Lots of people have made the claim that Apple's issues with IBM as the G5 developer/supplier were focused around price and volume - with lots of commentators cheerfully rewriting history to say that IBM wrote Apple off as a bad customer.
That didn't happen -at least not until after Apple made its decision. What I think did happen was that IBM failed to meet Apple's needs on both volume and performance while claiming to be making best efforts to meet those needs. At the same time those best efforts were failing, however, it successfully designed and built a three core, 3.2Ghz, Altivec equiped processor for Microsoft and a 4Ghz capable Cell board containing a G5+ master control CPU (again Altivec equiped) for itself and its partners Sony and Toshiba.
From Apple's perspective, however, the credibility -whatever the legal realities- of IBM's claims must have seemed doubtful (especially when it turned out that Microsoft's cost on the 3.2Ghz three core machine would be a bit less than Apple's cost on a single core 2.7Ghz G5).
Worse: if, as I believe, G5 supply and advancement constriction was the stick for beating Apple, the carrot was an offer to make Apple a Cell partner -something that would give IBM almost total control over Apple's software and supply line while weakening Apple's competitive position vis-a-vis Sony.
As a result Apple's decision to drop IBM as a supplier amounted to chewing off a leg to escape a death trap - sensible in the circumstances and something which would have been a non issue if Plan B had been updated in the last few years and a better alternative architecture selected.
Unfortunately, however, Plan B hadn't been updated for years; they made the Intel decision, have now seen some of the consequences, and therefore face some tough choices.
Option one, of course, is to just push ahead and hope things turn around. It's not an impossible bet: Apple has the market momentum to muddle along until Intel makes progress, there's probably room for some compromise on price, Apple can push the heat and power use envelope to its limits on the iMac and MacBooks to get a small performance boost out of increased megahertz, Intel's new Macboards will be marginally harder to break, and the $1,100 iMac can be refocused as the better family entertainment center device replacing the $499 MacMini.
Option two, however, is much better. To see what it is lets start by looking at three laptops: two real ones, and one that might have been:
The MPC7448 is a uniprocessor, 32bit (but with 36bit addressing), machine starting at 1.7Ghz. It's been commercially available since last June, with the dual processor MPC8641D series based on the same e600 core becoming available in large quantities pretty much on the same schedule as Intel's "Yonah" line - meaning it's available now but production ramp-up is just beginning.
In the table below, therefore, I'm comparing two real Apple notebooks, the Powerbook G4 and MacBook, to a hypothetical one: an entry level Powerbook e600.
|Powerbook G4||MacBook||Powerbook e600|
|Screen (Resolution)||15.2-inch TFT (1440x960)||15.4-inch TFT (1440x900)||15.2-inch TFT (1440x960)|
|Base Processor||MPC7447A 1.67Ghz||Intel Core Duo (1.83Ghz)||MPC7448 G4+ (1.7Ghz)|
|High end Processor||n/a||Intel Core Duo (2.3Ghz)||MPC8641D (2.1Ghz)|
|Graphics Processor (RAM)||ATI Radeon 9700 (128MB)||ATI Mobility Radeon X1600 (128MB)||ATI Mobility Radeon X1600 (128MB)|
|Base Price (USD, list)||$1,999||$1,999||$1,999|
|base DENmark score||537||442 (=2 x 241)||762|
|high end DENmark score||n/a||560 (=2 x 280)||1600 (Guestimated)|
|Estimated battery life (Claimed)||4 hrs (5.5)||3 (?)||>4 hrs (>5.5)|
|Estimated CPU cost as a percentage of list price||<4%||>13.5%||<4%|
The MacBook should do a lot better on less CPU intensive benchmarks, but overall this is hardly a contest. That hypothetical Powerbook e600 series fits with Moore's law to protect the brand while giving Apple the ability to deliver a lot more for a little less.
So what can Apple do? Two things:
- Re-establish the Mac as a premium product by dropping Intel as a CPU supplier and signing a three to five year deal with Freescale to use the e600 and subsequent e700 series processors.
- Make the MacOS X integration shell and related Apple software available as a software only package for Linux.
Taken together these two steps would strengthen the brand while buying developer loyalty, a new revenue stream, time, customer confidence, and market position.
Doing this would give Freescale the resources needed to expedite the production ramp-up on both chipsets, reduce Apple's board costs by up to 50% (because networking and memory controllers are built in), let Apple protect and advance the PowerMac and X-Serve lines (because even the low end 7448 uses 36bit addressing) and give Apple's customers back their traditional price, performance, design, and security advantages over Wintel.
Doing this splits Apple's computer business into two lines: one offering software and support for Lintel; one offering premium "it just works" packages built on Apple branded PPC hardware. That opens major new markets for Apple's developers and directly extends Apple's own markets and customer base into the Linux community.
This is a recipe for market penetration - a way to reach Linux and Windows users who would otherwise never touch a Mac. It's also a recipe for market consolidation - strengthening the halo effect and locking in sales in the music, video, and related home entertainment markets.
The key to all of this, of course, is customer confidence. What happens if Steve Jobs gets up in public and says "oops!"? The answer is well known, if counter-intuitive: customers trust companies that take responsibility for their mistakes. That's the fundamental lesson Procter and Gamble taught the world in the contaminated Tylenol case.
Bottom line? Steve Jobs, and Apple, could apply the Midas touch to MacTel simply by killing it -producing a far better outcome for Apple than the best they can hope for in sticking with Intel.