Apple Switches

by Kieran Healy on June 6, 2005

Steve Jobs “announced this morning”:http://www.apple.com/pr/library/2005/jun/06intel.html that Apple will ditch the IBM PowerPC processor and begin using Intel chips in its computers as of next year.

We pause for a moment to allow Mac users to digest that sentence.

It looks like IBM was unable or unwilling to make the kind of investment in the G5 chip that Apple needs. IBM is moving out of the commodity PC market anyway, and supplying Apple with chips accounts for only a tiny percentage (I think about 2 percent) of its total chip sales, most of which are in embedded systems and the like. So Apple has, it seems, been forced to jump to Intel. It’s been rumored for years that Apple have maintained an x86 build of OS X for years, and today Steve Jobs confirmed this was true. Note that — as far as I can tell — there is no prospect of you being able to buy OS X and install it on your Dell: Apple will still use proprietary hardware and while hackers might be able to get a system running, regular users will not. The company suffered too much in the “clone wars” of the 1990s, when it was possible to buy computers that ran Mac OS from companies other than Apple. Still, this is an enormous shift, one that Apple will have to work hard to sell to developers and users alike.

For developers, the question is how easy it will be to recompile code written for one architecture in order for it to run on the other. (The platforms differ in their “endianness”:http://en.wikipedia.org/wiki/Endianness, the convention about byte-ordering in code which is part of the reason applications compiled for x86 architectures will not run on PowerPC systems, and vice versa.) Apple says that stuff developed in their X-Code framework should be easy to recompile. Other developers may have to work harder. Apple’s goal is to have developers release their applications as “fat binaries”:http://en.wikipedia.org/wiki/Fat_binary, which will run on both kinds of chip. This way, end-users won’t be confused by having to choose which version of an application to download based on the chipset inside their computer — this is precisely the sort of decision that people buy Macs in order to avoid having to make. Apple says it also has a piece of software called “Rosetta” that allows applications written for the PowerPC to be emulated on Intel chips.

It seems the next version of Mac OS X (v10.5) will be called “Leopard,” so I am sure the marketers are already gearing up with change-your-spots slogans as I write. Personally, I hope Apple negotiates the transition successfully: Mac OS X is a very civilized operating system that’s simply a joy to use and doesn’t get in the way of your work. But it could be a tough road, as the incentive to buy an Apple computer in the next year (before the transition) is now very much reduced, and the likelihood of a wave of “FUD”:http://en.wikipedia.org/wiki/FUD very much increased. (I’m not sure that it’s rational to think that the incentive to buy has gone down: after all, most computer hardware is “obsolete” in six months anyway, if all you care about is the bleeding-edge of perfomance.) I imagine the key will be to find a way to make the transition as contained as possible, with the result that Apple keeps the same niche in the market that it presently occupies. It’s a smallish company that sells computer hardware on the strength of elegant industrial and software design. It’s not going to become Microsoft (why even try?), and the path-dependence in the marketplace means that it’s just a fantasy to believe that there are millions of irritated Windows users just waiting for the right moment to jump to Apple. People can stay irritated their whole lives, in my experience, and not change anything much about their situation.

Apple is a cash-rich company at the moment (thanks to the iPod), so perhaps this will help them over the hump. They’ve done something like this twice before, and brought most of their users along with them. I really, really don’t want to go back to using Windows, so I suppose I’ll be following along with Intel — er, I mean, interest.

{ 37 comments }

1

Steve LaBonne 06.06.05 at 2:19 pm

Baby, it’s cold outside. -Satan

2

Yusuf Smith 06.06.05 at 2:31 pm

Endianness isn’t really the reason PowerPC apps won’t run straight on Intel chips. They are totally different processors; one uses a complex instruction set (CISC, ie. the Intel range) and the other uses a reduced set (RISC, the PowerPC).

Most applications will recompile, however, because they are already based on technology that already exists for both platforms, like Unix, the GNU code libraries, and so on. MS Office for Mac, for example, is based on an Apple code library called Carbon. MS can (if they so wish) recompile Office for Carbon on Intel and it will run.

I find this whole switch puzzling – Apple have been saying for years that megahertz isn’t all that matters in getting performance out of a chip. Also, it’ll cast a serious shadow over the whole Mac platform at least until they bring out the Intel-based machines.

3

Kieran Healy 06.06.05 at 2:37 pm

I find this whole switch puzzling – Apple have been saying for years that megahertz isn’t all that matters in getting performance out of a chip. Also, it’ll cast a serious shadow over the whole Mac platform at least until they bring out the Intel-based machines.

It seems like they were driven to it by IBM, who don’t want to invest in making a version of the G5 that runs cool enough to put in a laptop, for instance, and the relatively small volume of Apple’s purchasing doesn’t give them sufficient leverage to force IBM to change their mind.

4

urizon 06.06.05 at 2:38 pm

Apple switches? I thought we were talking about corporal punishment, for a minute.

5

nick 06.06.05 at 2:42 pm

I find this whole switch puzzling – Apple have been saying for years that megahertz isn’t all that matters in getting performance out of a chip.

And this isn’t really about MHz; it’s about power consumption and cooling. In a desktop setup, you can bung a couple of G5s together and it flies. But in a Mac mini, or most of all, a laptop, IBM hasn’t delivered on its roadmap for a mobile G5 that runs cool and doesn’t drain the battery. And since the small-form models (laptops/minis) are the big sellers…

6

msf 06.06.05 at 3:01 pm

The OS won’t be a difficult port. The apps and device drivers will be more difficult. The OS is already availble for PCs (with a fair number of hardware restrictions) for free as OpenDarwin. MacOS is based on a FreeBSD kernel which runs on PCs. The issues will be primarily device support; but Apples have very limited device support at any rate due to the closed architecture.

7

Kieran Healy 06.06.05 at 3:04 pm

The issues will be primarily device support; but Apples have very limited device support

Why should this be an issue, though? Like I said, I don’t think their goal is to make the OS available for installation on any old commodity PC. Instead, they want you to keep buying Apples, just as before. So they can pick and choose what to support, as they always have.

8

theCoach 06.06.05 at 3:11 pm

Kieran,
It seems worth pointing out that the next Windows and Apple Operating systems will change in the future and that a better position might be to go with the Operating systems that will help you be the most productive. If it is a Windows system, go with that. Presumably, Microsoft will have a harder time making you more productive (or total benefit from your computer, however you want to measure it) because you are familiar with their software, but you should keep an open mind going forward and make judgements based on the products rather than the history of who makes them.

9

Andy 06.06.05 at 3:32 pm

AltiVec code is going to be the painful thing to re-write. Oh and yusef, the CISC versus RISC thing is tired. All modern x86 CPUs have reduced instruction sets. Check out some of the CPU articles at Ars Technica if you want. It’s not 1995 anymore.

10

John Quiggin 06.06.05 at 4:04 pm

This is going to be painful, no doubt. A possible silver lining is that it ought to be much easier to run Windows in parallel. I currently use the Virtual PC emulator which is impressive, considering, but still problematic. It may also reduce the difficulty of porting Windows software to Mac.

Can anyone point to info on these issues?

11

des von bladet 06.06.05 at 4:30 pm

Yeah well, like I said then (I was and am and probably will be so very prescient) Jobs II always was NeXT taking over Apple, isn’t it?

12

Precision Blogger 06.06.05 at 4:36 pm

Switching Endianness can be messy. Some programs will compile but mishandle pairs or quads of bytes. Stay tuned for some confusion…
– The Precision Blogger
http://precision-blogging.blogspot.com

13

Ben Alpers 06.06.05 at 4:47 pm

My old Pismo G3 Powerbook is coughing and wheezing. I was planning to purchase a new Powerbook this summmer. I was disappointed that I’d have to go with a G4, but it’s been clear for months that G5 Powerbooks were a long way away.

So here’s the question: should today’s announcement in any way change my decision? As I saw it, I was buying a G4 Powerbook, figuring that in a year or so, there’d probably be a G5 Powerbook that would put my computer to shame. Now there will probably be an Intel-based (G6?) Powerbook (assuming they’re keeping the “Power” name-set) in about the same timeframe that will similarly put my computer to shame. So as far as I can tell, I should just go ahead and do what I would have otherwise done.

Do others around here agree?

14

des von bladet 06.06.05 at 4:57 pm

Ben: Just do it, always. They’ll always be a newer spiffier one if you wait, same as there always has been, but you want one sooner.

Don’t get Brian Weatherson and DSquared bickering about angels, envelopes and infinity again, especially not if you don’t have forever to live.

15

Kenneth Chiu 06.06.05 at 5:22 pm

Whether or not endianness will make it difficult to recompile apps depends in large part on the quality of their source code. Good practice calls for not putting such dependencies in your source. For example, if you are sending a 4-byte integer over the network, you first convert it to a standard order using the htonl() function (host-to-network). The htonl() function is hardware-dependent, but is supplied by the development environment. So no changes need to be made to the application source code.

16

Barry 06.06.05 at 5:34 pm

Besides, if you get one of the last ‘genuiinnne powerPeeCee Macs, the last reeeeallll Apple’, you might be able to sell it for a decent price a few years down the road.

17

Keith M Ellis 06.06.05 at 5:35 pm

“Oh and yusef, the CISC versus RISC thing is tired.”

Really, there’s been a convergence. The RISC chips got more CISC-like and the CISC chips got more RISC-like.

18

chris lovell 06.06.05 at 5:59 pm

ben, I’m pretty happy with my Pismo. I’m mostly text-based, but I run Photoshop every so often and the performance seems fine. For most users I think the “need” to upgade is overhyped. Sure, if I did digital video work I’d be upgrading regularly, but I’m not replacing this thing until A)th l ogic board dies or B) a new job pays for a new powerbook.

19

Kenneth Chiu 06.06.05 at 6:02 pm

“Really, there’s been a convergence. The RISC chips got more CISC-like and the CISC chips got more RISC-like.”

I believe Yusuf’s main point is that even if they had the same endiannes, applications compiled for PPC won’t run on x86 simply because they have completely different instruction set architectures (ISAs). (Unless of course, some kind of emulation or binary translation software is used.)

It’s like trying to take a printer cartridge made for a particular manufacturer and trying to use it in a printer made by a different manufacturer. Unless they’ve taken special care to be compatible, it just ain’t gonna work.

20

PZ Myers 06.06.05 at 6:52 pm

I’m a Mac fanatic, and this decision doesn’t perturb me in the slightest. It’s never been about what CPU was under the hood, it’s all about the software. As long as I can continue to use Mac OS X rather than that clumsy POS called Windows, I won’t care if it’s on a PPC or a Pentium.

21

Jon H 06.06.05 at 7:46 pm

It shouldn’t be that big of a deal.

NeXT, after all, ported its operating system from the Motorola 68040 to Intel, Sun Sparc, and HP PA-RISC. And developers could build fat binaries supporting all four architectures by clicking some checkboxes and clicking “build”.

And that was a wee small company of about 250 people, porting to hardware they didn’t control. Then Apple bought NeXT and turned the operating system into OS X.

By comparison, it ought to be easy as pie for Apple, since they’re using basically a later, upgraded version of the same technology.

It’s not so much that Apple is moving from PowerPC to Intel, but rather that they moved from Intel to PowerPC (and Intel, in the shadows), and are now moving back to an Intel focus.

22

Jon H 06.06.05 at 7:53 pm

What will be interesting is that, after this transition, Mac software (from Apple and from third parties) ought to be pretty loosely tied to the hardware.

So if Apple needs to change *again*, it should be easier.

Theoretically, Apple could even adopt different CPU architectures for different parts of the product line.

If IBM pulls their thumb out and produces a kick-ass processor for graphics workstations that beats Intel, Apple could put that in a Mac desktop, while using Intel for the others.

Economically, that might only work for high-performance, high-margin, high-price, low-volume machines, but it’s something they’ll be able to do.

23

teece 06.06.05 at 8:03 pm

***usuf’s main point is that even if they had the same endiannes, applications compiled for PPC won’t run on x86 simply because they have completely different instruction set architectures (ISAs).***

Um, no. I have C and C++ source right now that compiles on any architecture that supports those languages. Indeed, the whole point of a compiler is to eliminate the need to give a hoot about the underlying instruction set. Take a look at the Linux kernel — the amount of it written in assembly, which is the only part that cares about the ISA or CPU, is very small. It all compiles just fine on any platform, with the exception of the little bits of assembly: tweaking for performance is all that is needed by architecture.

A kernel is about as CPU-dependent as it gets. And the Apple kernel already runs on x86. Hell, the whole OS already runs on x86 according to Jobs.

The only problem is a) badly written code, of which there is a lot, and b) the economics of getting non-Apple code ported.

The fact of the matter is that it’s been known for 15+ years that writing high-level code that cares about the CPU is bad practice unless you have really good reason. It’s a simple recompile to find the parts of your code that are badly written. The only major problem will be system libraries that may not follow you to the new architecture, as it sounds like may happen to Carbon.

Don’t get me wrong, it’s a non-trivial thing to implement and support, but the CISC/RISC G5/Pentium is not an issue at all.

24

Kenneth Chiu 06.06.05 at 8:16 pm

teece,

Note that I said, “…applications compiled for PPC…”. You are talking about the source. Others are talking about the compiled code. Machine language for PPC does not conform to the x86 ISA.

25

Jon H 06.06.05 at 10:56 pm

For what it’s worth, as of *today* Mac developers can download XCode 2.1 from Apple, and compile their applications for Mac and Intel.

26

99 06.06.05 at 11:23 pm

In the discussion about switching (as a former Apple owner who dropped about seven grand on a Centris over its lifespan), no one addresses the software licensing issue. Why should I have to buy a new license for the Adobe Creative Suite (or buy new fonts for that matter) if I decide to move platforms. Shouldn’t I be able to upgrade cross platform? The EULA is about one user, one processor. If I am discharging a machine, isn’t it fair that I can simply transfer my license? I have about five times the investment in software than I do in hardware. I moved to PC’s because when the hardware became problematic (and I work in a cross platform environment — I have three year old Macs ten times as buggy as my seven year old Win 98 laptop) it was cost effective to simply buy a new machine. Whereas I see some good things about OS X, it ain’t that stable (font management? Hello?), and I’m handcuffed by my software ownership. I could do like every other Mac owner out there and get pirated software, but I sold software at one point in my life, and am proud of the fact that it’s been literally years since I wasn’t a fully licensed operator of software.

27

teece 06.06.05 at 11:25 pm

OK, sorry Kenneth, but in that case this whole discussion would make no sense.

CISC/RISC is completely irrelevant. Sun binaries won’t run on PPC chips, and they’re both RISC chips. Hell, Linux x86 binaries won’t run on Windows machines, even though they share the same machine code. If we are talking compiled code, the whole talk of incompatibility woes would make no sense, so I think my misunderstanding was warranted ;-P

Little/Big endianness is tricky, because applications may recompile cleanly, yet give wildly wrong results, if they run at all. That is why it was brought up. But code would still need to be recompiled, and the RISC/CISC difference is moot. And modern Pentium/Athlon chips are already something resembling RISC chips any way today, as alluded to above. They translate the complex x86 instruction set into something simpler to be executed in hardware.

28

Bruce Wilder 06.07.05 at 12:29 am

How do you ask a customer to be the last customer to buy obsolete hardware?

29

Chris Lovell 06.07.05 at 2:02 am

Why, you sell it to him/her at at 95% discount.

Seriously, I’ll be happy to buy Apple PPC machines at greatly reduced prices. Even in the distant future when developers stop releasing OS X PPC binaries, the machines will still work, and they’ll still be able to run most open-source software. Bring me your used G4 laptops…

30

theCoach 06.07.05 at 10:16 am

The question that has to be answered is what happens to the owner of Adobe Photoshop 6 for the Mac?
a) Adobe provides a recompiled version of a now very old version for free or minimal amount.
b) Adobe offers normal (high) upgrade prices.

31

Jon H 06.07.05 at 10:20 am

bruce wilder writes: “How do you ask a customer to be the last customer to buy obsolete hardware?”

Apple will still sell PowerPC machines until some point in 2007. They’ll release the next version of OS X (10.5), for both PowerPC and Intel Macs, in late 2006 or early 2007.

I wouldn’t be surprised if the following version of OS X (10.6), which ought to come out around 2009/2010, also supports at least the later PowerPC Macs.

Considering that Apple can and will release software that supports both kinds of Mac, and isn’t cutting the PowerPC Macs off cold turkey, it seems to me that the PowerPC Macs aren’t going to go obsolete much faster than they otherwise would have. They will probably have at least 5 years of new OS support in store.

32

PZ Myers 06.07.05 at 10:32 am

Yeah, if there’s a G4 PowerBook fire sale, I’d leap to get another. These are phenomenally sweet machines, you know.

Also, speed is less of an issue nowadays. Way, way back in the dark ages, a speed bump from 1MHz to 3MHz was a big deal. Going from a floppy based system to one with a hard drive was a revolution. Now, though, it takes a pretty pathetic geek to think he’s got something over on someone else because his CPU is running at 3GHz rather than 2GHz. I’ve got a 1GHz laptop, and it’s fast enough for everything I do.

Actually, I’ve even got an old 33MHz 68040 laptop tucked away on a shelf somewhere. It’s heavy and klunky, the software feels a bit sluggish, and the display is like looking through a tiny little porthole…but it still can do most of what I like to do on a day-by-day basis. Not that I’d want to trade down, of course!

33

Keith 06.07.05 at 10:47 am

I’m with Chris on this one– I’ll take an obsolete G4/5 over a Pentium machine any day of the week, and twice on Sundays.

34

Patrick 06.07.05 at 11:04 am

“I find this whole switch puzzling – Apple have been saying for years that megahertz isn’t all that matters in getting performance out of a chip”

What? You mean a large corporation may have been twisting the truth in order to sell its products?

Listen up, Mac zealots, Oceania has ALWAYS been at war with Eurasia, and Eastasia has always been our good friend and ally. Get it right.

35

modus potus 06.07.05 at 3:22 pm

What? You mean a large corporation may have been twisting the truth in order to sell its products?

Like Intel? Notice how they aren’t bragging about the clock rate of their newest chips? Seems that even Intel figured out how to get more processing with less megahertz than their Pentium 4’s. The latter was designed primarily to obtain the fastest clock rate and not to perform the most processing for a given clock rate. Intel milked the megahertz myth for as long as they could, but even they are giving it up now, starting with the surprisingly Athlon-like Pentium-M.

36

hank 06.07.05 at 4:06 pm

I suppose the real reason for this is to make DRM easier?

Apple is an entertainment company, not a hardware company. Like Sony.

37

ArC 06.07.05 at 10:31 pm

_They’ve done something like this twice before, and brought most of their users along with them._

The stories I’ve read say each platform shift (680×0 -> PPC, OS 9 -> OSX) has cost Apple a significant chunk of users

Comments on this entry are closed.