2013: The year that the “ARMpire” strikes back

2013 is going to be a major year for both ARM and Intel. For years, it’s been obvious that the ARM and x86 architectures were on a collision course. The two sides skirmished in 2012 – Intel launched its first x86 smartphone and tablet products, Microsoft debuted Windows RT, and Qualcomm talked up its own plans for notebooks based on quad-core Krait processors. The actual amount of competition between x86 and ARM products, however, was relatively small.

In 2013, that’s going to change. This year, ARM and Intel are going head-to-head – but a lot of the early buzz is going to focus on the Cortex-A15, and A15-class devices like Qualcomm’s Krait and Nvidia’s Tegra 4.

The rise of quad-core smartphones

We’ve historically taken a dim view of the need for quad-core parts in tablets and smartphones given the inevitable battery life trade-off and the general lack of multi-core-optimised software. That hasn’t changed much, but quad-core phones are going to be big, big news the next few months. Expect a lot of companies to talk up raw performance capabilities, 1080p screens, and similarly power hungry features.

Unfortunately, there’s already plenty of evidence that the current crop of quad-core phones experience significant throttling under load. Tests of the Nexus 4 show a 50 per cent performance difference in GLBench’s Egypt HD test depending on whether or not the phone is benchmarked inside a freezer or at room temperature.

That problem isn’t going to go away and battery life will inevitably suffer because of it. The drive towards ever-higher screen resolutions will only exacerbate the issue. A 1920 x 1080 screen has 2.25 times as many pixels as a 1280 x 720 display, and driving that many pixels at an acceptable frame rate takes more power – both computationally, and to illuminate it adequately. It’s no accident that the Nexus 4 throttles most sharply in 3D tests. Recent power consumption tests really drive this home; the Cortex-A15 in Samsung’s dual-core Exynos dissipates much more power than anything else on the market.

Nevertheless, there’s opportunity here. Qualcomm’s quad-core Krait is already on the market, Nvidia’s quad-core Tegra 4 has just been unveiled at CES, and Samsung’s quad-core Cortex-A15 shouldn’t be far behind. Ultimately, quad-core is simply going to work better for tablets than it does for phones. The expanded form factor and larger battery gives manufacturers more room to tailor solutions to specific markets.

Dual-cores dig in

Despite the quad-core push, the 2013 smartphone sweet spot is going to stay with dual-core chips. The transition from Cortex-A9 to Cortex-A15 (or A15-class processors) will give dual-core smartphones their own performance boost, as will the adoption of next-generation GPU technology. The difference here is that while quad-core phones will be an ARM-only arena, Intel’s dual-core Medfield chip will make its own debut this year, with two times the graphics cores and two Hyper-Threaded 32nm CPU cores. It’ll also be Intel’s first solution with native LTE – to date, Qualcomm has had a virtual monopoly on 28nm LTE-capable radios.

Intel isn’t talking much about Merrifield yet, despite having announced the chip almost a year ago, but it has told us to expect more details at Mobile World Congress. We can extrapolate, however, based on what we saw with Medfield in the Orange San Diego. A dual-core Medfield at the same 1.6GHz should scale well compared to the single-core variant. Lightly-threaded benchmarks should pick up a 15-20 per cent performance boost, while quad-threaded tests could see gains of 50-60 per cent.

Much will depend on Android’s ability to distinguish between the physical second core and a Hyper-Threaded logical core; Intel has likely done a significant amount of work to ensure that software can take full advantage of the dual-core/quad-thread CPU.

So how well will Merrifield compete with the next-generation of Cortex-A15 dual-core phones? Like Medfield, we expect it to be middle-of-the-pack overall, with very strong performance in some areas and competitive (but not extraordinary) performance in others. It’s an important move for Intel because it demonstrates that the company can iterate and improve overall performance on a yearly basis, even if it doesn’t establish a commanding lead over other smartphone designs.

Tablets and notebooks

What do Microsoft’s Surface RT and Samsung’s latest Chromebook have in common? They’re both devices that claim to provide the basic functionality that modern users need, without an x86 chip under the hood. Windows RT sales have been tepid, at best, but that’s going to change as next-generation Cortex-A15 devices come to market. Similarly, Qualcomm has made it absolutely clear that it intends to launch notebooks with Snapdragon hardware under the hood.

Intel’s Core business is in no danger, but Clover Trail could be. The Windows tablet market is just emerging and there’s going to be a lot of mud thrown at the wall as manufacturers suss out what sort of devices consumers are looking for and jockey for position. That’s where Bay Trail, Clover Trail’s successor, is supposed to drop in.

In the past we’ve talked about Bay Trail to some degree – this is the quad-core, Out-of-Order-Execution, no-HT processor built on 22nm that Intel believes will offer superior performance per Watt to anything the ARM camp has to offer with the A9 or A15. It’s a big deal, with a lot of upgrades over Clover Trail, including USB 3.0 support on some models, but it won’t debut until the back half of 2013.

What do customers really want?

Bay Trail could put some serious hurt on the A15, not because ARM’s latest CPU is slow, but because its temperature under load is a challenge for tablet and smartphone vendors. GPU offloading could help (Nvidia may well focus here), as could manufacturing tweaks and more sophisticated software throttling. Here, the variables are all familiar, even if their values are still misty.

The really big question of 2013 is what sort of products consumers are going to want to buy. This will be the first time Windows, iOS, and Android devices have gone head-to-head. It’s the first time that x86 and ARM have gone toe-to-toe in the Windows market and we may see some competition in the Android arena as well. All of that makes for a very interesting year. And of course, it’s equally possible that for all the hundreds of millions poured into marketing, development, and sales, the Windows tablet market could fail to fly in 2013 on both x86 and ARM devices.