JON HONEYBALL
In what could be as big a landmark as the launch of the iPhone, Jon explains why Apple’s switch from Intel to ARM could change the world as we know it
In what could be as big a landmark as the launch of the iPhone, Jon explains why Apple’s switch from Intel to ARM could change the world as we know it.
Place a note in your diary: 22 June 2020 was the day when the PC industry changed course, and nothing will be the same again. As Bob Dylan grumbled, “the times they are a-changin’”.
What has brought on this change? The announcement of the move of the Mac line from Intel chips to customdesigned ARM processors.
“But wait,” I hear the naysayers cry, “no one buys Mac laptops and desktops Jon! They’re toys for those who prefer to inhabit the Club and First lounges at airports, the sort of person for whom the badge is more important than the inner capabilities. They have a small percentage of the marketplace. Who cares what processor they use? No one uses them for real work?”
They don’t stop there. “Windows laptops are where the action happens! Intel rules the roost, with a side order of AMD to keep it in check. No one takes Apple seriously in this space, so who cares what CPU is inside? And isn’t ARM the thing that runs your mobile phone? How is that going to work in a serious laptop?”
I would forgive you for all of the above and, if that’s what you think, I’m here to tell you that you’re right. And oh so very wrong.
I’m not going to go into minute detail about the new stuff for the simple reason that I can’t: much is still to be unveiled. Instead, let’s look at the structural view of what Apple has announced, see how it will impact the marketplace and predict the knock-on effect it will have on the rest of the industry.
First things first: this is not a hasty switch. This is a calculated decision, which has probably been in the planning for more than a decade. If you look carefully, you’ll see a clear path that gets us from there to here. The rise of the iPhone and iOS on ARM processors. The introduction of custom Apple processors in recent iPhones and iPads. The additional custom and unique hardware processors that it has created, such as the T2 processor, right down to the silicon in the AirPods Pro with enhanced Bluetooth capabilities.
Take a deep breath because there’s more. The ultimatum that iOS developers keep up to date with the development technologies or get dropped from the App Store. The ability to persuade users to update their OS almost immediately when a new version comes out. An App Store that handles the purchasing process so the customer remains a customer of Apple, not the third party. The code-checking by Apple before an app can go live in the App Store, and Apple’s non-negotiable signing of the code, allowing it to kill off something that turns out to be nasty.
Now take stock of one clear fact: this trillion-dollar company is going all-in on the transition away from Intel processors into its own “Apple Silicon”, a new generation of system on a chip (SoC) hardware based around its own implementations of the ARM processor code.
This isn’t a niche company having a little flounce with the big boy Intel, engaging in a battle where the winner is obvious and the upstart will get squashed. Quite the opposite: this actually draws a line under the x86 instruction set and, finally and irrevocably, sets the date of the decline of Intel as a force in the laptop and desktop space. Intel has already failed in the mobile space and its bloated slowness, tardy releases and inability to scale to the latest fabrication techniques mean that the writing’s been on the wall for quite some time.
Microsoft has attempted a few half-hearted slaps around the face of its old friend. First there was Windows RT nearly a decade ago, when Windows 8 shipped. But despite it showing the promise of “all you like about Windows but without the x86 legacy and hence malware”, it fell flat on its face. The developer story wasn’t engaging and even Microsoft’s own Office group needed to fall back on x86 code cross-compilation – a technique that Microsoft forbade any third-party developer from using. Intel wasn’t bothered then and just rolled onwards.
Then came the Qualcomm work, bringing an ARM version of Windows that would run x86 32-bit code too. The first devices had good battery life, as you’d expect from an ARM-based device, but their performance sucked. And compatibility was hit and miss, to put it politely. Later versions jacked up the performance although at the expense of battery life, while Win64 Intel emulation is still missing in action.
“This sets the date of the decline of Intel as a force in the laptop and desktop space”
Microsoft can, and probably will, walk away from Windows on ARM. After all, what’s in it for Nadella and co? Microsoft has form too, having walked once already.
And that’s the core of the issue. Apple can’t simply walk away from this decision. It is betting the entire company on this move. Because it’s not only about moving macOS to ARM: it’s about creating a seamless development and delivery platform that stretches from the iPhone to the Mac Pro.
This is not the ridiculous task you might imagine. A year ago, Apple brought out Mac Catalyst, a tool that let developers take existing iPad apps, crosscompile them and deploy them onto macOS. Catalyst goes even further now, with muchenhanced support for all the desktop OS features you would want: proper dialog boxes, resizable windows, multiple window features, menu bars. And guess what? If you’re an Apple iOS developer, your app will be available to macOS users shortly, unless you specifically disable that cross-porting.
Then we need to consider the transition itself from Intel to Apple Silicon. Firstly, there’s a full build of Apple Silicon macOS 11 (note the increase from 10) available today for developers to start testing on. It’s a fairly peanuts rental cost – around $500, but you have to return it to Apple at some point in the future. Inside is a modified iPad Pro platform: it’s essentially the same as the iPad Pro, except it has 16GB of RAM rather than 6GB.
The platform includes several USB-C ports and the usual bits and pieces you will need, but this is in no way representative of what will actually ship as real hardware. Why? Because this is a low end hackedtogether testbed and nothing more. It doesn’t have lots of things it will need to have as a shipping product. For example, there’s no Thunderbolt support at all. That needs an Intel chipset so what will Apple do here? I suspect that the Apple Silicon will just go straight to USB 4, which has Thunderbolt support baked in. Look at the lack of Wi-Fi 6 support on the macOS platform – why bother doing this when your forthcoming
SoC platform will probably have that built-in, and maybe even go further?
So there’s an ARM platform box now, with hardware coming at the end of the year. I suspect that Apple will be moving all the hardware over by the end of 2021 at the very latest. This is not a slow transition – this is a stop, move and reset.
Does this mean that current Intel-based Mac devices are now end of life, which would be worrying for the future of the MacBook Pro 16in laptop I just bought at a cost of over four grand? No, not at all: macOS will be on both Intel and ARM for the foreseeable future, at least for the workable life of the existing hardware. New apps will ship in a “fat binary” format containing both the Intel and ARM executables.
Think this is clever? The greatgrandfather of macOS, the NeXTSTEP platform that Steve Jobs developed when he was away from Apple in the late 1980s, was available on four different CPU platforms: Motorola 68000, Intel i486, SPARC and PA-RISC. Apps developed for it could come as a four-way fat binary.
Indeed, there was a utility tool called “Lipo” that could “liposuction” out the versions you didn’t need on a computer to reduce disk space. Apple, through NeXTSTEP, was doing all of this over 30 years ago.
So what happens to my existing apps that are Intel based? If I buy an Apple Silicon computer for Christmas and I want to make use of the superb machine-to-machine system transfer capabilities that Apple customers have enjoyed for years (and which make Windows customers green with envy), what happens then? Well, there’s a technology called Rosetta 2 that is an on-the-fly code translator
– it takes your existing Intel app and, on installation or in itial execution, recompiles it to ARM code.
Again, Apple has history here: the first Rosetta was used to help ease the transition from the PowerPC platform onto Intel. It worked very well, and was supported for a surprisingly long time after the Intel hardware had long taken over the entire platform.
Rosetta will even work for OS-level kernel extensions (KEXTs). Want more? There is a hypervisor capability built into the Apple Silicon SoC, meaning you’ll be able to boot other operating systems. As it currently stands, these will have to be ARM versions of whatever OS you want to load. Microsoft could make Windows for ARM available, but currently it’s only available for OEM customers. Maybe it will change its mind.
Parallels and VMware have already announced support for their hypervisor platforms for the Apple Silicon platform. When asked directly whether they will support x86 code and operating systems running within that hypervisor session, they are coyly referring back to their Appleapproved announcement and say they have nothing more to add at the moment. I suspect that this is going to be an emerging story.
At the keynote, Apple showed its entirely suite of applications running on the iPad Pro-derived development platform. Including Final Cut Pro showing three 4K video streams running concurrently, and Logic Pro doing a complex mix. It demonstrated a port of Microsoft Office, and Adobe’s Photoshop too. Obviously it’s far from clear as to how well these run on this hardware, and how much more work needs to be done to get a well performing program, but both companies are onboard.
Get on – or be left behind
For more mainstream developers, the story is essentially simple. If you have been developing for iOS and iPadOS using Apple’s tools and frameworks then there’s little you need to do. If you have been developing for macOS using its tools then there’s work, but it shouldn’t be onerous. If your development is creaky around the edges then you have work to do. But Apple is not going to sit around waiting for you to catch up – you’re either getting on this train or it’s leaving the station without you.
The comparison with the Windows ecosystem couldn’t be more stark. On the one hand, Microsoft desperately wants to jettison Win32 and all that legacy stuff. It wants us to move to managed apps running on Windows 10X. But the roadmap here is far from clear, and the deadlines keep slipping. Microsoft just wrings its hands and wails that it can’t get rid of support for old stuff because its customers wouldn’t accept this.
The reality is that Microsoft could and should already have a solution in place whereby you could take an existing desktop computer running some legacy app, boot from a USB key, and migrate the entire platform over to a cloud-based Azure-hosted machine with remote access. After all, Microsoft is all about Office and Azure/cloud. Windows is legacy, and all attempts to move that forward have stumbled and fallen. There just isn’t the commitment to make it work; there’s lots of noise and bluster and then it falls quiet. Windows RT was one example. Windows 10 on ARM will probably be another. The Windows Phone platform died an ignominious death, and Microsoft’s app store aspirations are feeble.
That’s the problem: Microsoft is about cloud and “office stuff running everywhere”. In other words, cloud back-end services being talked to from Microsoft subscription-based apps running on whatever you want to use. There’s no room in this future for innovation on Windows, and it shows.
Apple, however, is betting the farm. And here’s the kicker. The move significantly weakens Intel, not because of the loss of chip sales, but because it will suddenly have a competitor capable of world-class
SoC design work, which is absolutely determined to make it work. Qualcomm’s work on ARM processors for Windows laptops? Qualcomm couldn’t really give a fig about that either – it’s an irrelevance in the bigger picture.
How do you like them Apples?
Let’s fast-forward a year into the future. Apple Silicon SoC is doing better performance than anything Intel can ship, with better leadingedge connectivity, GPU, artificial
intelligence engines and every tool you could want on the silicon itself. Significantly better battery life, due to the move to asymmetrical multiprocessor design where you have a mix of high-power fast cores and low-power-consumption slower cores, and a scheduler that allows the system to move the threads of the workload around in real-time as the workloads demand.
Will Apple take the opportunity to drop prices? Possibly – after all, you can buy an iPad now for £399, including VAT. An iPhone starts at just over £400. And yet Apple takes almost all of the profit from the mobile phone market globally. What happens if Apple does the same thing in the laptop and desktop world?
What will the impact be on the second and third-tier vendors, clinging on to a shrinking marketplace, reliant on a disinterested Microsoft for Windows, and a chip vendor in Intel that keeps tripping up and falling over?
So yes, the announcements of 22 June 2020 are profoundly important to this industry, and to the customer. A few issues ago, I said that the iPad Pro 12.9in with the Magic Keyboard and Pencil was the most impressive computing platform I’d seen in the last decade ( see issue 310, p52). That was no exaggeration. The potential for a unified smartphone/tablet/ desktop ecosystem, with modern app development and none of the historical crapware, will come as a blast of cold air to the incumbents, who have been slow, lazy and unresponsive for far too long.
And their sad, self-serving whining will get a wake-up call when they see how Apple has engaged with developers in a way that Microsoft has forgotten how to do. Who have engaged their customers with stores, online presence, safe app purchasing, and a relentless determination to keep customers’ data and privacy safe. How it is doing better silicon than anything Intel can manage, and how it will be beholden to no one. Oh, and did I mention the trillion dollar-sized company with hundreds of billions of dollars in the bank?
We most definitely are living in interesting times. Buckle up, this is going to be fun.