This ignores shared memory between GPU and CPU, akin to Intel iGPUs of yore. It also ignores the less efficient memory management of using 16KiB pages, unlike x

This ignores shared memory between GPU and CPU, akin to Intel iGPUs of yore. It also ignores the less efficient memory management of using 16KiB pages, unlike x86 using the more space-efficient 4KiB pages. This is a macrumours.com article, though.