Ever since Apple announced the A12Z and its move from x86, there have been questions about exactly how these ARM chips will perform and what we can expect from them. The first benchmark results are starting to appear from Apple development kits and as long as you take them with a mountain of salt, they are quite interesting.
What we have to work with here is Geekbench. Geekbench tends to be a very effective test for Apple CPUs, but in this case we are talking about Apple CPUs that run the x86 version via emulation. Although Geekbench favors Apple CPUs more than x86, running the application through an emulator will adversely affect performance.
Also, note that the application only reports four cores. The A12Z is nominally an eight-core chip, with four large, four small. It is not clear whether these development systems use only “large” cores, or if the application simply does not detect them correctly or if it is a limitation of the emulator. Regardless, it is very early and these are the first results.
Here are the data as they arrived at Geekbench 5.
We see single-threaded scores of 844 and a multiple-threaded score of 2958, which produces a scaling factor of 3.5x. On the x86 side of the equation, there is the 13-inch MacBook Pro, with scores of 1218 and 4233. This also addresses a scaling factor of around 3.5x. Likewise, the 13-inch Macbook Pro is approximately 1.44 times faster than the A12Z in both single-thread and multi-thread mode.
One thing to keep in mind is that emulation performance can vary drastically depending on the application. Some programs may run on relatively small penalties, while others crater and die. Rosetta 2 is specifically designed to avoid these results, but historically, there is a bad corner case or two lurking somewhere in any emulator. Some applications are more difficult to emulate than others. But the result of this effect is that we don’t really know if that 1.44x advantage that the 13-inch MacBook has is the product of the emulator’s handicap or if it is a good aspect of CPU performance. IPad Pro data suggests it may be the first.
If we assume that the A12X in the iPad Pro is valid enough for the A12Z, we can verify ARM’s native Geekbench performance, even if in iOS, not in macOS. Here we are examining 1120 single-core, 4650 multi-core, with a scaling factor of 4.16x. The 13-inch MacBook Pro is only about 8 percent faster than the iPad Pro in single thread and 10 percent slower in multi threads.
Frankly, this should send to frisson of fear through Intel and AMD. The implication of these findings is that the gap between the 13-inch Mac and the A12Z is largely the result of emulation. It’s not a guarantee, because the differences in the operating system matter in situations like this, but it certainly seems that most of the penalty that the A12Z is carrying is related to the emulation of the x86 code.
Apple’s annual record of delivering new performance improvements is significantly better than Intel’s at the moment. AMD can make a much stronger discussion about its recent improvement, thanks to Ryzen, but the huge 1.52x IPC improvement from Excavator to Ryzen skews the comparison a bit. To put it bluntly, AMD’s improvements over the past three years would be a little less impressive if the Bulldozer weren’t a terrible chip to start with.
We are in a strange situation at the moment. Intel has always been Apple’s primary supplier, but AMD today sells better-performing mobile CPUs, making them the most obvious comparison point. The 4900HS appears to score 1116 single-core and 7013 multi-thread. x86 MT is not, at least, in no immediate danger, in absolute terms. Keep in mind that the 4900HS also consumes much more energy than Intel or Apple chips.
What we see here is not proof that Apple will launch an ARM MacBook chip that can compete with the best Intel and AMD products it can offer. But it certainly places a limit on expected performance, excluding unusual emulator oddities that Apple will spend in the next few months to crack down on. X86 companies may want to ask mobile CPU designers to put in an extra cup of coffee.
Final note: these kits are not the CPUs that Apple will ship to customers and do not represent the final performance.
Characteristic image of Apple.