Analysis Apple is becoming increasingly irritated with its prime PowerPC provider, Motorola, to the extent that it talking to fellow PowerPC partner, IBM, about how the platform can continue to evolve without the chips-to-cellphones giant's participation, sources close to the Mac maker have claimed.
We can't say we're entirely surprised. Like a marriage, the PowerPC alliance has seen its fair share of fights between partners, threats of divorce and, ultimately, reconcilliation. Steve Jobs' decision to can Apple's clone programme didn't go down to well with Mac OS licensees Motorola and IBM, and those two later fell out over how the processor should be extended beyond the G3: Motorola wanted to leap ahead with the G4 and its vector processing technology, AltiVec, while IBM believed that there was still plenty of mileage to be made out of the G3 architecture.
G3 vs G4
With hindsight, IBM was probably right. It has at least announced a 1GHz G3-class processor, and while Apple and Motorola sources claim the upcoming G5 goes like the proverbial off a shovel, the G4 has largely failed to provide the significant performance lead over the G3 that both companies promised. Yes, clock-for-clock, the G4 is faster and does provide higher clock speeds than the G3 too, but the real booster, its AltiVec engine, hasn't exactly inspired all those third-party Mac programmers out there to use it in their own apps.
"The G4 delivers superior performance in vector operations to its Intel counterparts, but it stacks up very poorly in terms of integer and floating point performance scores when stacked up against the P4," claims our Apple source. "This is because several key design improvements in the G4 that were proposed did not make it into the final product, which is what explains insignificant [real-world] performance advantages of the G4 over the G3 in non-AltiVec applications. The G5 will end this nonsense: in many tasks, there have been fourfold increases in integer and floating point performance."
But though it shows plenty of promise, the G5 remains unnannounced, with no official release schedule, and its Q1 2002 ship date remains tentative at best.
To all intents and purposes, then, the G4 can be regarded as little more than a faster G3. It should have been much more and the fact that it hasn't been must largely be down to Motorola. The bug in the original PowerPC 7400 that prevented it running efficiently above 500MHz arguably forced it to redesign the chip from the ground up in order to deliver higher clock speeds. That's what it did with the 7450, the second-generation 'G4 Plus'.
It's currently doing the same with the 7460, the third-generation G4-class CPU - codenamed Apollo - which will take the G4 to 1GHz and beyond. Why? Because, we hear, the 7450 won't run at over 900MHz, just as the 7400 couldn't exceed 500MHz.
Apple vs Motorola
Apple's current processor strategy is tied to Motorola's ability to get ever faster desktop processors to market. Given Motorola's record, it's no wonder Apple might be feeling frustrated. Megahertz Myth comments are all very well, but the bottom line (for now, at least - we've yet to see how AMD's new naming scheme tunes buyers into new performance metrics) is that it needs higher and higher clock speeds to avoid appearing to have been left way behind by the x86 world.
Who's in charge here?
To be fair to Motorola, its focus is directed toward other markets and Apple is only one of many, many PowerPC customers. Cisco, for instance, has committed itself to using the 7450 in next-generation routers. If the networking market picks up and begins to approach its size last year, Cisco could well become a bigger customer than Apple (assuming it isn't already). And at least Cisco won't constantly whine about not keeping up with Intel et al and won't impose unrealistic deadlines - something, we hear, Apple does rather a lot.
Essentially, both Motorola and Apple each want to define the pace and method by which PowerPC platform should evolve, and each wants to do so according to their own markets - the broad embedded arena for Motorola, or the narrow high-end desktop for Apple. These viewpoints aren't mutually exclusive, but they do bring very different priorities.
The logical move for Apple, then, is to take on the development of high-end PowerPCs itself - to steer the evolution of the platform in the direction that will best serve its desktop and notebook sales.
We've been hearing for some time suggestions that Apple might well pursue just such a course by buying Motorola's PowerPC assets. As we've seen, that would bring Apple considerable freedom of movement, but the downside is both the upfront cost and the ongoing expense of running a processor developent programme. That's a very expensive business, even if you're a fabless - ie. you don't physically make the chips yourself - chip company. Sun continues to develop its own processors, as does IBM, but both Compaq and Hewlett-Packard (even if they don't merge) have been chosen to end the development of their own Risc processors - Alpha and PA-Risc, respectively - and buy Intel's Itanium instead.
One alternative we've suggested before would be for Apple to form a closer partnership with IBM and together buy Motorola out. Apple benefits because it takes tighter control of PowerPC development than it has now and shares the risk and cost with IBM. IBM wins because it acquires technology it currently has to license off Motorola. It also gains marketshare and the customers that go with it. Unlike Apple, it has the resources to sell high-end PowerPCs to a broad, embedded customer base.
All this has to be seen in the context a sale of Motorola's chip division, the Semiconductor Products Sector, something the company has threatened to do should the operation fail to improve its financial position.
Bringing in Big Blue
Our Apple sources suggest that Steve Jobs has indeed discussed the matter with IBM chief Lou Gerstner, though lacking Jobs' diary, it's impossible to confirm this. It would hard to imagine that the two have not conversed on the subject, however. At the very least, we believe Apple has discussed signing IBM as a second source of G5 processors, just as, in late 1999, it brought Big Blue on board to produce G4 CPUs after it emerged Motorola was having problems making enough of them. Apple's move came after the G4's infamous '500MHz' bug, which force the company to reduce the clock speeds at which it had begun shipping its Power Mac G4 desktop line.
Contractual issues may make it harder for Apple to bring IBM on board as a G5 supplier, but past history has shown that it may have to. It may also look for alternative suppliers, such as TSMC, the world's largest chip foundry.
Our source's comments suggests that these are indeed the lines on which senior Apple figures are thinking, though it should be pointed out that it could equally simply be what the lower orders think what their bosses are up to. We'd happily take Jon Rubenstein's call to put us in the picture...
And if Motorola delivers the G5 on time and with the requisite level of performance, such contingencies may prove academic while the two companies continue to co-operate as before.
Exit PowerPC, Stage Left?
But what if it all goes horribly wrong? Mac OS X does, thankfully, provide Apple with a much more straightforward PowerPC Exit Strategy, though as our source notes, only if it's willing to drop Classic compatibility and Mac OS 9. Native OS X apps could be modified for a new architecture and recompiled relatively easily, but that's not the case with older software.
But shifting to, say, second-generation Itanium (aka McKinley) or Sun's UltraSparc III - both platforms suggested by our source and certainly the most likely alternatives - would require a significant effort to smooth the path for developers who have already had to go through one major transition (to OS X) as it is.
We can only see Apple making such a move if it had literally no other choice, such has been its investment - in money and intellectual capital - on PowerPC. That won't stop it exploring the potential of such platforms, and its important not to mistake experimental systems never intended to go beyond the lab for a strategic change of direction.
The bottom line is that everything hangs on the G5. Relations between Apple and Motorola may be strained, but that's no reason to assume that their differences are irreconcilable. But Apple would do well to make some contingency plans and, if our source is correct, it's wisely doing just that.
The marriage has, after all, been through more rocky times before. ®