Apple is set to wrest control of the PowerPC platform away from long-time partner Motorola through a clause in its agreement with the chip maker that allows it to buy Motorola's PowerPC assets for $500 million next year.
That, at least, is what is suggested by an email allegedly sent by an Apple insider to the MacOS Rumors Web site. It's also supported by a recent recruitment drive at the Mac maker.
The email is an attempt to put the record straight on a number of comments previously made by MacOS Rumors relating to the apparent cooling of Motorola's interest in high-end (ie. desktop) PowerPC processors. Such chips are, of course, bread and butter to Apple and essential for its on-going business strategy. With Intel set to release a 1.8GHz Pentium 4 next month, and AMD about to offer a 1.53GHz Athlon 4 that offers comparable performance, Apple's CPUs are once again looking - even if they're not in fact - well below par.
Motorola and the G4
Accelerating PowerPC up to 1GHz is the job of Apollo, which the email confirms as the PowerPC 7460. Interestingly, the Apple source suggests this will be the chip that brings G4 technology to the iMac and iBook rather than Apple's pro boxes.
As a member of the 74x0 family, Apollo is an enhanced version of the current 7450, developed around the same architecture but build on IBM's silicon-on-insulator technology to cut its power consumption and production costs, and boost performance. That's good for Apple, because it gives it an enhanced G4 for portables and iMacs.
Motorola, the source claims, remains committed to the development of Apollo. That sounds about right. With the 7460, Motorola will get a G4-class chip even better suited to embedded applications - which are what the company is really interested in.
Incidentally, Motorola's next chip release is the PowerPC 7440, which is a low-power version of the 7450. It's due to sample next quarter and go into volume production in Q4, in time for new Macs in Q1 2002, perhaps.
In the meantime, a Power Mac G4 at that speed being launched late August, reckons the source (though whether it will ship then remains to be seen). Early summer (Macworld Expo New York) will see 800MHz and 933MHz Power Mac G4s announced.
The source says all three boxes use "G4 processors". Apple's pretty free with the term G4, using to refer to any G4-class CPU. Since Motorola hasn't officially launched the 7460 yet, and given the source's earlier comments that if won't be used in professional-oriented desktops, we reckon the 800MHz, 933MHz and 1GHz parts are 7450s.
Moving up to the G5
The source claims the 7460 will be released alongside the G4's successor, the G5, codenamed Goldfish, the chip that will take the Power Mac family above 1GHz. Of the G5, the source claims it will tape out this coming autumn and go into production ramping through Q4 2001 in time for "systems [to be] ready for release at the January Macworld show".
Release maybe, but not shipment. While we can imagine G5 taping out in a September/October timeframe as the source suggests, such a rapid ramp seems unfeasible, no matter how much Apple CEO Steve Jobs browbeats Motorola into speeding the process up.
Still, with a plentiful supply of fast 7450s and the publicity generated by launching high-speed G5s, Apple may be tempted to rush the release, as it did with the 733MHz 7450 last January, more to stress to the world that it is catching up with x86.
The source claims early G5 samples are hitting 1.6GHz and averaging 1.13GHz. If so, further development work between now and the end of the year, should see the lower of those two clock speeds settle as the chip's bottom end. Top end chips may well ship at 1.6GHz - it all depends on the yields Motorola is getting.
This timetable takes us into 2002 and the point at which things really get interesting. Says the source: "Apple has the option in 2002 of buying the entire PowerPC assets from Motorola for $500 million."
If the G5 is indeed shipping early Q1 2002, Apple will already by looking at its successor - we'll call it the G6; the source certainly does. Motorola hasn't committed itself to the G6, says the email, and that may be sufficient incentive for Apple to take over the development of high-end PowerPCs.
The relationship with Motorola hasn't been an entirely happy one over the last few years. Conspiracy theorists will blame Jobs' decision to kill the Mac clone programme, but that's unlikely to be an issue - it hasn't proved to be one for IBM, after all. No, it's Motorola's preference for the embedded market, where it makes its money, that has made it difficult for the company to focus on the tasks that Apple would like it to devote its attention to. And given Motorola's poor financial performance over the last two or three quarters, it's likely to be even less willing to devote significant development resources to high-end PowerPCs than it is now.
If so, a breach will open between Apple and Motorola sooner or later. The source suggests Apple is anticipating a quick break rather than a prolonged one, and is already preparing for it.
"Apple has invested 50 million dollars in developing a 0.10 micron lithography process for future PowerPC processors last year," the email claims. "Apple has also contributed a large amount of engineering staff to the project, and in fact, much of the design work on the G5 is being done in Building 2 of the Cupertino R&D campus, rather than at Motorola's Austin Texas facilities."
Circumstantial evidence bears some of this out. In May, we reported that Apple was advertising for a "Microprocessor Development Project Manager" to "manage the development of new high-performance microprocessors for future Apple computer products". The applicant needed at least three years' experience "managing high-performance processor development through production ramp".
At the time we speculated that this appointment might mark a rekindling of Apple's interest in PDA processor design, but in the light of the Apple (allegedly) staffer's comments, it takes on a new significance.
Apple does a Transmeta
But before putting two and two together to make five, is it feasible for Apple to spearhead PowerPC design? It does. Apple already has plenty of experience here, developing its own chipsets - the logic that links the processor to the rest of the system - and don't forget it bought Raycer a year or so back. Raycer was developing 3D graphics acceleration silicon, and while there's not much call for that these days, the acquisition has provided Apple with further chip design expertise.
The templates for Apple are companies like ARM, Transmeta and Imagination Technologies. These so-called 'fabless' chip makers do all the design work and either license the results or contract out manufacturing to designer-less chip foundries. It's a business model that works, and one that's increasingly replacing monolithic operations - companies that design and produce chips - like Intel and Motorola.
And Apple won't be alone. It has what appears to be a healthy partnership with IBM. Big Blue seems rather more keen on the PowerPC platform than it appeared to be a few years back, partly through the work it has done for Apple and through its partnership with Nintendo. It has already begun plotting out its PowerPC roadmap out beyond 2GHz.
At the very least, IBM provides Apple with world-class production techniques as a contract manufacturer. Apple could also look to the rising stars of chip production, such as the Taiwan Semiconductor Manufacturing Company, whose 0.13 micron process is arguably ahead of IBM's, one reason why it has displaced Big Blue as Transmeta's chief supplier.
And if it all goes pear-shaped? "Apple has several contingency plans regarding future hardware," says the email. "Mac OS X has compiled successfully on three different RISC architectures of which I will not name."
If one of them is Compaq's Alpha, we can now discount it - the Big Q just sold it to Intel in return for a commitment to Itanium, Intel's 64-bit server chip. Though, as one Reg reader points out, that may leave Apple with a handy pool of unemployed engineers from which to recruit.
Speaking of Intel: "[OS X on Intel] is very unstable because of the Intel architecture's legacy architecture, making them the most computationally inefficient processors period."
There's some truth there, but the vigour of the damnation counters the argument that the email is authentic - it's just a little bit too redolent of all that infantile 'my CPU is better than your CPU, because you CPU is poo' stuff.
In any case, Apple has far more pragmatic reasons for avoiding x86 - be it Intel or AMD - platforms, primarily the loss of hardware sales.
Is the email authentic? It's impossible to say. However, certain parts of it describe plausible directions for future Apple's processor strategy that are supported by other evidence. We remain cautious, and we call on Apple to clarify its strategy, preferably with its partners. We understand its reticence, but having explained where it wants to take its operating system, it really is high time Apple outlined its vision for the platform on which that OS is based. ®