A new analysis by AnandTech of the A8X’s GPU used in Apple’s new iPad Air 2 has revealed some surprising details regarding the new chip’s design and performance. While AnandTech initially thought that that GPU was based on Imagination’s PowerVR GX6650 – a 6-cluster GPU that currently represents the largest of Imagination’s GPU designs – additional investigation reveals that the GPU design has likely been customized by Apple for even higher performance, resulting in a GPU design that AnandTech has dubbed the GX6850. GFXBench measures the iPad Air 2’s performance at double the fill rate found on the A8-equipped Phone 6 Plus, and notes that Apple tends to have a preference for larger bus widths and lower clock speeds for the sake of energy efficiency, suggesting that Apple chose to build a custom eight-cluster design with this in mind. An unreleased die shot of the A8X confirmed to Anandtech that the new GPU design is essentially just two of the A8’s four-cluster GX6450s stacked together. Interestingly, the analysis notes that the iPad Air 2 is “overweight” in terms of GPU performance compared to the iPhone 6 Plus, providing roughly 30 percent better performance per pixel.