Will Apple Switch the Mac to ARM? Why the Rumors Do — And Don’t — Ring True.
That’s one cynical way to interpret the story from Bloomberg News saying that Apple is exploring ways to move its Macintosh line of computers away from Intel’s chips and toward using its own internally designed line of chips.
Apple’s last shift in chip technology happened during 2005 and 2006, when it pivoted away from the old IBM-made PowerPC architecture and instead embraced Intel’s processors, which already run inside most of the world’s personal computers.
One side benefit that resulted was that Macs soon had the ability to optionally run Microsoft Windows and other operating systems, too. One of the most popular software products for the Mac is Parallels, a virtualization program that allows users to install and run Windows side by side on the Mac.
Bloomberg’s story says that Apple engineers have “grown confident” that its own line of chips — the current top of the line is the A6X inside the newest iPad — will eventually have enough computing muscle to run a full-featured Mac, and not just an iPad or iPhone.
Such a change would no doubt hurt Intel, already fighting to maintain its spot as one of the tech industry’s agenda-setting companies, as the PC market contracts and its lack of participation in the mobile market becomes ever more glaring.
The thinking goes that, in time, Apple will want to offer a more unified computing environment across all of its platforms — phones, tablets and PCs — and one key way to make that happen is to have a single chip architecture inside them all.
It isn’t crazy, and you just know that somewhere in some lab in Cupertino or Austin, there is a hopped-up prototype Mac running some weird iteration of OS X on some hopped-up prototype A-chip, just to see if it can be done. As the late Steve Jobs once said about the prospect of switching to Intel, but before it happened: “We like to have options.”
Certain pieces of the puzzle are in place; others have yet to be properly put in place, none of them impossible. Probably the most important one was the introduction by ARM — the British chip designer whose cores form the basic designs around which Apple’s A5, A6 and A6X chips are built — of the Cortex A57 and A53 cores last month.
These are the first 64-bit ARM cores ever, and being 64-bit capable is a must for a Mac. Why? Memory. A 64-bit chip can address a lot of memory, much more than an older 32-bit chip can. Take the base configuration of a MacBook Pro. It ships with eight gigabytes of DRAM memory on board and, depending on the model, can be expanded to 16GB. The iMac maxes out at 32GB of RAM. The muscular Mac Pros can in some configurations take up to 6GB of RAM. All that RAM requires a 64-bit chip, and before last month, a chip based on an ARM core couldn’t get there. Now it can.
The fundamental problem will be one of performance. Intel’s history of chip designs have always tended to emphasize boosting the overall computing power of a chip, and it has done this better than anyone else. There’s a reason that Intel chips are found in most of the world’s PCs servers.
And while I’m painting this in broad brushstrokes, this emphasis on computing muscle has, over the years, caused Intel to lag ARM-based chips when it comes to power efficiency. In the same way there’s a reason that Intel chips are inside most of the worlds PCs and servers, there’s a reason that ARM-based chips, built by companies like Qualcomm, Nvidia, Texas Instruments, and now Apple, are inside most of the world’s mobile phones, smartphones and tablets.
So it would seem that the state of play right now has the best chip going where it’s best suited. Intel chips go in Macs, and ARM-based A5, A6 and A6X chips go in the iPad and the iPhone. Each is the best tool for the task at hand.
But could ARM designs catch up with Intel enough that an ARM chip could be as good — or better, as it would have to be — to knock Intel out of the Mac, as the Bloomberg story suggests? They could. I talked with analyst Nathan Brookwood about this. He said that while ARM chips generally don’t match Intel’s on performance right now, Apple has the in-house expertise to design one that could get there. It would take a few years, but it could be done. “There’s no reason that an Intel chip couldn’t arbitrarily be made to have the same power efficiency as an ARM chip. There’s also no reason that an ARM chip couldn’t be faster, with the right hardware resources brought to bear,” he told me. “It’s all a matter of implementation.”
If a company decided it wanted to design an ARM chip that was, as Brookwood put it, “hell-bent on performance,” it could be done. “You could get a pretty fast machine,” he says.
Trouble is, it would have to be not only be fast, but have a really excellent road map, lasting well into the future, that not only met but exceeded that of Intel. That’s a tall, tall order.
And there would, of course, be numerous complications, none of them insurmountable.
Remember that when Apple shifted from PowerPC to Intel, it had to provide a cushion to all those software developers. Software written for the old chip had to work on the new, and vice versa. This was done primarily through emulating the old chip on the new. Apple called the technology Rosetta.
It turns out that there’s a path for a new Rosetta-like technology on ARM. A Russian start-up company called Elbrus Technologies has developed emulation technology that allows an ARM chip to run software developed for Intel’s x86 chips. (Read about it here, at EETimes.) Advantage ARM.
What about Intel? It has advantages, too, none of them trivial. One is the best record in the world of consistently delivering chips that outrun every other chip in the world, when it comes to raw performance. It also has the world’s best chip-manufacturing technology and expertise. And it has the benefit of a fruitful, seven-year relationship with Apple.
It’s a toss-up for now. If the environment shifts in such a way that ARM designs are bolstered by Apple’s own considerable and growing chip-design expertise, and its desire to rely less and less on outside partners to control its destiny, Apple could indeed move the Mac to an ARM-based chip of its own design.
If it does happen, it will not happen quickly. It would take at least two years, and then, if history is any guide, the transitional phase during which software developers and customers alike would have to have their hands held would last another two years. It is something that Apple does well.
Another thing that Apple does well: The precise opposite of what some rumors suggest it will. The story is properly hedged with the classic “to be sure” paragraph that says no final decision has been made on any of this. The speculation will grind on via the rumor mills. And until there’s more evidence on the table that this transition is happening, my advice is to treat it as nothing more than that: A rumor.