Apple’s A9X goes head-to-head against Intel’s Core M in ARM-x86 grudge match

For the past few years, the biggest CPU performance debate has been ARM versus x86. ARM has historically dominated mobile devices, while Intel owned the datacenter and PC industries. But as mobile power consumption shrank and smartphones took off, both Intel and ARM started eyeing each other’s turf.

Anandtech’s iPad Pro review puts Intel’s Core M directly against Apple’s most recent A9X SoC to determine how the two chips stack up against each other. This is easily the best comparison we’ve seen to date, for several reasons. First, the iPad Pro’s form factor and size compare well with multiple systems in the ultraportable and ultrabook market, which makes it less likely that inherent thermal restrictions will impact the comparison. Second, Apple’s A9X is the most advanced SoC on the market today, with the highest efficiency and best overall performance.

Anandtech’s entire review is worth reading, but we’re going to focus on the low-level SPEC CPU 2006 tests. The test systems included a 2015 MacBook, an Asus T300 Chi, and the Asus UX305CA. The latter system used a Skylake-derived CPU, while the other two used Broadwell-based Core M. We’ve excerpted a representative sample of the results below:

SPEC CPU 2006 performance. Data by Anandtech

The full suite of tests shows the A9X and Macbook each taking five tests, and our excerpt reflects this balance. The primary difference between the two platforms is that when the Intel CPUs win, they tend to win by much larger margins. Overall, the Core M still has an intrinsic performance advantage over its ARM-based rivals.

Skylake offers a 10% average performance improvement over Broadwell, despite a lower peak frequency. Data by Anandtech

There are also some interesting comparisons between the Broadwell and Skylake results. The Core m3-6Y30 has a lower turbo clock than the Core M-5Y31 (2.2GHz instead of 2.4GHz), but it improves on the Broadwell processor by an average of 10%. This puts it on stronger footing against the Apple A9X, but doesn’t fundamentally change which benchmarks belong to which camp.

Intel vs. ARM: Who’s winning the long-term fight?

There are two ways to evaluate this match-up. One — and it’s the one I generally tend to favor — is that physics and process are the primary barriers to faster performance these days. Both Apple and ARM have made huge performance strides in the last five years, but those improvements have required higher TDPs, more sophisticated power management, and better software support. This view argues that Apple and ARM have made such huge advances because they’ve been adopting techniques and technologies that companies like Intel and AMD started using a long time ago.

If you hold this view, than ARM and x86 performance will inevitably approach each other, assuming equal tool and compiler support, but will be hard-pressed to systemically beat the other by large margins. What we see in SPEC CPU 2006 are two CPUs that appear optimized for different workloads, as opposed to one CPU that’s just unilaterally better than the other at every kind of task. Independent evaluation of ISA power efficiency has generally failed to find evidence of any kind of systemic power advantage for one ISA over the other.

The other argument is that Apple’s chip division and its pure-play foundry partners will continue to close on Intel’s CPUs before beating them altogether. Intel’s process technology might still be more technologically advanced than its rivals, but that technical advantage isn’t translating to the kind of cost savings or performance scaling that would allow Chipzilla to compete effectively against the ARM ecosystem. Proponents of this argument point to Intel’s non-presence in smartphones, and its incredibly expensive contra-revenue strategy in tablets, as evidence that Santa Clara has either hit the wall with CPU design or can’t get its cost structures under control.

There’s a third possibility in play, though it isn’t anchored in the relative hardware capabilities of any vendor. Right now, there are three computing ecosystems that exist across a variety of form factors and devices: Windows, Android, and iOS. It’s possible that the computing landscape of the next decade will be shaped as much by software platform as┬áby hardware decisions. There will inevitably be some crossovers, but Windows 10 Mobile has fractional market share, and Intel’s piece of the Android market depends heavily on shipping devices contra-revenue (meaning at a loss).

The PC market has taken a beating in recent years, but it still accounts for hundreds of millions of devices, while data centers, business systems, and the HPC world are all strongholds of x86. “Android, iOS, or Windows?” has already replaced “Does it have Intel inside?” which itself supplanted “Is it 100% IBM-compatible?”

Most mainstream consumers never had more than a vague idea what CPU drove their computer hardware. In the future, they may not care at all.