iPhone 5's custom ARM core could offer over 2X performance


— 1:58 PM on September 17, 2012

After being out of town at IDF last week, I spent part of Saturday morning trying to get a fix on what I thought of the iPhone 5. The media coverage was largely disappointed in tone, which wasn't a surprise since many details about the device had leaked beforehand, so Apple lost the shock value of past unveilings. Worse, however, was mainstream media coverage of the specs and such, since they literally offered no sense of key technical information.

What I mean is that listing "Yes" for 4G LTE support offers zero insight into what could be substantial hardware differences. One phone could support LTE with two hours of talk time and another with ten, but both would show up as "Yes." These giant tables of specs are virtually useless without better info and more context.

Such things were no idle concern for me, since I intend to upgrade from my iPhone 4 soon but was on the fence about the iPhone 5 (or, perhaps, the Lumia 920 or something).

Looking over the various specs tables out there told me something important: that we had very little meaningful information about the iPhone 5's hardware, especially the new A6 SoC. Apple, though, claimed a 2X increase in both CPU and graphics performance over the iPhone 4S—and they have a very good track record in recent years of delivering on such claims. That's a big deal, in my view, since it also means a ~2X increase over the iPad 3 in CPU performance, and that tablet seems very fast in daily use. This realization led me to seek out a hands-on video of the iPhone 5 in action, because I'm weird enough that I come at these things backward from everybody else, I guess. Anyhow, I soon found AnandTech's hands-on video, which showcased absolutely silly fast speed from an iOS device, in my estimation.

At that point, I figured Apple had implemented the new ARM Cortex A15, as rumored in places. Turns out Anand was tracking this story and found out something unexpected: Apple built its own custom, ARM-compatible CPU core for the A6. Looks like they didn't buy PA Semi for nothing. Then someone leaked some purported GeekBench results for the iPhone 5 that show it to be well over twice the speed of the 4S on average—and up to six times faster in certain tests, especially in memory-intensive workloads.

If those numbers turn out to be correct, it means a lot of things.

For one, it helps explain the how behind a pretty big generational leap, since the iPhone 5 is ~20% lighter, ~18% thinner, and ~2X faster than the 4S. That may not be innovation along the vectors you care about—I was hoping for a much larger display, for instance—but it's undeniably solid progress along the fronts Apple has pursued for multiple generations. On those fronts, it takes Apple well beyond today's competing phones, in terms of the size/weight/speed combination.

Also, it means Apple has arrived as a chip design company, and it raises all sorts of possibilities going forward, including Macs eventually making the transition to a different, higher-performance custom ARM core from Apple itself.

I'm sick as a dog and only following this story from a distance, but it seems to me the flock of "expert" media that have sprung up in the wake of the "post-PC era" have completely missed this story. I suppose reviewers would have gotten some clues eventually about the performance, once they used the phone for a while. Or, I dunno, perhaps folks have been trained by Apple's "post-PC" hype to truly believe that speed no longer matters in a computing device. Clearly, Apple doesn't buy its own rhetoric there. (Blame Apple, too, if you wish, for obscuring just about everything related to the A6 SoC.) Whatever the case, there are layers here that were previously unexposed, and they run contrary to the initial spin about the fundamentals of this product.

   
Register
Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.