Skip to main content

Apple A7 SoC: Two Cortex-A57 & two Cortex-A53-class SoC, Rogue & >20,000 MIPS

Nearly 10 years after the first desktop processor moved from 32-bit to 64-bit (that was the Athlon 64 launched on 23 September 2003), Apple has unveiled the first 64-bit processor for smartphones, the A7.

Apple said yesterday that it is a product based on a “desktop-class” architecture, in other words, this is an ARMv8 part which may well be based on a big.LITTLE configuration that uses custom versions of the Cortex-A57 and the Cortex-A53.

We know the processor has two general-purpose registers, two floating-point registers, more than one billion transistors and a 102mm^2 die size. In comparison, an Intel Ivy Bridge processor with four cores, a 22nm manufacturing process and 1.2 billion transistors occupy 160mm^2.

Apple also said that it is up to twice as fast as the A6 (a dual-core part with a clock speed of up to 1.3GHz, with a 97mm^2 die area and fabbed on a 32nm process) with a 40x CPU performance compared to the original Apple iPhone. Back then, Apple used an ARM1176JZF-S from a Samsung processor, one that was underclocked by a third, and is even slower than the chip used by the Raspberry Pi.

This chip reaches 515 DMIPS (1.25 DMIPS/MHz, 1 core, 412MHz) which would mean that the A7 hits 20,500 MIPS. That’s around 40 per cent more than the dual-core Samsung Exynos 5250 that powers Samsung Chromebook (3.5 DMIPS/MHz, 2 cores, 2GHz) but still a far cry from the 30,000 DMIPS the Exynos 5 Octa is rumoured to be hitting.

But it is not all about raw power. In the case of Apple, it is perhaps more important to stick to a power budget and, given that the battery life of the iPhone 5S is actually similar to the iPhone 5 under similar scenarios, one can suspect that the same 1,440mAh battery is being used (turns out that the iPhone 5S has a 1570mAh model and the iPhone 5C has a 1,507mAh battery, respectively nine and 4.7 per cent higher respectively).

As for the GPU, the inclusion of OpenGL 3.0 means that it is a dead cert that ImgTec PowerVR Series 6 is being used. Apple said that the current GPU is twice as fast as the one powering the iPhone 5S and 56x faster than the original iPhone but didn’t actually specify what the unit of comparison is.

The PowerVR MBX Lite in the original iPhone had a fill rate of 1 MT/s and 100MPx/s at 200MHz. The Tri-core SGX543MP3 that powers the iPhone 5 reaches 35MT/s and 1GPx/s at the same clock speed while hitting 21.6GFLOPS. One can assume therefore that the Series 6 GPU on the A7 will hit around 44GFLOPS.

Third on the list is the M7, which Apple presented as being a motion coprocessor, a piece of silicon that is dedicated to processing specific data. It is interesting to see that we’re starting to see a trend with the return of coprocessors which were big business back in the days of the 486 and before. The Moto X, with the X8 SoC, has two dedicated DSPs, one for contextual computing and the other for processing natural language.

So to recap, the A7 is probably a quad-core part, consisting of two big cores and two small ones, implemented in a big.LITTLE configuration, etched on a 22nm process or finer with a PowerVR SGX Series 6 GPU.

Desire worked at ITProPortal right at the beginning and was instrumental in turning it into the leading publication we all know and love today. He then moved on to be the Editor of TechRadarPro - a position he still holds - and has recently been reunited with ITProPortal since Future Publishing's acquisition of Net Communities.