Apple has never been one to publically disclose the specifics of the hardware it uses in its mobile devices – at least, not to the level that most hardware enthusiasts would prefer. Instead, companies like Chipworks are tasked with dismantling and dissecting hardware down to a level that leaves few questions unanswered as they’ve recently done with Apple’s new A8 SoC.
Chipworks partnered with Ars Technica for analysis on the project and after going back and forth, they’ve come to some pretty solid determinations as to the makeup of the A8 chip.
The analysis validates Apple’s claim that the chip was fabricated on a 20-nanometer process, this time by Taiwan Semiconductor Manufacturing Company (TSMC) instead of Samsung. As such, it’s one of the first SoCs to get such treatment which allowed Apple to add in additional features while reducing the die size by about 15 percent.
The die shot clearly reveals just four GPU cores which immediately rules out the six-core PowerVR GX6650. Instead, it appears Apple has selected the smaller four-core GX6450 which is the direct successor to the G6430 found on last year’s A7.
The CPU block, meanwhile, is a bit more complicated. The analysis concludes that it is definitely derived from Cyclone (Apple’s first 64-bit ARMv8 core) with a few additional tweaks to make it even more potent than the original design. What’s more, the die shot also suggests that the L2 cache has been reorganized to a per-core design since there is no obvious single block of L2 on the A8.
Last but not least, the chip appears to be using the same 4MB of SRAM cache memory used to service the GPU and CPU.