Bombs Away

Wednesday, 8 June 2005

There were two major reasons why I didn’t think Apple would move the Mac to x86 Intel processors:

1. To maintain compatibility with existing Mac software, they’d need a way to run existing PowerPC Mac software on Intel-based Macs at a reasonable speed, and I didn’t think that was possible.

2. They couldn’t just start selling x86-based Macs out of the blue. They’d have to pre-announce them to give developers time to build native x86 Mac software. But once they pre-announced Intel-based Macs, sales of existing PowerPC Macs would likely tank.

On Sunday morning, I wrote:

The only way this makes any sense is that there’s something else. Something big. Not that CNet and the Journal have the story wrong, but that they only have part of the story — and the part they don’t have is what’s going to knock our socks off.

My two guesses as to what the “something else” could be were (a) that Intel would be making PowerPC chips for Apple; or (b) that Apple had found a way to emulate PowerPC-compiled software at acceptable speed on Intel processors.

I placed my bet on (a), and, obviously, I lost. The answer was (b). But I was right that one or the other had to be the case.

I bet on (a) — Intel-produced PowerPC chips — because I thought it would solve both of the above problems. The correct answer, (b), only solves the first problem. My guess is that it was simply out of the question, completely unrealistic, to expect Intel to produce PowerPC chips; I’m guessing for anyone genuinely familiar with the semi-conductor industry, it was as ludicrous as a proposal for Coca-Cola to start producing Pepsi.

Rosetta — the technology that allows existing PowerPC software to “just work” on Intel-based Macs — is the missing link that makes this transition possible. “Emulator” is perhaps not quite an apt description; Apple seems to prefer the term “translator”. The specific description I’ve heard is that it is “dynamic binary software translation”. I’m curious to know more about how it works, but the only important questions are whether — as it was described in a slide during the keynote — it’s “Fast (enough)”, and how many important apps run under it. We should find out soon enough, when benchmarks start leaking from seeded developers. (Their NDA forbids publishing benchmarks based on the developer transition kit hardware, but come on, you know they’re going to leak anonymously.)

Here’s the overview of how Rosetta works from Apple’s “Universal Binary Programming Guidelines” documentation:

When an application launches on a Macintosh using an Intel microprocessor, the kernel detects whether the application has a native binary. If the binary is not native, the kernel launches the binary using Rosetta. If the application is one of those that can be translated, it launches and runs, although not as fast as it would if run as a native binary. Behind the scenes, Rosetta translates and executes the PowerPC binary code.

Rosetta runs in the same thread of control as the application. When Rosetta starts an application, it translates a block of application code and executes that block. As Rosetta encounters a call to a routine that it has not yet translated, it translates the needed routine and continues the execution. The result is a smooth and continual transitioning between translation and execution. In essence, Rosetta and your application work together in a kind of symbiotic relationship.

Rosetta optimizes translated code to deliver the best possible performance on the nonnative architecture. It uses a large translation buffer, and it caches code for reuse. Code that gets reused repeatedly in your application benefits the most because it needs to be translated only once. The system uses the cached translation, which is faster than translating the code again.

Rosetta is not just a side note or afterthought; it is an essential component in the transition strategy. It is not going to be marketed with trumpets and banners. Rosetta’s role is that of the unsung hero.

There’s nothing glamorous about it, because in an ideal world, on the day Apple begins shipping Intel-based Macs, all Mac OS X software will have been updated to run natively on both architectures. But the world is not ideal, and that is not going to happen. A year from now, at WWDC 2006, most Mac OS X applications will be ready to run natively, but not all. (And I can hear it now: those developers who haven’t yet shipped universal binaries will be labeled “laggards” by Steve Jobs during the keynote.)

Don’t forget also, however, that even those apps that are updated will in many cases be available only as paid upgrades. Will there be universal binaries of the Adobe and Microsoft suites? Yes, definitely. Will they be free updates to the current versions? I doubt it.

If you’re holding on to any not-latest-and-greatest software, apps that you perhaps only use occasionally — or something you no longer actively use, but which you keep around for opening old files — you’re going to want those copies of your software to work on an Intel-based Mac.

Unlike the Classic environment — which runs apps in a visual ghetto — most users may not even notice that a few of their apps are running under Rosetta. (Assuming, again, that Rosetta’s performance is adequate.)

Apple won’t trumpet Rosetta because they want to trumpet native apps, and for good reason: apps recompiled for Intel Macs will be faster. It’s similar in many ways to the 68K emulation that Apple delivered in the transition a decade ago: everyone wants native apps for the new architecture, but everyone is glad they can still execute their existing apps when necessary.

Judging from the preliminary documentation, however, Rosetta is far less comprehensive than the 68K emulator. The 68K emulator could pretty much emulate any software — apps, drivers, control panels. Rosetta is limited only to applications, and only certain ones at that.

There’s a wee bit of uninformed hysteria regarding the fact that Rosetta effectively emulates only a G3 processor; the Rosetta documentation states it won’t run “code written specifically for AltiVec” or “applications that require a G4 or G5 processor”.

But in most cases, apps that are capable of taking advantage of AltiVec, fall back to non-AltiVec code branches when running on G3 processors. Exhibit A: Photoshop; if you have a machine with AltiVec, it uses it, if you don’t it continues to work just fine.

One potential hiccup is that Rosetta is an all-or-nothing affair for each application. Again quoting from the Rosetta documentation:

Rosetta must run the entire process when it translates. This has implications for applications that use third-party plug-ins or any other component that must be loaded at the time your application launches. All parts (application, plug-ins, or other components needed at launch time) must run either non-natively or natively. For example, if your application has both an x86 binary and a PowerPC binary, but it uses a plug-in that has only a PowerPC binary, then your application needs to run non-natively on a Macintosh using an Intel microprocessor in order to use the non-native plug in.

The Osborne Effect

So, the first reason I didn’t think this would happen was an issue of could — I didn’t think Apple could execute existing PowerPC code on x86 machines with reasonable performance. And I’m happy to say it looks like they’ve proved me wrong.

The second reason was an issue of would — I didn’t think Apple would be willing to pre-announce a dramatic shift like this, knowing that it would immediately detract from sales of existing PowerPC-based Mac hardware.

And I was wrong on that point as well. Apple is marching into this transition chin-first, and they’re simply going to take their lumps.

The question is not whether sales are going to be hurt; the question is how badly. Especially as we get closer to the release of Intel-based Macs next year, sales are going to drop. I’ve already heard from numerous people claiming they’re delaying planned hardware upgrades until the new Intel-based machines arrive. It doesn’t matter if such reactions are irrational or emotional or uninformed — what matters is that there exists X number of people who would have purchased new Mac hardware in the coming months, but who instead are now planning to wait for next year’s Intel-based Macs.

Apple’s attitude is clear: we’ll take a couple of quarters of weak sales now, and make up for it next year when the new systems ship. With growing iPod revenue, it’s entirely possible that Mac sales could take a complete nosedive for the remainder of 2005 and Apple could remain profitable. (It’s also possible that we will see some outstanding price cuts on existing product lines in the coming months.)

This transition period is going to be hard on Apple, not hard on Mac users. If you’ve just purchased a machine recently, or need to buy one soon, you’re no worse off than you would have been if Apple had remained committed to producing new PowerPC hardware — today’s machines would have been obsoleted by even-better machines next year no matter what processors they contained. And no one is abandoning PowerPC software development. I see no reason to expect Intel-only Mac software in the near future. Universal Binaries take full advantage of both Intel and PowerPC Macs. That’s worth repeating: Universal Binaries take full advantage of both Intel and PowerPC Macs.

The “Osborne Effect” is named after the Osborne Computer Corporation, who had a successful personal computer in the early 1980s called the Osborne 1. Founder Adam Osborne began hyping their next-generation machine before it was built, and his hype was so effective that customers stopped buying Osborne 1s in anticipation. They went bankrupt and never finished the project.

Nothing so spectacular is going to happen here. Even in the worst-case scenario for Mac sales in the next 12 months, Apple is in no risk of going bankrupt.

Why Switch?

This announcement has caught both Apple’s customers and the rest of the industry by surprise. Especially given the fact that sales of PowerPC-based Mac hardware have never been stronger than they are now. Apple’s 43 percent increase in quarterly sales from a year ago is simply outstanding.

As trite as it sounds, I’m inclined to believe Jobs’s explanation during the keynote, that Apple believes it can make better computers down the road with Intel processors, and that the difference is enough to justify this painful transition.

Maybe the PowerPC roadmap looks good, but the Intel roadmap looks better. Or maybe they see the PowerPC roadmap as downright bleak — e.g., say, no G5 PowerBooks in the foreseeable future, even if Apple wanted to stick with the PowerPC.

The thinking seems to be: better to act now, from a position of strength, and absorb the costs of the transition while the company is doing well, rather than wait a few years and risk being forced to act from a position of weakness or desperation.

“I stood up here two years ago in front of you, and I promised you this,” Jobs said during the keynote, in front of a slide picturing a 3.0 GHz PowerMac G5. “And we haven’t been able to deliver that to you yet.” He went on to say that they’d like to be offering G5 PowerBooks, but can’t.

I’ve seen some interpret this as petulance or spite — that this switch is just Jobs picking up his ball and leaving for another playground because he feels IBM has embarrassed him. That interpretation is foolish. I really think Jobs was just being honest, or at least as honest as he could be in a public statement.

I think it boiled to a choice between two difficult options: either initiate a painful and expensive transition to Intel processors, or stick with PowerPC and fall behind.