Apple has shipped the first Developer Transition Kits (DTK) to help MacOS developers prepare to switch from x86 to ARM. Even if Apple actually prohibits benchmarks, all sorts of DTK values are online in Geekbench. They show how well Rosetta 2 works, so Apple’s own is called a dynamic binary translation for the ARM chips.
The Developer Transition Kit contains an A12Z, i.e. the SoC, which also calculates in the iPad Pro (Early 2020). The Geekbench results were created with the x86-64-bit version of the benchmark, so they run under Rosetta 2 instead of native. It is striking that the A12Z only uses four of its eight cores – presumably the quartet of the fast vortex instead of the more efficient Tempest. This is understandable, since the Rosetta 2 layer is used and some performance is lost due to the translation.
With around 820 points on one core and around 2,800 points on four cores, the A12Z calculates faster than a Snapdragon 8cx (test) despite dynamic binary translation if it runs the Geekbench natively with ARM code. Qualcomm’s current SoC reaches 700 / 2,600 points, the modified version called Microsoft SQ1 in the Surface Pro X is at least a little faster. The A12Z also pulls in the Geekbench multicore test with the quad-core Core i5-1030NG7, which is soldered in the Macbook Air Early 2020 (test). However, it is not enough for an identical single-thread performance of roughly 1,000 points.
Apple Silicon should be fast and economical (Image: Apple)
However, the A12Z should have little in common with the Apple Silicon, which will be installed in future Mac systems such as a Macbook Air or an iMac or a Mac Mini. According to unconfirmed reports, an A14 derivative is to be used here, which is produced in TSMC’s N5 EUV process. The SoC should have eight fast cores (Firestorm) and at least four efficient cores (Icestorm).