There has been plenty of speculation as to what Apple (NASDAQ:AAPL) plans to bring to the table with its next generation iPhone, often referred to as the iPhone 6s. There has also been much debate as to which semiconductor foundry will ultimately end up manufacturing the bulk of the A9 chips that are expected to power the device.
What there has not been much talk about, however, is what kind of improvements we can expect to see in going from the A8 inside of the iPhone 6 to the A9. Here is how I am currently thinking about it.
Time for a beefier graphics processor
When Apple released the A8, I was a bit surprised to see that Apple had chosen the Imagination PowerVR GX6450 graphics block -- I thought that the move to the much denser 20-nanometer technology would give Apple the area budget to implement the more powerful GX6650.
Since Apple is reportedly moving to a 14/16-nanometer FinFET manufacturing technology, the company should be able to fit more performance into a given power envelope. Imagination says that on an apples-to-apples basis, its Series 7XT GPUs should offer between 35% and 61% more performance than comparable Series 6XT processor at the same clock speed.
Furthermore, if the 14/16-nanometer technologies that Apple is reportedly building the A9 on allows the company to increase the clock speed of the A9 graphics block over the A8, the performance improvement should be even greater than the numbers Imagination provided.
Is this enough to retake the crown from Samsung?
The AnandTech review of the Samsung (NASDAQOTH: SSNLF) Galaxy S6 seems to indicate that the Exynos 7420 inside of the S6 has the fastest mobile graphics processor currently shipping, courtesy of the ARM Mali-T760MP8. The chip is implemented in Samsung 14-nanometer technology.
The GPU inside of the Samsung chip looks to be anywhere from 35% to 50% faster than the GX6450 that powers the Apple A8 chip.
If Apple implements the GT7400, and if it can achieve higher clock speeds with that graphics processor than the GX6450 inside of the A8, then the architectural improvements coupled with the frequency uplift should allow the A9 to outperform the Exynos 7420.
A big improvement in CPU performance likely coming up, too
In terms of CPU performance, the Exynos 7420 bests the A8 in all three of the performance tests that AnandTech ran. The largest performance delta -- seen in the WebXPRT test -- came in at just over 26%.
If Apple simply ports the "Enhanced Cyclone" cores over to a 14/16-nanometer process, then a clock speed boost -- perhaps to 1.8GHz -- would be enough to give Apple back its CPU performance lead.
That said, I do not think Apple is hiring top notch CPU architects just to port older designs to newer processes. I suspect that the CPU cores inside of the A9 will be substantially enhanced over the CPU cores inside of the A8. I also believe that, in addition to per-GHz performance enhancements, Apple will clock the A9 CPUs higher than the A8 CPUs.
Given that Apple keeps hiring top notch CPU talent, I would be floored if the A9 did not deliver a very healthy CPU performance gain over the A8.