Apple announced plans to transition from Intel CPUs to its own series of CPUs, dubbed Apple Silicon, in June. The transition will occur over the next two years, and Apple announced the first machines today – the MacBook Air, Mac mini, and 13″ MacBook Pro. This transition has been a long time coming as Apple’s A series CPUs have outperformed laptop and desktop CPUs for many years. Apple is finally unleashing them into Macs.
There was a lot of speculation about what Apple would release and how similar it’s strategy would be to the Intel transition of 2006 – 2007. Would Apple Silicon Macs look like Intel Macs or would they introduce revolutionary new designs? What CPUs would they run – the A14 from the iPhone, an A14X shared with a forthcoming iPad Pro, or something totally different? How fast would the CPUs be? Which models would include them? Would they be available in different speeds?
Those questions were answered today. Similar to the Intel transition, Apple Silicon Macs are indistinguishable from their Intel counterparts. The Air looks exactly like an Air, the mini like a mini, the Pro like a Pro. There are a few minor feature bumps here and there but nothing major. Apple is saving its radical designs for later. This release strategy follows the Intel transition almost exactly, also focusing on the mini and the MacBook Pro prior to other models. All models cost the same, or less, than the machines they replace.
Apple Inside
In terms of CPUs, Apple has decided to brand its desktop CPUs as the M series (not to be confused with the M series of motion coprocessors in iOS devices). The M1 appears to be the same across all models, save for a minor difference in the number of GPU cores on the two Air models. Apple has not released any frequency specifications and there are no build-to-order options available. All models are powered by an M1 with 8 cores (4 high performance, 4 high efficiency) and 7 or 8 graphics cores. Performance will differ between models mostly due to how much cooling they have – the less cooling available, the lower the sustained performance ceiling. The Air will have the lowest performance because it has no fan, followed by the MacBook Pro with a small fan, and I believe the mini with a larger fan.
The main foci of Apple’s 45 minute announcement were performance and power usage, neatly summed up as performance-per-watt, the same metric Apple used to justify the Intel transition. It touted a bunch of performance numbers across the three models – 2.8x – 3.9x faster Final Cut Pro, 2.3x faster Lightroom, 3x – 4x faster gaming, depending on the model, beating nearly all systems in the same price range. The are also far more power efficient, allowing MacBooks to last up to 20 hours on a single charge. This performance combined with power efficiency results in the highest performance-per-watt in the industry.
The M1 also adds a few additional capabilities to the Mac that iPhones and iPads have been enjoying for years including faster unified memory, dedicated machine learning hardware, instant wake, high-performance automatic encryption, and activation lock.
How Does it Compare?
Apple’s performance numbers are fine, but they don’t specify exactly what was tested, so it’s hard to judge exactly how much faster they will be versus existing models. Those performance numbers are also based on native apps, not apps that are emulated through Rosetta 2. Most apps will fall into the second category for quite a while until developers rebuild their apps. According to Apple’s documentation and developer testimonials during the announcement, porting to Apple Silicon is low effort. Even so, it will take time. Adobe, for instance, is porting its apps over, but they won’t start being available until December.
To make this more real, let’s consider my current hardware. I’ve had my MacBook Pro for a few years now, a refurbished 2016 model. It works fine for Lightroom but things could always be faster. I enjoy the screen space of a 15″ or 16″ model but I really enjoyed the size and portability of my 13″ MacBook Pro. Could an Apple Silicon MacBook Pro be in my future? Let’s see, with the help of a table. Below I’ve listed the specs and performance of my MacBook Pro, Apple’s highest-end Intel models, the A14 iPad Air, and a supposed A14X that actually might be the M1 to see what the advantage is.
| Model | Specs | Performance (Single / Multi Core) |
|---|---|---|
| My 2016 MacBook Pro | 2.7 GHz i7 (4 cores) | 802 / 3255 |
| Highest end 2020 13″ MacBook Pro | 2.3 GHz i7 (4 cores) | 1245 / 4551 |
| Highest end 2019 16″ MacBook Pro | 2.4 GHz i9 (8 cores) | 1118 / 6980 |
| iPhone 12 Pro | 3 GHz A14 (6 cores) | 1581 / 3822 |
| “A14X” leaked performance | 1.8 GHz (8 cores) | 1634 / 7220 |
Some interesting observations can be drawn from this table. First, my Mac is slower than everything currently available. Even a refurbished Intel machine would provide a 40% – 50% boost over what I have. My iPhone 12 Pro is nearly twice as fast in single core performance and around 30% faster in multi core, probably averaging out to the same 50% boost overall. But I can’t run regular Mac apps on my phone. Given that the M1 is most definitely faster than an A14, the “A14X” numbers look realistic and could be real. We’re talking 30% – 60% faster than the highest end 13″ MacBook Pro, and even eclipsing the highest end 16″ model by a bit. That translates to a 2x speed boost over my current Mac in a smaller package and with a better keyboard. Not bad.
The iPhoneification of the Macintosh
Apple Silicon brings several advantages and also several changes to the Mac. The M1 is a system on a chip, meaning that nearly everything necessary to run the Mac is in one package – the CPU, the GPU, the memory, the coprocessors, the power controller, and the I/O controllers. I think this effectively kills third party memory upgrades. The MacBook line has not allowed users to install additional memory for years, but the Mac mini has always provided user-serviceable memory slots. I’m 99% sure those are gone for the M1. The memory is built into the package and is limited to 8 or 16 GB as manufactured. I’m not sure what Apple will do for the Mac Pro, but it may follow a similar pattern, just with larger memory ceilings available as build-to-order options.
I believe this also kills off discrete GPUs used in high-end Macs like the large MacBook Pro, iMac, and Mac Pro. Apple talked a lot about its Unified Memory Architecture and how it provides blazing performance by being shared between the system, the Neural Engine, and the GPU, exactly the way the A series does. This only works because the CPU, GPU, and Neural Engine are manufactured into in the same package and share an ultrafast data path. Doing this with memory slots or a separate GPU slows everything down significantly by forcing data to travel across the system bus. If Apple wants to keep leveraging the performance of Unified Memory it can’t support discrete GPUs. There is an extensive talk from WWDC 2020 about the differences between AMD / Nvidia GPUs and Apple GPUs. Apple’s product information lists discrete external GPUs as compatible with its Intel machines. This information is absent for Apple Silicon Macs.
I believe this will simplify the Mac lineup in the same way it simplifies the iPhone and iPad. Base hardware will be similar across all models, with major distinctions being form factor and storage space. Users will choose between a few memory configurations, but that will be it. There won’t be multiple CPU speeds to choose from or optional GPUs. I think Apple will standardize on certain combos across its lines. The “consumer” models will get one CPU/GPU configuration and speed while the “pro” models will get another. There won’t be 3 different CPU and 3 different graphics configurations for a single model as there are today (e.g. 16″ MacBook Pro).
With the power of Apple’s integrated solution comes a loss of customizability to which power users are not accustomed. Normal computer users probably won’t notice as they are driven by price and whether it “does what it’s supposed to do”, but us tech heads will have a bit less to debate over. The addition of Activation Lock will add a level of complication for the used market as buyers will need to make sure their Macs are not locked before purchase, similar to current iOS devices.
What’s Next
Apple has stated that it will take two years to transition to Apple Silicon. It provide similar estimates for the Intel transition and delivered significantly earlier, migrating in under a year. There are rumors that Apple will release a redesigned iMac in the first half of next year. This may bring with it the 16″ MacBook Pro. I think the Mac Pro will be last and might push closer to that two year timeframe.
In the near-term there are three Apple Silicon Macs to choose from, each with a variety of form factors and price ranges. Expect most of Apple’s software to go native very quickly, with major applications from Adobe, Microsoft, and other firms going native soon after. Smaller apps will continue to run under emulation and may suffer a bit, though Apple stated that some apps still run faster under emulation than natively on the previous model. Apple will tune performance over time and things will get faster and snappier as apps become native.
While I’m very excited, I’m not running out to buy an Apple Silicon Mac immediately. I’m very excited to read reviews and performance analyses, to learn about the internals of the design, and to see how the M1’s additional capabilities are used. The Mac world is in a very exciting place and I’m excited to observe it for now and see how the transition plays out.
Relatively Ambitious 