The iPad 2 sports Apple's brand-new A5 processor, about which, as usual, Apple is saying very little. Apple's official spec sheet says a dual-core processor clocked at 1GHz, but not whether it uses a version of ARM's Cortex-A8 or Cortex-A9 design, or another ARM-compatible architecture designed by Apple's chip division.
Comparing the A5 directly to existing designs can help. The only dual-core ARM processor on the market right now is Nvidia's Tegra 2, found in devices like the Motorola Xoom and Motorola Atrix 4G. If the two chips are of similar architectures, they should perform similarly on raw measures of processing power.
Getting to those measures is a bit tricky. You can't compare anything that involves the browser or Javascript because then you're really comparing Android and iOS, and you can't use anything that compares graphics, because then you're comparing Nvidia's GPU to the Imagination PowerVR SGX543 in the iPad 2. Also, benchmarks have to run similarly on both Android and iOS.
I found three benchmarks which purport to do this: Linpack, a simple math benchmark; the Stockfish chess program, which calculates possible chess moves; and a little app called BenchIt, whose CPU component just calculates Pi.
Here's what I found. Initially it doesn't appear to make much sense.
Explaining the Results
Let's start with the assumption that on a pure processor benchmark, the first-gen iPad and the HTC Inspire 4G should perform at least somewhat similarly. The iPad's A4 chip and the Inspire's Qualcomm QSD8255 are both 1GHz Cortex-A8 compatible designs. (Of course, if Apple is doing something odd with the iPad's clock speed, this comparison goes out the window.)
Then let's add the fact that we know the Atrix and Xoom, with their dual-core Cortex-A9, are much faster than the Inspire. Where does that leave the iPad 2?
It leaves us with two bad benchmarks, first of all. Linpack turns out to be more reflective of a device's Java engine than of its processor, according to Linpack for Android creators Greene Labs. If the iOS version of Linpack is using the new Nitro JavaScript engine, that could explain the iPad 2's high score. I've left this number in, though, because this makes Apple's point that well-written software matters even more than hardware for many aspects of user experience.
On BenchIt, both classes of Android devices are faster than their iOS counterparts. Comparing them within their own OS classes, though, the Atrix and Xoom are 2.42 times as fast as the Inspire, while the iPad 2 is 3 times as fast as the iPad 1. Okay.
Stockfish presents yet another face. There, the iPad and HTC Inspire perform similarly, the two Motorola phones perform similarly, and about twice as fast...and the iPad 2 is smack in between.
What does this mean? I have no idea, but I'm working on it, and I could use your suggestions. Full results are below. At first glance, the dual-core A5 chip does not appear to be very similar to a Cortex-A9, but rather is probably a highly proprietary design that is also compatible with the ARMv7 instruction set. Qualcomm's upcoming Krait chip takes the same approach, as does Marvell's Armada. As in many other things, it looks like Apple is truly blazing its own trail when it comes to mobile processors.
