The economical cores of the Apple M1 are 2 to 5 times slower than they actually are

The Apple M1 is equipped with an eight-core, four-speed and four-speed CPU, making all the Apple Silicon Max and the latest iPod Pro released so far. These are less powerful, but they consume significantly less, and they are used to extend the autonomy of devices for secondary tasks, require less power and run in the background.

This asymmetrical design is well known in the mobile world because Apple has been using it on all chips for iPhone and iPod since 2016. Their arrival at Magos offers a new opportunity: we can analyze what is happening at the CPU level, how each type of hub is used and the comparative power of the two types. Many theoretical experiments were thus carried out, but the developer Howard Oakley found a method Perform a test in real condition.

This developer used the method again Service quality (QoS) Force the MacOS code to run on powerful cores named Firestorm, then renamed the thrifty cores to Easterm. The quality of this service is a way of indicating the priority of a task to the computer, which, as described in the previous article, explains why the Apple M1 Max looks so fluid:

Using this technique he measured the time required for each of the four side hearts to perform the four basic tasks. This is a calculation performed on floating point numbers each time, a basic task often used to measure the performance of a processor. Four different methods were used for this calculation, from the lowest level (assembly code) to the highest, the Swift code as seen on the MacOS processor. Performance on the Mac is measured every time with the basic MacOS installation, without touching the computer and without the addition of additional kernel extensions, which is up to the theoretical tests conducted so far.

The results obtained by the developer show that the thrifty cores are really slow, but the difference can be well defined with optimal code. With method Apple Simt, The low-volume solution provided by the manufacturer is twice as slow as the economical four-speed, a good performance when you consider that only a quarter of the energy of the fast centers is required.

The gap is widening and we are moving away from low-level languages. With the basic Swift, economical cores are three times slower, but with the addition of more complex code, applications are often required to be five and a half times slower. In use, fast cores always have an edge, which makes sense, but even five times slower when done on economical cores, the asymmetry of the Apple M1 is still effective.

Computers spend more time on non-hasty tasks and may need more time to complete without problems. For Magos, this is especially true with spotlight scheduling or photo library analysis to give two examples. These tests demonstrate that in some cases, a developer can improve his code to control the gap and make the most of the economical cores for his application.