3 ms·
As a rough rule of thumb a core's performance tends to up as the square root of the silicon it has once you start going superscaler. The four little cores on a
by Symmetry 11y ago
As a rough rule of thumb a core's performance tends to up as the square root of the silicon it has once you start going superscaler. The four little cores on a typical octocore Android device are about as bit as one of the big cores. And the A57 you'd find a new octocore phone using are about half the size of one of the Twisters in an A9. Apple and an octocore Android phones are spending more or less the same amount of silicon on application cores.
- rayiner 11y agoOn a 20 nm process, the A8 CPU block (2x Twister cores) measures 12.2 mm^2.[1] The Exynos 5433 big block (4x A57) measures 15.1 mm^2, while the little block (4x A53) measures 4.6 mm^2.[2] The A8's quad-core PowerVR GPU measures 19.1 mm^2. So the extra space used in the Exynos configuration is almost enough to upgrade from a 4-core to a 6-core GPU. [1] http://www.anandtech.com/show/8562/chipworks-a8 http://www.anandtech.com/show/8562/chipworks-a8 [2] http://techreport.com/review/27539/samsung-galaxy-note-4-with-the-exynos-5433-processor/2 http://techreport.com/review/27539/samsung-galaxy-note-4-wit...
- Symmetry 11y agoThat's a bit misleading since the A8 CPU block doesn't include the 3MB L3 cache which is counted as it's own module. On the Exynos the 2MB cache in the A57 module and the 256KB cache in the A53 module are counted as part of their blocks.
- rayiner 11y agoFair point. But the A8 CPU block does include the 1MB of L2 shared between the two cores.