New leaked benchmarks that supposedly show Apple’s developer transition kit running Geekbench 5 Pro natively on the Mac mini are starting to make the rounds — meaning unlike the benchmarks that had leaked previously, which showed Geekbench 5 running virtually via Rosetta 2 on the transition kit, someone has supposedly found a way to run the program on the kit itself. But like the last benchmark rumours, there are caveats to these as well.
Screenshots provided to 9to5Mac of the alleged new results show a single core score of 1,098 and a multi-core score of 4,555. Previous results showed 844 for single-core and 2,958 for muti-core. At first glance, the disparity could be explained by the difference in how Geekbench is being run. Running any app natively verses non-natively is going to yield faster results since the program can run directly on the hardware without being translated by another program first.
However, if you look at the other information provided by the screenshots, you’ll see that the processor is listed at an Apple A12Z Bionic @ 2.49 GHz, which, OK, makes sense because the developer kit is also running on a A12Z Bionic ARM processor — but those scores are more in line with the second-gen 11-inch iPad Pro.
The Pro also uses the A12Z Bionic at the same processor speed, but that processor is running on eight cores, where previous developer kit benchmarks showed a processor with half the number of cores and a lower 2.4 GHz clock frequency. Not to mention the model of the device being tested in the supposed new leaked benchmarks is an iPad Pro 11-inch (2nd generation), which also confirms that it’s running on eight cores.
Screenshot: 9to5Mac, Fair Use
Screenshot: 9to5Mac, Fair Use
It appears that these benchmarks are not the developer kit at all, but just another iPad Pro. Another look at more recent iPad Pro benchmark results in Geekbench 5 seems to confirm this. And as Gizmodo noted previously, it appears the developers kits are only running on four cores instead of the rumoured twelve that Apple’s actual ARM processors are supposed to have once it starts shipping laptops with the new chip.
According to 9to5Mac, whoever was able to run Geekbench 5 Pro natively on the Mac mini developer kit allegedly did this by “booting into recovery, turning security features off, and codesigning apps,” which means that person was able to trick the Mac’s operating system into thinking that Geekbench 5 Pro not been altered or corrupted since it was digitally signed by the original programmer. Although, I’m not exactly sure why they would need to do this in the first place since Geekbench 5 Pro can run on iOS and is available in the App Store.
It also doesn’t explain the sudden appearance of four extra cores, according to the benchmark results. However, there is a way to disable the number of cores a program uses to run at the OS level, whether that’s tinkering in the BIOS, OS settings, or running a third-party program. It’s likely that the Apple’s Mac developer transition kits came with XCode, and with XCode you can limit the number of cores the computer uses to run programs. Since Apple made this transition kit just to assist developers in porting over their apps to ARM, it’s possible that they came with four out of eight cores disabled.
Why this would be the case, if it is the case, I have no idea. But that would also mean that the transition kits are running on just the iPad’s A12Z Bionic ARM processors — which means all these “leaked” benchmarks are nothing more than scores for a processor that has already been in existence for a while now.
The actual Apple ARM processors that will make their debut at the end of this year are rumoured to have a total of 12 cores, eight of which will be dedicated to performance. I just hope that performance is better than what we’re seeing from the iPad’s A12Z Bionic right now. I mean, just look at something like an early 2019 iMac with an Intel 6-core i5-8600 (two generations ago) — that yields a single core score of 1,155 and a multi-core score of 5,407, much higher than the 8-core A12Z Bionic.