Ever since Apple announced the A12Z and its transition from x86, there have been questions about the performance of these ARM chips and what we expect from them. The initial benchmark test results began to appear in the Apple development kit, as long as you are not bored, they will be very interesting.
What we need to use here is Geekbench. Geekbench is often a very powerful test for Apple CPUs, but in this case, we are talking about running the x86 version of the Apple CPU through emulation. Even if Geekbench does favor Apple CPUs more than x86, running applications through the emulator will affect performance.
Also, please note that the application only reports four cores. The A12Z is nominally an eight-core chip, with four large and four small. It is unclear whether these development systems only use “big” kernels, or applications just cannot detect them correctly, or whether this is a limitation of the simulator. In any case, it is still early, and these are early results.
This is the data into Geekbench 5.
We see that the single-threaded score is 844 and the multi-threaded score is 2958, which produces a scaling factor of 3.5 times. On the x86 side of the equation, it is a 13-inch MacBook Pro with scores of 1218 and 4233, respectively. This also calculated a zoom ratio of about 3.5 times. Similarly, in single-threaded and multi-threaded mode, the 13-inch Macbook Pro is 1.44 times faster than the A12Z.
One thing to remember is that simulation performance may vary intensely Depends on the application. Some programs may receive relatively small punishments, while others may crash and die. Rosetta 2 is specifically designed to avoid these results, but historically, any simulator has one or two latent extremes. Some applications are more difficult to emulate than others. But the result of this effect is that we really don’t know whether the 1.44 times lead of the 13-inch MacBook is a product of emulator impediment or CPU performance is very good. The iPad Pro data indicates that it may be the former.
If we assume that the A12X in the iPad Pro is a good replacement for the A12Z, then we can check the ARM native Geekbench performance, although in iOS, not macOS. Here, we are studying 1120 single-core, 4650 multi-core, scaling is 4.16 times. The 13-inch MacBook Pro is only 8% faster than the iPad Pro in single-threaded mode, and 10% slower in multi-threaded mode.
Frankly speaking, that should send a Friesen Fear of Intel and AMD. The implication of these results is that the gap between the 13-inch Mac and A12Z is largely the result of simulation. This is not a guarantee, because operating system differences are important in this case, but it seems that most of the punishment A12Z bears is related to emulating x86 code.
Apple’s annual record of providing new performance improvements is much better than the current Intel. Thanks to Ryzen, AMD can make a stronger argument for its own recent improvements, but the huge improvement of 1.52 times IPC from Excavator to Ryzen makes the comparison a bit skewed. Frankly speaking, if Bulldozer is not such a bad chip from the beginning, AMD’s improvement in the past three years will be less impressive.
At present, we are in a strange situation. Intel has always been Apple’s main supplier, but now AMD sells higher-performance mobile CPUs, which makes them a more obvious view. The 4900HS seems to have a score of 1116, a single-core score of 7013, and a multi-thread score of 7013. In absolute terms, x86 MT is at least not immediately dangerous. Remember, the power consumption of 4900HS is also much larger than Intel or Apple chips.
What we see here does not prove that Apple will launch a MacBook ARM chip comparable to the best products Intel and AMD can provide. But this will definitely have a detrimental effect on the expected performance. Unless there is an abnormal simulator quirk, Apple will cancel this phenomenon in the next few months. x86 companies may want to ask mobile CPU designers to add more coffee.
Final note: These kits are not CPUs that Apple will deliver to customers, and do not represent final performance.
Featured pictures of Apple.