How Might Apple’s ARM Silicon Perform in Future Macs? We Ran Some Benchmarks!
Macs will transition to Apple-designed chips that, according to our tests, are already better-performing (in some respects) than some Intel CPUs in today’s Macs. But the software changes required to support the transition could be more difficult and important.
By Tom Brant
Macs will begin ditching Intel’s chips in favor of Apple’s own silicon by the end of the year, the company announced this week as part of its WWDC keynote. But we don’t need to wait until then to take a stab at how the switchover is likely to affect the Mac’s computing performance.
In fact, a comparison we just did of recent-model Intel-powered MacBook Pro laptops with a 2020 iPad Pro and an iPhone 11, both using Apple’s latest A-series processors, suggests that the first Mac with Apple silicon should be at least as powerful as Cupertino’s current slate of tablets and phones, and could top the muscle of the current entry-level MacBook Pro.
A big caveat before we get down and dirty into those tests: This performance comparison is purely theoretical at this point. Apple’s current processors themselves in its mobile devices are capable on a physical silicon level, and likely to get even better. But a lot of the performance that future Intel-less Macs will offer will depend not just on hardware changes, but also on the software improvements necessary to accommodate them.
Microsoft’s own similar effort to transition some Windows laptops from Intel processors to Qualcomm ones is evidence of the multiple software hurdles and compatibility issues that can complicate such undertakings. (See: Microsoft’s Surface Pro X.) But as long as Apple is able to overcome these, the most powerful mainstream laptops of 2021 and beyond might very well come with a piece of Apple silicon as their brains.
How Apple Prepared to Ditch Intel
Apple’s own A-series chips have powered Apple mobile devices for more than a decade. Unlike Intel Core and AMD Ryzen chips, which use a microprocessor architecture known as x86, Apple’s processors are based on a design licensed from ARM.
These processors, which also handle graphics output and are capable of accelerating artificial intelligence (AI) computations, have grown immensely more powerful since their debut. Apple was even confident enough to make the claim in 2018 that the iPad Pro was faster than 92 percent of all portable PCs sold between June 2017 and June 2018.
With that kind of power potential in future silicon releases, it became clear that Apple might not need Intel to supply its Mac business anymore. But although the hardware was ready, the software was not. Rewriting the code of macOS and Apple’s own apps to run on an ARM instruction set instead of an x86 instruction set is a massive undertaking on its own. Getting an army of third-party app developers to do the same is exponentially more difficult.
The first task, it seems, has largely been completed, with Apple announcing at WWDC this week that it has ARM-native versions of its existing apps, and that it is has been working with key companies such as Microsoft and Adobe to work up native versions of essentials such as Office and Creative Suite. The second task will gear up this summer, with many developers receiving a specially modified Mac mini, dubbed the Developer Transition Kit, with an Apple A12Z Bionic processor and a new version of Xcode to help them rewrite their apps.
As it happens, the A12Z is the same CPU found in the latest revision of the Apple iPad Pro we tested. And that got us to thinking.
What Existing A-Series Says About Future Apple Silicon
Once those Mac mini Developer Transition Kits begin shipping out, we’ll have our first real-world look at how well Macs could run on Apple silicon. But because many cross-platform performance benchmarks already exist, we don’t need to wait until then to take a first, very preliminary stab at what kind of potential performance they and future Apple-silicon-based Macs might offer.
For the purposes of this sneak peek, we ran the same benchmarks on a 2020 13-inch MacBook Pro with a 10th-generation Intel Core i5, an iPhone 11 with an Apple A13 Bionic, and a 2020 12.9-inch iPad Pro with the same A12Z Bionic chip that will power the Developer Transition Kit. We also ran a 2019 13-inch MacBook Pro, using last-gen Core i5 silicon, through the same wringer.
Now, of course, all of this heavily simplifies matters and takes some major liberties. The iPad Pro runs iPadOS, not a native version of macOS that these future ARM-based Macs will. The iPhone runs iOS, of course. RAM allotments differ across the devices, and the way RAM is used by the different OSs also differs. And there’s no guarantee that any “real world” future Mac will use the A12Z Bionic that the developer kit and the iPad Pro do. Indeed, it’s very likely that the new Mac silicon will be very much its own thing. This is just a theoretical horsepower illustration based on the silicon that exists today, in the devices that exist today.
Also note that the Core i5 in the two MacBooks we used were chosen because (1) they were what we had on hand, on short notice, and (2) they are the entry level for the current smaller-screen MacBook Pro. These chips are equivalent to Intel’s U Series, the company’s mobile processors meant for slim laptops. The bigger MacBook Pro (today’s 16-incher) uses Intel’s more muscular H Series mobile chips, up to a Core i9, while Apple’s Mac mini desktop uses a true desktop CPU. Apple has shared zero details so far regarding what the possible performance grades could be of its coming home-baked silicon. We’re comparing what exists today, but it is very possible that coming Apple chips could debut on multiple tiers of performance for bigger or smaller devices, like today’s Intel silicon. We just don’t know yet.
Nonetheless…Let’s Get Benching
The benchmarks we used include Geekbench 5, an industry-standard method of evaluating CPU performance across Android, iOS, Windows, and macOS devices. We also ran some browser-based performance tests that measure the capabilities of the entire computer, including memory, storage, and graphics, in addition to the CPU. Finally, we ran a standalone graphics test, GFX Bench 5, that simulates the kind of graphics pushed by 3D games, running on Apple’s latest Metal API.
From just a cursory glance at the Geekbench results, it’s immediately apparent that the latest iPhone and iPad Pro offer theoretical performance that rivals (and, in some cases, exceeds) that of both the latest MacBook Pro and its 2019 predecessor, which has an 8th-generation Intel Core i5 chip…
Geekbench offers two sets of results in its CPU test. The single-core numbers give an indication of the processor’s inherent efficiency, while the multicore results suggest its maximum performance. Overall, they suggest that Apple’s A13 Bionic design, on this test at least, could be slightly more efficient than Intel’s latest 10th-generation Core i5 on a per-core basis. The A13 Bionic recorded the highest single-core result (1,330), as tested in the iPhone 11.
But the A13 Bionic in the iPhone 11 is designed for power efficiency rather than raw speed, so it has a lower multicore score than the other chips we tested. The A13 Bionic has two 2.65GHz high-performance cores and four lower-power cores, for a total of six. The Geekbench single-core score here represents the ultimate performance of one of the high-performance cores. The A12Z, being less power-constrained, has four high-performance cores running at 2.49GHz, along with four lower-power cores, for a total of eight. So, more high-performance cores, and more CPU cores in general, means a higher multi-core score.
Could Making Things Bigger Eliminate Throttling?
Dissipating heat is the most important challenge for any device with a microprocessor, and solving that challenge is fraught with tradeoffs. Phones and most tablets lack cooling fans of any kind. While most laptops include them, making them too big or powerful will result in an ungainly, noisy device. So the space allotted to active cooling hardware, processor thermals, and other essential components, such as battery cells, is an ever-shifting matrix of compromises, so long as your device has to be portable.
This conundrum helps explain what you see on our GFX Bench 5 test results below. The iPad Pro’s ability to achieve an average of 112 frames per second (fps) on the GFX Bench Car Chase scene is an impressive feat, twice as good as the 2020 MacBook Pro’s score. But that’s the mean value from three different test runs, and our testing showed that some throttling is likely occurring. The second test run, with the iPad already warm from the first one, achieved just 87fps.
Letting the device cool down before running the third iteration resulted in an even better score of 123fps. The fact that we were able to achieve the same results (albeit much lower ones) on each of three test runs with the MacBook Pro is an indicator that its throttling logic and cooling capabilities are more refined, as you’d expect from a much larger device.
A further wrinkle: Of course, whether Apple will rely on self-cooked integrated graphics or may farm out graphics acceleration, for some systems, to discrete-chip partners such as AMD (and its Radeon mobile wares), is entirely up in the air, and may well depend on the specific computer in question.
The Advantage of the Intel Ecosystem
Indeed, Intel has been making microprocessors far longer than Apple has, so it has the benefits of decades of experience troubleshooting issues that go beyond the CPU to affect other computing components as well as the operating system, cooling, and power management.
That could help explain why the performance differences on the more comprehensive WebXPRT 3 and BaseMark Web 3.0 tests follow a different pattern than the GeekBench and GFX Bench tests. Both of these tests saw one or the other MacBook Pro performing slightly better than the iPhone and the iPad Pro, although the margins are (mostly) modest ones.
On our last browser-based test, JetStream 2, the A12Z Bionic in the iPad Pro once again took the lead in this group. JetStream 2 combines a variety of JavaScript and web assembly benchmarks, covering a variety of advanced workloads and programming techniques, and reports a single average score.
Although JetStream is essentially evaluating a device’s web browser, the CPU, memory, and other components in the device determine how quickly the browser starts up, executes code, and continues to run smoothly over time.
Rosetta 2 and Beyond
The fact that the Core i5 MacBook Pro offers a better score on some performance benchmarks even though the A13 Bionic in the iPhone 11 is theoretically more efficient indicates that the quality of the Mac’s software is just as important as-and in some cases more important than-its forthcoming Apple silicon.
This is part of the reason why many consumers today consider Macs naturally to be more powerful than iPhones, even though Apple’s silicon has already reached parity with Intel’s by some measures. While you can run some powerful apps on an iPad, they’re typically not as feature-rich or as mature as their Windows and macOS counterparts. This is especially true of apps like Adobe Photoshop, upon which multimedia professionals making up a large portion of the Apple customer base tend to rely.
So even though the first Mac with Apple-made silicon is scheduled to hit stores by the end of the year, that silicon’s raw capabilities will probably matter less to most prospective buyers than whether or not the existing macOS software ecosystem can take solid advantage of it.
Based on prior experience during the switch from PowerPC processors to Intel ones in the mid-2000s, Apple knows that the hardware will be ready before the software. So it’s resurrecting its Rosetta emulation program to allow apps designed for Intel processors to run virtually on the new Apple ones. Emulation is a stopgap measure, a little bit like towing your car rather than driving it.
For now, all eyes are expectedly watching the Mac mini developer kit as the first concrete sign of how an Apple-silicon Mac might perform. The current Core i5-equipped Mac mini is a bit more powerful than a similar 13-inch MacBook Pro, with multi-core GeekBench scores around 5,000. It also has more room for cooling. Will the A12Z Bionic manage to take advantage of it?
Even though the A12Z Bionic is not necessarily going to be an exact analogue of what comes in the first ARM-based Macs (we don’t really know what the characteristics of the silicon will be), a lot is riding on the Developer Transition Kit-not so much from the point of view of the hardware itself, as from what developers do with it. We already know that Apple’s mobile chips can rival the capabilities of Intel’s under certain circumstances. But whether or not software developers will adapt quickly to them is anyone’s guess. If they do, Apple will have staged a coup: emancipation from Intel with similar or, possibly, even better computing performance than before. If the software transition is rocky, customers might wait it out, hold onto their older Macs longer than they might otherwise have, or opt for the next generation of Intel-powered Windows laptops instead.
Whichever scenario shakes out, it’s clear that Apple’s chip-design prowess is already well established. Now, a more difficult task must begin in earnest: transforming a software ecosystem over which the company has lesser control.
[Thanks to Sascha Segan and Wendy Sheehan Donnell for testing assistance on this story. -Ed.]
Originally published at https://www.pcmag.com.
