Apple inside: the significance of the iPad's A4 chip
Just one day after the unveiling of the iPad, Apple's long-rumored tablet has ruffled feathers and turned heads. With a familiar interface, Apple's momentum, iTunes integration, and a price that the even non-geek can easily fall in love with, it has all the makings of a hit. But deep inside lies something even more revolutionary.
At the heart of the iPad lies a tiny sliver of silicon. A game changer within a game changer. That's Apple's A4, a system-on-a-chip (SoC) that reportedly combines a low-power CPU, a graphics processing unit (GPU), and other hardware, much of which is still confidential. What we do know is that it finds an almost ideal balance between battery life and speed, such that the iPad can animate and zip about at a pace that iPhone 3GS users could only dream of.
History behind the “magic”
The roots of the A4 begin in 2005 with a then three year old startup called PA Semiconductor. At the 2005 Fall Processor Forum, PA Semi outlined a vision for a SoC PowerPC architecture, the G5-derived PWRficient family. As the name implied, the focus was on designing a high-performance multicore PowerPC chip that was ideal for mobile devices.
That very same year, Apple had to deal with what executive Tim Cook called “the mother of all thermal challenges”—putting a G5 processor into a PowerBook. IBM had never intended for the G5 to operate in a machine as thin as a notebook, leaving Apple with nothing but the old PowerPC G4 chips for the remainder of the PowerBook line. With PC notebooks outpacing Apple's own technology, Apple had to look for solutions, and fast.
It was in these dire circumstances that Apple took notice of PA Semi, and the two became close partners in plotting out the future of PowerPC for Mac hardware. Then came Apple's fateful WWDC 2005 announcement, that the switch to Intel was now in progress; future relations between Apple and PA Semi appeared to be severed.
But PA Semi's 150 engineers still delivered on their promises. The sole member of the PWRficient family was released in February 2007, first to the company's closest partners, followed by a wider release later that year. It boasted a very impressive set of specs, including two 64-bit PowerPC G5 cores with two DDR2 memory controllers on a single chip. It ran at 2.0 GHz with an average of 13 watts of power consumption and a peak of 25 watts. Meanwhile, Intel's more traditionally-designed Merom Core 2 Duo LV L7700—the closest competition at the time—could only run at 1.8 GHz with a max of 17 watts.
Early the following year, Apple quietly purchased PA Semi, with many in the tech press speculating that Apple's intent was to put its newly-acquired talent to work on chips made for future mobile devices, such as iPods and iPhones. Then came a long period of silence that has lasted until now.
CPU/GPU hybrids today
Now we have the Apple A4, a brand new design for a SoC produced and owned by Apple, using the same ARM architecture that powers the iPhone. The A4 runs at 1GHz and supposedly helps the iPad achieve a maximum battery life of 10 hours, thanks to its tight placement of circuitry and small form factor. In integrating a CPU with a GPU, it follows in the footsteps of other energy-efficient SoC processors, like Nvidia's Tegra.
Until recently, the CPU and the GPU of a computer were separate pieces of hardware, often found on different boards or positioned far apart on the same motherboard. Through SoCs and more ambitious designs, attempts have been made to fuse the two components together.
Intel's Larrabee project tried to tie a GPU so closely to one of its own future desktop/laptop CPUs that a discrete GPU would become not only obsolete, but impossible to add to this design. Tragedy struck, and Intel missed its benchmarks. Now Larrabee has been reduced to such a degree that it will never see a wide release.
Intel had a another project up its sleeves, an SoC that wedded the Atom CPU to a GPU for LG-branded smart phones, called Moorestown. But that project has only just been shown off at CES and is still a ways away from hitting stores.
Nvidia's Tegra, on the other hand, has seen applications in portable media players, the Google Chrome OS-based netbook, and the Boxee Box. While all of these products are interesting designs, Nvidia is only a chipset provider. It cannot dictate the exact specifications of any of these devices beyond what's supplied by its very own Tegra.
New Frontiers Ahead
With the A4, Apple still maintains its long-standing relationship with ARM while delivering on performance, with a design that no competitor can use in its own products. More to the point, the A4 puts a very critical part of Apple's iPad under its very own control. And that move is unprecedented.
Going back to the earliest days of the Mac, Apple chose Motorola's 68k series of chips to power its Macs because they offered better performance than Intel's equivalent technology. In the early ’90s, the company migrated its Macs to the PowerPC architecture when Motorola couldn't deliver a 68k processor as fast and as energy efficient as Intel's Pentium series. Then, when the major vendors behind the PowerPC couldn't keep pace with Intel's Pentium IV and AMD's Athlon series, Apple switched its Macs once more—this time to Intel's own Core series.
Today, Macs remain beholden to Intel's specifications. If Intel can't keep pace, Apple will have to find yet another vendor for CPUs. But now, with the iPad's A4, Apple has demonstrated a new option: It has the ability to take existing designs and repurpose them to give its own products better performance than the competition.
It's extremely unlikely that Apple could leverage ARM architecture for the Mac, as those processors are specifically made for low-power devices, not high-performance workstations or general computing. However, if it could make headway through patents and cross-license agreements, Apple may someday be able to do this with another architecture, such as the now-standard x86-64.
While this fits in well with Tim Cook's declaration that Apple believes in controlling the technologies that run its own products, we're still a long way away from returning to the days when a vendor made every single component of the computer by itself. Economically, it doesn't make sense for one manufacturer to have that kind of overhead. That's how prices remain low, and why IBM doesn't produce every single component of the PC all by itself.
But it does demonstrate a brighter future ahead for Apple, and many in this industry have been shaken by the possibilities. While the idea of Apple exerting even more control over its products is a fascinating possibility, there's also an undercurrent of concern. An Apple that possesses firm control of every component of its devices has the potential for greatness—but with great power comes great responsibility.