Don’t panic! Apple adopts Intel processors

The confusion and furor surrounding Apple CEO Steve Jobs’ announcement at the WordWide Developers Conference that future Mac, after Jun 2006, will use Intel processors inside is totally unfounded. Nothing changes now, very little changes in the next year and longer term the future for the Mac got a little brighter. Although the decision caught me by surprise, as I thought about it, and listened to what was said in the keynote, I could see why it made sense.

If we look short term, the decision makes little sense. Right now a G5 (Power PC, aka PPC) PowerMac has very similar performance to the best workstations on the PC/Intel platform running Windows and the G5 will cost less than a similarly performing PC workstation. At the low end the Mac mini is competitively priced to a cheap Dell or other name brand. (Macs are not price competitive with off-brand PCs, the so called “white box”.) So, why put the developer community, and developers within Apple, through the pain of a processor shift?

For the future (“we have to do it for the children”) and because it’s really not that painful for most developers.

Right now a G5 PowerMac is very performance competitive with the best offerings from Intel. What Apple have been privy to, that rest of us haven’t, is the future of both Intel processors and PPC processors. Based on that future Apple decided they had no choice but to make the change. In the future, the performance-per-watt of power of a PPC chip will be “15 units of processing” according to Mr Jobs. The same watt of energy would give 70 units of performance on an Intel processor. Without knowing exactly how those figures were derived, and what it means for real-world processing power it seems like a significant difference. It was enough to push Apple to make the change.

Not that there’s anything wrong with the PPC architecture: IBM continue to develop and use it at the high end and PPC chips (triple core “G5” chips) will power the Microsoft XBox360. The sales of chips to Microsoft will well and truly outweigh the loss of business from Apple. It is, however, a crazy world: next year will see a Microsoft product powered by PPC and Macintoshes powered by Intel!

Steve Jobs demonstrated how easy it will be for developers to port applications to OS X Intel. In fact, he confirmed long-term rumors that Apple have kept OS X running on Intel processors with every development on OS X – Mr Jobs demonstrated and ran his keynote from an Intel Macintosh. For most applications a simple recompile in the Xcode developer environment will suffice – a matter of a few hours work at most. Moreover, even if the developer does not recompile, Apple have a compatibility layer, called Rosetta, that will run pure PPC code on an Intel Mac. Both platforms are to be supported “well into the future”.

During the keynote Mathematica was demonstrated (huge application, 12 lines of code from 20 million needed changing, 2 hours work) as were office applications. Commitments to port Adobe’s creative suite and Microsoft’s Mac Business Unit software were presented. Apple have been working on Intel-compatible versions of all their internal applications according to Mr Jobs. [Added] Luxology’s president has since noted that their 3D modelling tool modo took just 20 minutes to port, because it was already Xcode-based, and built on modern Mach-0 code.

Remember, these applications are for an Intel-powered OS X Macintosh. No applications are being developed for Windows. In fact, after the keynote Senior Vice President Phil Schiller addressed the issue of Windows. Although it would be theoretically possible to run Windows on an Intel Macintosh it will not be possible to run OS X on anything but Apple Macintosh.

Apple’s Professional Video and Audio applications might not be as trivial to port although most of the modern suite should have no problem. LiveType, Soundtrack Pro, DVD Studio Pro and Motion are all new applications built in the Cocoa development environment and will port easily. Final Cut Pro may be less trivial to port. It has a heritage as a Carbon application, although the code has been tweaked for OS X over recent releases. More than most applications, Final Cut Pro relies on the Altivec vector processing of the PPC chip for its performance. But even there, the improvement in processor speeds on the Intel line at the time Intel Macs will be released are likely to be able to compensate for the loss of vector processing. At worst there will be a short-term dip in performance. However with Intel Macintoshes rolling out from June 2006 it’s likely we’ll see an optimized version of Final Cut Pro ready by the time it’s needed.

[Added] Another consideration is the move to using the GPU over the CPU. While the move to Intel chips makes no specific change to that migration – Graphics card drivers for OS X still need to be written for the workstation-class cards – Final Cut Pro could migrate to OS X technologies like Core Video to compensate for the lack of Altivec optimizations for certain functions, like compositing. Perhaps then, finally we could have real-time composite modes!

Will the announcement kill Apple’s hardware sales in the next year? Some certainly think so but consider this: if you need the fastest Macintosh you can get, buy now. There will always be a faster computer out in a year whatever you buy now. If your business does not need the fastest Mac now (and many don’t) then do what you’d always do: wait until it makes sense. The G5 you buy now will still be viable way longer than its speed will be useful in a professional post-production environment. It’s likely there will be speed-bumps in the current G5 line over the next year, as IBM gets better performance out of its chips. We are waiting for a new generation of chips from Intel before there would be any speed improvement. If Apple magically converted their current G5 line to the best chips Intel has to offer now, there would be little speed improvement: this change is for the future, not the present.

So, I don’t think it will affect hardware sales significantly. As a laptop user I’m not likely to upgrade to a new G4 laptop, but then there will be little speed boosts available there in the next year anyway. But as a laptop user, I’m keen to get a faster PowerBook and using an Intel chip will make that possible.

Although I have to say I initially discounted the reports late last week because, based on current chip developments, there seemed little advantage in a difficult architecture change. With the full picture revealed in the Keynote as to the long term advantages and the minimal discomfort for developers, it seems like a reasonable move that will change very little except give us faster macs in the future.

How could we have any problem with that?

[Added] Good FAQ from Giles Turnbull at O’Reilly’s Developer Weblog

5 replies on “Don’t panic! Apple adopts Intel processors”

Leave a Comment
  1. Unfortunately, it’s not quite as simple as you make it out to be. I’d just like to address a few things:

    For most applications a simple recompile in the Xcode developer environment will suffice – a matter of a few hours work at most.

    This just isn’t true for a large number of applications. And of course, it’s mostly the important ones. For example, one application I wrote for a customer was written using a framework based on Carbon. I wrote it shortly before the company that makes the framework announced that it will no longer be updated. Therefore, that application will have to be completely rewritten for an x86 Mac. I don’t know how my customer will feel about the cost of that.

    On the video production side, right now all AE host apps that are shipping are still using the CFM binary format (for plugins, anyway). There will be no CFM on Intel, so your existing plugins will need to be rewritten for Mach-O, at the least, and of course, ported to Intel, as well. Neither of those is a huge problem for developers, but it means time that’s not being spent fixing bugs, writing cool new plugins, etc. And it will likely mean upgrade fees for users, as well as some of your favorite plugins just no longer being available, if the vendors decide not to make the jump, or leave specific products behind.

    Another consideration is the move to using the GPU over the CPU.

    This isn’t a realistic option for lots of types of processing. First, it’s main benefit is for data that needs to be displayed. Getting data back off the GPU is extremely slow. Slow to the point that you’re better off doing the calculations on the CPU in the first place. Then there’s the issue of precision. Currently, some GPUs only use 24-bit floats, and some don’t support full IEEE compliance, so you don’t get the same results you get with CPUs. There’s also the issue of reproducibility. An ATI GPU may actually produce slightly different results than an NVidia GPU, for example. It may or may not matter for any particular application. Regardless, even if developers do want to use the GPU it’s a change they have to start making now, so it’s no better of an option than rewriting for Intel, I suppose.

    On the plus side, maybe this will make it easier for video card vendors to write their Mac drivers, and maybe we’ll start seeing some high end video cards on the Mac. I can dream, can’t I?

  2. Great to get some actual developer feedback Darrin. It seems that those who will have the least problems are the developers working only on the modern code base – all Cocoa all the time :). Makes me more concerned over Final Cut Pro, although I believe they dropped the CFM for Mach-O at version 4.

    Philip

  3. My main concern is that the G5 has some life left in it, because we could be waiting 2 years for the high end to go Intel.
    Little speed bumps do not hack it. It needs something like the core technology to keep the faith and I suspect Apple will not put the resources behind it.

  4. Great article Philip…and great feedback from you too Darrin! Philip – I think you made an excellent point that a lot of people are overlooking – the fact that Apple made this move to get faster processors 1, 2, 5 years from now, not tomorrow. There are a lot of people making the point that Apple’s own site says PowerPC is a lot faster than Pentium for certain tasks (which it is), but in 2 years from now, it won’t be.

    I’m mad at IBM for dropping the ball and not following through with an architecture that had a lot of promise for us. And while I worry about how painful a transition it is going to be (which upgrade will work with which box natively, and have plugins, etc), I really respect Apple for making the move when it was appearent that IBM wasn’t going to get where Apple needed them to be. I also commend Apple for keeping a (no doubt expensive) backup plan by developing all flavors of OS X and supposedly their apps for X86. That move will hopefully help make the transition a less painful one.

    Hopefully Apple will be able to get and take advantage of the latest (and future) processors from Intel, with 64 bit and dual core. This may also put to rest which OS is best (from a speed standpoint, because we all know which one works and looks the best!). When the OS’s are running on the same processor with like architecture and hardware around it, you know there are going to be direct comparisons. I’m anxious to see that day!

Comments are closed.
Send a Message