Have a look at this story over at news.com. Whoever this Andrew Neff fellow is over at Bear Sterns is, he sure isn't very in-tune with the technological aspects of his outlandish claims. The idea that Apple can transition their entire architecture over to x86 within a matter of two to four years is just crazy. I am not saying this will never happen or that it cannot happen, but Neff has totally discounted the PPC architecture. Apple has openly invested heavily in the upcoming G5 architecture, to the point where they had Apple engineering teams working alongside Motorola engineers.
There are quite a few reasons Apple cannot just switch gears from PPC to x86. Publicly only the Darwin open source kernel can be compiled and run directly on the x86 architecture. There is a lot more to the MacOS X architecture than just Darwin. The most important aspects of OS X (Aqua, Classic, Cocoa, Carbon, and Quartz) cannot just be recompiled for x86. Almost all of these aspects have been heavily optimized for the PPC architecture. These parts of OS X written for the PPC instruction set will not run natively on x86 w/o either an almost complete rewrite or some type of an emulation mode. Emulation is not the way to go when performance is what you're after. Ask Linus over at Transmeta about how Intel beat their Transmeta (code translation) architecture over the head.
So what can Apple do? PPC as powerful as it is just doesn't seem to have the type of R&D power behind it like x86 has with Intel and AMD. Well in my humble opinion Apple has two options:
1. Outright buy the PPC intellectual property from Motorola and take over the R&D aspect of the PPC architecture and then do the FAB with some independent FAB out in Taiwan.
Problem with this option is that it is a very expensive and investment intensive undertaking. Its a fact that designing new processors and new architectures is hard beyond definition and to keep this up over time is even more impossible. However, Apple has a lot of experience with the PPC architecture and being in firm control of the architecture gives them obsolescence protection like none other (Motorola has less interest in keeping the PPC architecture fresh, they make more money with cell phones and embedded designs). Furthermore this also gives them the ability to maintain their codebase, platform dependency and closed architecture advantages.
2. Slowly start rewriting Aqua, Cocoa and Quartz for the x86 architecture and then at the same time write a very very tight emulation layer to run Classic apps the same way you can currently under OS X. While the code monkeys are doing those rewrites start investing in a closed architecture like they have now with the PPC but with x86. I think the XBox is a good proof-of-concept that closed architecture x86 is possible (its not uncrackable but its complicated enough to keep the 5Cr1P+ K1Dd13$ away). Based on this same train of thought imagine for a minute if Apple joined nVidia and together they designed some kind of platform based on x86 processors and nVidia's nForce chipset. The hardware synergy on the x86 platform is awesome to say the least, but Apple's forte has always been amazing software. The amount of rewrite for all the iApps plus a lot of the core system level utilities would be just a tremendous amount of work.
These two options would pose a staggeringly difficult choice to have to make. On one hand Apple gets to hold on to their PPC roots and knowledge but take a huge risk by getting into the cut throat processor development arena. On the other hand Apple has to make huge sacrifices to their tried and true PPC roots and attempt to move their unique style to an entirely new platform.
In my opinion the latter choice would offer far more payoffs because of the style of partnerships that could be made with hardware vendors like HP and Sony. Whatever choice Apple makes it will be the choice that will either make or break them. Even with their godly industrial design and user interface teams the marketing hype can't keep up with the widening MHz gap (G4 tops out at 1GHz, P4 tops out at 2.53GHz going to 3 by Q4 '02).
So with all that said and done, heres a toast to APPLE! Truly a computer for the rest of us!
P.S. Its important to take your glasses off before going into the shower.
Posted by Mr. Keyur at August 3, 2002 03:46 PMInteresting article. But, I don’t understand this
fascination with clocks. I always thought the
measures of quality were instruction cycles, not
mere clock cycles. I also think thermal efficiency
is still a good measure of reliable design. I have
been trying to figure out why computer development
has stalled. I used to think it was the monopolies.
Now I think it must be the stupidity of the user.
Best regards.