the last 5 or 7 years of Processor history. The Pentiums have been evolving according to Moore's law -- about 58% performance improvement per year. The PowerPCs have been exceeding that by about 10% per year -- or about 68% performance increase per year. The trends in the future look even better for PowerPC, since Intel is about to make a desperate move towards RISC (with Merced), and the PowerPCs size and heat advantages will offer even larger benefits in multiprocessing and multi-cored solutions.

Created: 03/13/98
Updated: 11/09/02

Top of page

Top of Section