The x86 Mac that is not a PC

An innumerable horde of "analysts" and "experts" have predicted - not just recently but ever since Apple launched Mac II, the business Mac - that Apple would inevitably have to shift their hardware architecture to an x86-based one. Continuing to state the obvious, factual arguments for this (as opposed to the "Don't fight Goliath" yells) are probably more well founded now than they have been since the days when Apple was stuck using Motorola M68040, at an embarrassingly low clock frequency, for over 2 years. Adding to the embarrasment, the first generations of Apple's Saviour, the PPC CPU, were vastly less capable than the 68060, the last of the M68Ks, even though it ran at a slower clock frequency... 66 MHz

But I'm digressing - as I am wont - from my pretext: Intel processors are leaving PPC's in a cloud of laboratory dust, racing towards their (realistic) year-end target of delivering a Pentium II capable of running at 3GHz. Excuses abound, such as, "You are just caught up in the Mega Hertz Myth", or the more desperate and logically incoherent, "They are just pouring millions and millions of extra transistors into their designs in order to try and alleviate the inherent weaknesses of the original 4bit architecture!".

That may be so. But what is wrong with doing that if it can be done and it works? After all, Motorola started throwing pipeline stages at their G4 designs as the only way to get a financially viable yield of over-500 MHz units. As for the MMMyth, it is only true that far - and by now we are a bridge too far (or rather, have remained at the MPX bridge far too long!). Mixing metaphors? you bet. Mixing problems? only to a certain extent. Until the GPX bus is replaced by something faster, there is not much point in moving CPU speed beyond 1.25 GHz - even if it were possible. And as for "My architecture is more elegant than yours": that one doesn't move many boxes from the feeding lines into buyers' homes.

I don't like it, but I can't suppress it any more: the Macs are falling behind. Strangely, the financial "double-dip" has alleviated this a little: lower sales figures can be hidden behind the general gloom and doom. But only for so long. And then what to do?

The trusty gossip columns and rumor discussion groups are full of hopes that the Big Blue cavalry will ride into town on October 15, bringing with them salvation and speed, bread(th) and bits, safety and sales. Any realistic analysis, however, shows this to be a "medium term" option at best. Short term it looks gloomy no matter in which direction you scan the horizon.

So why not bite the bullet (hoping it's not going to dum-dum you) and go x86? One suggestion has been to get AMD to make a modified version of their upcoming Pentium killer - nick: "Hammer" -, peel off its x86 code translation layer, add a PPC code translation layer, add an AltiVec, add a code translator layer for that.... no, you are right, it hardly seems an option.

Which leaves us with x86 pure, whether AMD or Intel flavor (there, I said it!). A Hailstorm of protests have battered this idea, most of which point out that if Mac OS (X?Y?X86?) is made to run on a PC-compatible computer, it will mean the end of Apple as a financially viable company: users will buy Dell hardware (or whatever you want to call it) and run MacOS Whatever on it, leaving little in the manner of revenue on the Apple Shop desks. Developers will be tempted to spend time on only one x86 version of their software - the one with the largest user base. Apple's profile will not be anything like that - more a hole, for previous pundits to hide in. MacOS Something will end like less than that - just a "Yellow Box"-like subset under Windows and nothing more. Can you say B?

So why mention it at all? Because the x86 CPU is just that and nothing more: it does not imply a metamorphosis of any chipset touched by it into a Windows-compatible pumpkin. It is possible to build a non-compatible x86 based model, using a blend of standard components, some central proprietary units, the very newest standards, and a very different architectural topology. Here we are not talking a PC with a proprietary "bootstrap" ROM which will be hackable with a modicum of ingenuity, we are talking something so different that no present Windows version will be able to run on it - and future ones will demand so much development work that it will not be feasible.

But what do I hear in the blue yonder? A clarion call of "AltiVec"? Indeed a corner to paint oneself into if ever non-corner compatible brushes existed! But this brush is different. If one can pair (or more) AMD processors with a VMX-compliant DSP processor, using one of the fast new bus protocols, such as HyperTransport, developed specifically for inter/chip communications - wouldn't that make AltiVec code (much of which is handwritten assembler code) a reasonably simple task?

Of course, Classic would have to be abandoned and Carbon demoted in favor of Cocoa (which is better anyway, due to its object/code nature rather than to any cross-platform capability (though the latter is a secondary benefit deriving from the former)). In other words, we are talking a last-ditch, white-of-the-eyes strategy.

This is not a statement, not a rumor, not a *Confirmation* but a proposition. It is not a short-term solution - at most a medium-term one - and I don't know enough about hardware development to be able to state confidently that it is doable. Others must tell me, as I hope you will.

It is possible. But it is Via Dolo Rosa.