After years in the shadows, Intel commits to Atom as a top-flight processor

The PC market is evolving rapidly and Intel is determined to adapt alongside it. That fact was much in evidence during the company’s Q2 conference call this week, in which it sharply cut estimates for the back half of the year, but simultaneously affirmed its dedication to ramping new products based on Haswell and Bay Trail. This was Brian Krzanich’s first earnings report as CEO of the company, and he gave a limited update on how Intel is responding to changes in PC demand and product mixture.
The greatest change is this: Atom isn’t just a component of Intel’s strategy — going forward, it’s a major pillar. The plan, Krzanich said, is to “move Atom even faster to our leading-edge silicon technology and focus on the SoC integration of key components like graphics, communications, and other devices.” That’s a significant change, even from several years ago, when Intel announced it would accelerate the previous Atom roadmap.

Atomic angst

When Intel launched Atom in 2008, it envisioned the processor as carving out a new market in mobile devices that could never have used a conventional x86 processor. MID designs, however, were in their infancy and customer interest was low, so Intel also created a new type of sub-notebook. Dubbed the netbook, these explicitly low-end, low-cost, small machines were meant to appeal to a limited group of users that wanted a second small machine for occasional work.

Everyone, including Intel, was surprised when netbooks turned out to be an explosive growth opportunity rather than a minor pitstop on the way to bigger things. It took Intel months to ramp Atom production to meet demand for the chips and it presented the company with a serious problem. MIDs were too different to be a threat to the company’s entrenched mainstream x86 products, but netbooks could’ve cannibalize Core sales if the company wasn’t careful. To prevent this, Microsoft and Intel both implemented strict restrictions on what sort of peripherals a netbook could or couldn’t carry. Atom’s performance, already weak at launch, was virtually ignored in the years that followed — each die shrink or new iteration focused on cutting power consumption and little else.
That began to change when the iPad caught fire and Intel decided to make a serious push into the smartphone and tablet space, but Atom still lagged behind. Again, concerns about cannibalization came to the fore. Unfortunately, Intel just doesn’t have a choice any longer — its smartphone business is following a decent trajectory, but Windows 8 tablets have been a flop. It needs parts it can sell into the $199 to $299 tablet space while competing with ARM designs as well as AMD’s Temash, and that means it’s time for Atom to cut its teeth on new nodes and aggressively adopt new technologies as Intel deploys them on other parts. With the bottom of the PC market dropping out as consumers shift to Android tablets, Intel needs a part it can deploy across both Android and Windows ecosystems.

Expect differentiation

This doesn’t mean Atom and Core are going to move in strict lock step. Remember, one of Intel’s goals for Bay Trail is to deploy the chip in smartphone SoCs, where every square millimeter is a precious resource. It may not make sense to deploy a feature like AVX2 on Bay Trail — widening all of the chip’s registers to 256 bits isn’t much of an issue for Haswell, but it makes a difference for power consumption in a tiny mobile part.

That doesn’t mean Atom will never add AVX2 (Intel hasn’t said one way or the other on Bay Trail). What we expect is that Intel will continue to aggressively deploy a mixture of performance-boosting and power-saving characteristics, and we’ll see each deploy when it makes sense to do so. Intel isn’t giving up on Core, or on the idea of driving big x86 cores into a wider range of devices, but Atom is increasingly critical to the company’s long-term stability and success. Intel is claiming to expect up to 150 tablet design wins for Bay Trail, at prices as low as $150. Meanwhile the company isn’t backing off margin claims either; Intel believes 55 to 65% is still a realistic broad range for its products through the indefinite future.