The transition of Macs to the SoCs developed by Apple will take place from the end of 2020, but the Developer Transition Kit – a Mac Mini equipped with Apple A12Z Bionic and 16GB of RAM – is already in the hands of the developers who have to adapt their applications to the new platform.
Thanks to this we were able to discover the performance of the SoC – the same that we find on iPad Pro 2020 – struggling with the Rosetta 2 virtualization system, which allows you to run any existing macOS application even on Macs with chips developed by Apple.
The results of the first benchmarks showed us to what extent the process affects the performance of the SoC; the scores obtained on the virtualized version of Geekbench 5 show us values of 833 points in single core and 2,582 in multi core, which significantly differ from what can be obtained natively on iOS.
The data reported in our review of iPad Pro 2020, in fact, show us 1,119 points in single core, 4,665 in multi core and 9,625 in the GPU test. Today, however, we have the opportunity to compare these values with those obtained from a second benchmark of the Developer Transition Kit; we are talking about a test carried out with a version of Geekbench 5 natively performed on macOS 11 Big Sur and A12Z Bionic platform. Here are the results:
As you can see from the tables, the scores obtained this time are very close to those recorded on iOS. We are talking about 1,098 points in single core and 4,555 points in multi core. The score recorded in the Compute test dedicated to the GPU, which tears the result obtained by iPad Pro and shows us a value of 12,610, is surprising.
In short, the native execution of the benchmark allows us to confirm that Apple A12Z Bionic maintains (and in some cases improves) all the high-level performance to which it had accustomed us in the iOS environment. To make a comparison at Apple, we point out that a MacBook Air with an Intel Core i5 processor gets 1,160 points in single core and 2,950 points in multi core. Recall that the first Apple Silicon it should be a much more advanced and performing 12 core solution than A12Z Bionic.
The most careful will have asked two questions: how was it possible to natively run Geekbench 5 on the Developer Transition Kit? For what reason, to the voice Model, is the name "iPad Pro 11" listed and not that of the Mac Mini? The answer to these two questions comes from Steve Troughton-Smith, who explained on Twitter how it is possible to convert any iOS and iPadOS application in a few steps to run it natively on a Mac with Apple SoC. Below we offer a series of screens that show different known applications (such as Pages, YouTube, Spotify, Google Maps, Adobe Draw and many others) in action on macOS 11:
Smith explains that thanks to the reintegration of some frameworks previously removed from macOS (such as OpenGLES and various classes such as UIWebView) and reintroduced with Big Sur, it is now possible to make sure that iOS and iPadOS applications are run natively without any modification to the code; the developer simply needs to enable compatibility with macOS at compile time.
By intervening in this way you will have the possibility to port the apps without the least effort, even if this is not the best procedure. Applications converted in this way will lead you to believe that they are running on an iPad Pro with iPadOS 14 (why Geekbench detects that device and not the Mac) and this will limit the possibility of resizing and orienting the work window.
In the absence of optimization, in fact, iPad applications can be run in windows that imitate landscape or portrait orientations, but no intermediate solution that can ruin the basic interface (if the app supports the possibility of being resized, it can also be done on macOS, but it is still necessary to make changes in order to avoid unexpected behavior of the interface), while iPhone apps will run in small windows. In this case it is also impossible to assign custom gestures via the multi touch trackpad.
In short, it is clear that the work to move applications from one operating system to another is undoubtedly much simpler and more immediate than could have been initially imagined, however a good dose of work is required so that the final result can work properly even in desktop environment.