Intel Corp (INTC)

[RobertG]

There will be blood: why Apple and Intel are destined to clash

By Jon Stokes | Last updated 33 minutes ago

http://arstechnica.com/apple/news/2010/07/there-will-be-blood-why-apple-and-intel-are-doomed-to-clash.ars?utm_source=rss&utm_medium=rss&utm_campaign=rss

A long time ago, in what seems like a galaxy far, far away, Apple and Google were the Brad and Jennifer of the tech world. Google had the cloud, Apple had the clients, and both companies were such industry darlings that fantasizing about world-changing Apple/Google team-ups was a popular pundit pasttime. The love extended all the way up to the top of Apple, with Google CEO Eric Schmidt taking the stage alongside Jobs at WWDC 2007 for the iPhone's unveiling. Jobs was supposedly so pleased with the prospects for Apple/Google collaboration that when he learned of Google's secret "Gphone" project (Android), he reportedly hit the roof.

Given how much engineering effort the company has focused on its iOS-based products—to the point where most of the CPUs in Mac hardware have fallen a full generation behind—it's hard to imagine that Jobs isn't feeling similarly betrayed by yet another Apple partner's full-court press into the smartphone market. I'm talking, of course, about Intel, which just hired former Apple and Palm VP Mike Bell to head up its smartphone efforts, and which is rumored to be contemplating a purchase of key baseband chipmaker Infineon. Intel is jumping into the smartphone market with both feet, and the company's goal isn't just to create a direct iPhone competitor—no, Intel wants to empower a whole ecosystem of iPhone competitors based on x86.

Is it a coincidence, then, that Apple doesn't appear the least bit inclined to bring its Mac line up-to-date with the latest 32nm parts from Intel? No, it isn't. But the story of the present and future of Apple's relationship with Intel is considerably more complicated than "Betrayed Apple Flees Intel for AMD."

Lagging on Intel, watching AMD
According to the Mac Buyer's Guide, the Mac Pro averages 236 days between refreshes; but it has been 510 days since it was last updated. The iMac was last updated 279 days ago; its average is 221. (Rumor is that updates are coming Tuesday, which, if true, will be just 19 days after I bought a 27" Core i7 iMac.) On the portable side, the Macbook Air has gone 413 days without an update, when its average is 255 days. Finally, the Mac mini was recently updated within a normal time interval, but it was a cosmetic update only—the machine kept the positively geriatric Core 2 Duo. All told, Apple looks set to skip Intel's 32nm Westmere generation almost entirely, and this was after the company appeared reluctant to upgrade to Nehalem. There are probably a few reasons why Apple has been slow to refresh its Mac line.

First, there's the fairly remote possibility that Apple could be seriously considering AMD. Intel's upcoming Sandy Bridge will, by all accounts, be an incredibly strong family of processors, so it's hard to imagine that Apple will jump ship for AMD at this point. But Bulldozer is a truly novel architecture in many respects, and with all such attempts at radical (as opposed to incremental) improvements, it's likely to either really rock or really flop. If the project works and AMD can deliver a cheap, high-performance Sandy Bridge alternative that doesn't waste any die space on an integrated GPU, then Bulldozer would be a great option for Apple. The company would be free again to choose between ATI and NVIDIA discrete GPUs, without having to engineer around Intel's IGP.

So while we find it unlikely that Apple will move to AMD on the desktop, a combination of a successful Bulldozer launch from AMD and a string of high-profile Intel smartphone wins would put such a platform shift well into the non-crazy category.

As for other reasons why Apple is apparently skipping Westmere, we can offer two more possibilities. One possibility is related to the fact that Intel's desktop Westmere parts are an evolutionary step backwards from a system architecture point of view. We've covered this elsewhere, but to recap: Arrandale and Clarkdale move the memory controller off of the processor die and return it to the northbridge, where it lived prior to Nehalem. This move boosts memory latency, which is not good, but the real killer is that Westmere also puts the northbridge, which contains Intel's ho-hum integrated graphics processor, into the same package as the processor die. Apple can't be pleased with having to engineer around Intel's IGP in the MacBook, and the prospect of doing this on the desktop has to be particularly unappealing. It's also probably not possible to do in the MacBook Air due to space constraints.

Sandy Bridge moves the IGP and memory controller back onto the processor die, which should at least boost performance by lowering memory latency again, even if you do still have to deal with the IGP issue.

The other possible reason that the Mac line is in such sorry shape is that Apple just doesn't care very much about the PC anymore. Jobs has compared the desktop PC to a pickup truck—a utility-oriented niche that will remain popular but will nonetheless be a niche. It was clear from Apple's press conference on the iPhone antenna issue that Apple's engineering focus is now squarely on its (non-niche) iOS products, which represent the future of the company. Sure, Macs are selling well, and Apple can still steal desktop and laptop PC marketshare from the Windows crowd. But the desktop PC market as a whole is mature and commodified, and it simply can't compare to the booming mobile market as a source of the kind of revenue growth that can justify Apple's stratospheric stock price.

This is especially true in a challenging economic environment where consumers and businesses are lengthening their hardware upgrade cycles, and where the real growth story for commodity PC hardware seems to be not on the desktop but in the data center. Outside of a few important niches—none of which are large enough by comparison to the iPhone and iPad markets to be worth devoting any real engineering effort to—the desktop PC is largely an Internet client.

So the Mac will remain with Apple as a legacy product that still throws off a great revenue stream in return for a minimal amount of investment, but the Mac does not represent the kind of explosive, long-term growth potential that can justify the company's position at the very top of the technology market capitalization heap. Or, to put it another way, Apple will be happy to take PC reference designs from Intel (or AMD?) and repackage them in nice mobile and desktop enclosures—effectively outsourcing PC R&D to the chipmaker—while focusing its own engineering efforts on differentiating its post-PC products from the rest of the market... a market that will soon include Intel-made smartphone reference designs.

The reference experience
You may wonder just how serious of a threat Intel is to Apple's smartphone interests. Depending on how you define "serious threat," the answer can go a number of ways. If by "serious" we mean that Intel is serious about the smartphone market, then the threat is definitely very serious.

At the most recent Intel R&D day, Intel CTO Justin Ratner made reference to "smartphone activity" in Intel's labs, activity that he said the company will soon talk about more publicly. These remarks were in the context of his answer to our question about possible blowback from Intel going into competition with its customers by designing full-blown, market-ready products.

When we suggested that in moving up the stack from hardware and into software, services, and complete products, Intel might alienate some of its key customers, Ratner answered by reassuring us that Intel's product teams design only "reference experiences." Because Intel's products, he explained, come in reference form, a company can either adopt the reference design wholesale by just rebranding it and going straight to market, or it could replace parts of the design with custom engineering.

This reference talk is probably not at all comforting to Apple, because with Intel's smartphone efforts, Ratner is essentially saying, "we plan to design a complete reference smartphone package and then offer that to every single one of Apple's competitors, who will either go straight to market with it or will improve on it and then go to market."

The next question, then, is whether Intel's smartphone effort can seriously threaten the iPhone. This is unlikely, at least in the near to medium term.

It certainly will be possible for Intel to squeeze its Moorestown platform into a smartphone form factor, especially since that form factor now includes jumbo devices like the HTC EVO 4G, but it's also possible for someone to train a dog to walk on its hind legs—a neat trick, but a lot of effort to achieve something unnatural and suboptimal.

With its own newly acquired in-house ARM expertise, Apple should continue handily cleaning Intel's clock in smartphones until some combination of architectural tricks and performance advantages eliminates enough of Atom's performance per watt disadvantage to make it a competitor. That day may come, but it won't come this year or next.

Trouble ahead
Apple will probably adopt Sandy Bridge, and Intel will keep plugging away at its effort to empower a whole new ecosystem of iPhone competitors, so there's unlikely to be any change in the status quo this year. But the writing is on the wall for the Apple-Intel relationship: this will end badly, just like the Apple-Google romance of 2006-2007. There's no way that Apple is going to sit idly by while Intel builds a hardware/software platform to compete with iOS, and then gives that platform away to Apple's competitors. And there's no way that Intel is going sit idly by and watch Apple's ARM-based products rake in billions in revenue without trying to take them down.

It seems likely that if AMD can get its act together, stay solvent, and ship a family of desktop and server processors that are competitive with Intel, then Apple may well break up with Intel entirely. At this point we won't make an outright prediction of an Apple-AMD shift, because the smaller chipmaker's upcoming server and mobile products will either make it or break it—if Bulldozer in particular flops or faces heavy delays, there won't be an AMD for Apple to migrate to. This is a tall order for AMD, but it's doable, and given the lay of the land with Apple and Intel it's no stretch of the imagination that Apple is rooting for AMD to come through. And if AMD does, then the odds of Apple moving to the rival chipmaker are somewhere north of 50 percent.

(Note: In acknowledgement of the fact that some of the response to this article will focus on the possibility that Apple might buy AMD, let me say upfront that I find this very unlikely under any set of circumstances. As I said above, the desktop PC market is mature, and it makes very little sense for anyone at all to buy AMD in order to keep producing a line of desktop PCs, even if that line is still growing in units shipped. Only someone interested in the much more lucrative and rapidly expanding data center market would be interested in AMD, and Apple doesn't have a real data center product. Acquisitions are about enabling growth in a hot new market, and not about sustaining revenue in a mature one. The ARM acquisitions that Apple has made recently are about the former, and an AMD acquisition would be about the latter.)