% fortune -ae paul murphy

Will Apple switch to Intel?

There's a report, apparently in today's Wall Street Journal but nicely summarised on the Mac Observer to the effect that Apple has been negotiating use of Intel CPUs. As part of the report there's speculation that Apple could abandon the PowerPC for Pentium and the usual less than astute observations that this could bring down Apple's prices while improving the Mac's performance.

There is reason to believe that Apple would talk to Intel and may have some news to announce soon, but there's no truth anywhere in the assumptions most of these "journalists" are making about Apple's motivation.

First, if you compare equivalent systems, Apple's products are cheaper than Dell's. Try it for yourself, go to Apple's on line store and pick a standard configuration laptop, desktop, or server and then build the nearest equivalent you can using Dell's on-line pricing. What you'll find may surprise you: Dell's low end products are cheaper, but by the time you add components to gain comparability they cost more -and that's true right across the line with Apple's high end product a good $1,500 less than Dell's nearest comparable workstation.

Second, when you compare performance it's important to remember that the operating system and applications environment count too. Mac OS X does a lot more than Windows/XP, but needs more system resources to do it. If, therefore, you want to make a reasonable guess about the relative performance of Mac OS X and Windows/XP on the same 3.5Ghz P4 box, you need to first level the software playing field. That's relatively easy: just look at grid style super computers where everybody uses whatever Unix is optimal for their hardware and runs the same applications. Compare, for example, performance at the Virginia Tech machine made up from Apple's X-serves to the NCSA machine in Illinois made up from Dual Xeon Dell servers and what you find is that the Apple boxes contribute 30% more bang per box than the Dell's do.

There's a double bottom line here: Apples to Apples, Macs are cheaper than PCs, and their processors are faster too.

So why would Apple be talking to Intel? Assuming the conversation has been about primary system CPUs I can think of three possible reasons:

1. due diligence. If Apple is about to make a CPU decision - whether that's simply the renewal of a major contract or a major change doesn't really matter- they have an obligation to their customers and shareholders to carefully check out the competition.

2. with the goal of assessing the costs and consequences of adopting the Itanium. Apple has a serious problem with IBM -a company that has not met its CPU supply obligations to Apple and has all but announced its intention to take Apple down roughly ten seconds after Steve Jobs eventually retires. Right now, if Apple sticks with the dual core and later 970 architecture they'll find themselves behind the performance eight ball with respect to Microsoft's XBox -for which IBM developed the CPU.

   Converting to the forthcoming IBM Cell processor isn't a winner either. Not only does this continue an uncomfortable relationship, but it involves tremendous technical change and traps Apple into running MacOS X on the same basic box IBM and Toshiba are prepping for an assault on Wintel's Asian markets.

   The Itanium hasn't been a winner, but is actually quite a reasonable CPU design with some potential to give IBM and the PowerPC a run for their money. It's certainly significantly faster and more secure than x86. I think Apple's adoption of the Itanium highly unlikely, but Apple's people would be remiss if they didn't consider it.

3. As part of someone's takeover of Freescale Semiconductor. I don't know what rights, if any, Motorola's spin-off retained on the PowerPC but if Freescale has any rights at all it has to be a tremendously attractive takeover target for both Apple and Intel. For Apple because it might offer a way of balancing IBM's market power, and for Intel because Microsoft's new relationship with IBM threatens to leave it the odd man out in the next generation PC processor wars.

Oh and the right answer? Just for the record my belief is that Apple's best route out its strategic CPU dilemma is UltraSPARC. I haven't seen BSD on Niagara (Sun's eight core, thirty two thread, 65 watt SMP chip) but if the Solaris systems now in lab and user test are any indication this could be a winner. Individual processes run more slowly on Niagara than on, say, the UltraSPARC IV, but more of them run in parallel and the power benefit should be enormously attractive to Apple -and the software barrier to SPARC is much lower than it is to either Itanium or Cell.