This week saw the announcement of a new MacBook Air, Mac Mini, and new iPad Pros. The Mac Mini was well overdue for an update (the update took 1475 days) and the MacBook Air update got rid of the last non-Retina Mac, but the star of the show was the new iPad Pro. The new iPad Pros contains an 8-core CPU and a 7-core GPU which allow them to do a lot of work quickly. Additionally, the 1TB storage option unlocks an extra 2GB of RAM, for a total of 6GB. Apple is positioning the iPad Pros as mobile powerhouses for people doing creative graphics works, video editing, & CAD. Based on the results of the benchmarks, Apple isn’t wrong. They’re now calling the iPad Pro “Like a computer. Unlike any computer.”, which is interesting. The marketing has shifted such that the device is supposed to be a replacement for your traditional desktop/laptop computer since the iPad is portable, but can be transformed with a keyboard case or connected to an external display or other devices via USB-C. For the general consumer who just browses the web, plays a few games, and does a little bit of photo/video editing along with some light document work, this makes sense. What is new is that the expandability of the iPad Pro with USB-C now allows for power users to do even more with the iPad than they could before, almost eliminating the need for dedicated workstations. So, we know what Apple engineering can do when heat is a concern, but what would they be able to accomplish when heat wasn’t much of a concern because there is an active cooling solution in place (aka the Mac)?
It hasn’t gone unnoticed that Intel has had some issues with their 10nm process, resulting in small tweaks to their existing 14nm CPUs: Skylake->Kaby Lake->Coffee Lake and not much movement on mobile either with Whiskey Lake and Amber Lake. This is in addition to the CPU industry as a whole dealing with the resulting changes due to Spectre, Meltdown, and the new PortSmash architectural vulnerabilities leading to somewhat degraded performance due to kernel mitigations and the disabling of simultaneous multithreading (SMT). The uncertainty of Intel’s roadmap and the delays due to the need to change the architecture to mitigate security issues has lead to delays in Mac launches since Apple is tied to Intel’s release schedule. So, how would they fix this? Make their own silicon.
Apple announced back in 2005 that they would transition all of their Macs and software over to the x86 architecture from PPC starting in 2006 and finish in 2007. The software transition required the creation of Rosetta, the binary translator for PPC code to run on Intel CPUs. It was not installed by default in macOS 10.6 (2009) and was removed with macOS 10.7 (2011), giving developers 5 years (starting at WWDC 2005) to port their code over to the new architecture. Fast forward to today, and Apple has a 7nm ARM SoC process down pat and, as of the time of this writing, the only manufacturer who has achieved this milestone. Because of this, they have achieved equal (some times better) performance on their mobile chips compared to Intel and AMD desktop CPUs. This means that an ARM powered Mac would perform great with applications compiled for ARM, but the real challenge is x86 emulation on ARM performing well. Microsoft has started to try this out of low-powered ARM windows devices and, it hasn’t gone well and has severe limitations. Additionally, Intel has been threatening ARM chip manufacturers over x86 emulation. Assuming Apple doesn’t get into a legal battle with Intel, they will need to create Rosetta 2 to help all of the existing software run on an ARM powered Mac and it has to be near equal performance to be successful.
There is another question about ARM based Macs though: will Apple also provide the GPU? Apple has been using AMD GPUs for some time now (much to my chagrin, Nvidia has better performance) due to a falling out of sorts between the two companies and power concerns (Nvidia cards are much hungrier than AMD’s). In my mind, I would think that the CPU transition would be more important over the GPU transition, so launching with an ARM CPU and still have dedicated AMD GPUs would allow for more time to create a SoC with a desktop class GPU in it. However, if Apple wants to do everything in one go, it’ll take more time and leave the Mac lineup in Intel’s hands for much longer.
Regardless of how long it’ll take, there is no doubt that an ARM powered Mac is coming. However, I tweeted the position that I’m in now: I need a new Mac, but I know that if I get one now, the money that I spent will be worthless when Apple drops Intel support. My 2008 MacBook Pro is still alive an kicking, so I want to ensure that a ~$3000 computer purchase lasts at least 6 years. I’m betting that 2019 or 2020 is the launch date for ARM Macs and I really hope that they are good.