The Second Coming of Windows on ARM (Premium)

Lost amid all the excitement around Windows 11 are some much-needed improvements to Windows on ARM. Is the renaissance finally here?

Well, not quite yet.

We’re still missing one crucial piece of the puzzle, which is, of course, a new generation of Qualcomm Snapdragon chipsets for PCs that will finally elevate the performance reality of this platform to the level of the promises. Today’s ARM chipsets—including the Snapdragon lineup and Microsoft’s barely different SQ-series—are not powerful enough. More specifically, they do not meet the promise of “Core i5” levels of performance. Maybe they meant 5th-generation Core i5.

This is something I’ve seen again and again, and most recently when testing modern ARM-based PCs like the Lenovo Flex 5G and the HP Elite Folio (which I’m still currently evaluating). There’s a lot to like about these thin, light, and silent laptops. But performance—and, for now at least, compatibility—issues always undermine the experience.

But with this week’s announcements about ARM64EC and native Office for ARM, things are finally falling into place. All we need is new underlying chipsets. And for the passage of time, because some key compatibility improvements won’t happen until Windows 11 ships this October.

To better understand the problems with this platform, it might be helpful to review the complaints of the past. Since its announcement at a Qualcomm event in late 2016, Windows on ARM (WOA) has been problematic. It’s a 64-bit ARM platform, but to date, it can only run 32-bit Intel/x86-type apps via emulation, and because the chipsets are so slow running Windows to begin with, the emulation experience has been lackluster.

But it’s worse than that. WOA, to date, is also incapable of emulating the more common 64-bit Intel/x86-style apps that most users rely on. And it cannot use Intel/x86-style drivers, which are common and usually include better device functionality. For example, a WOA system will generally connect to and work with most printers, but the custom software that enables advanced settings—edge-to-edge printing, or whatever—is unavailable because it’s tied up in the driver install.

And WOA has other compatibility issues that are a bit harder to describe. Shell extensions, input method editors (IMEs), assistive technologies, cloud storage apps, and other utilities that modify the Windows user interface do not work on the platform (unless they’ve been recompiled for ARM, which few have). And games that require older versions of DirectX won’t work.

In the years since Microsoft announced WOA, it has, of course, released the platform on a handful of PCs and has evolved it with new capabilities, while Qualcomm, its silicon partner, has evolved its chipsets to work better with PC workloads. Things have moved slowly on both sides of the fence, more so with Qualcomm, frankly, but progress has been made.

The biggest change, of course, is the addition of x64 (64-bit Intel/x86-style) app emulation, which is currently in testing and will be formally added to the platform in late 2021. You can only experience this work by enrolling a WOA PC in the Dev channel of the Windows Insider Program. Until this week, that channel was not devoted to a particular Windows version, but as of yesterday, the Dev channel is now targeting Windows 11. So if you’re in the Dev channel, you’re getting Windows 11. And it includes x64 emulation.

That’s huge. As I’ve often noted of WOA emulation in general, running necessary software slowly is better than not having it at all. And in my personal experience, there are key apps—most notably Affinity Photos, but also Adobe Photoshop Elements before that—that are not available in WOA unless you have x64 emulation. This one feature shifts a WOA PC from “nope” to “just about there.”

Of course, x64 emulation doesn’t solve every problem, and WOA users are most likely going to need to live without Intel/x86-style drivers and older games forever, for example. But one announcement this week speaks to that bit about utilities that modify the Windows user interface, and to more advanced scenarios in which complex apps like Office and Adobe Photoshop are difficult to port to ARM because of their extensibility features.

I’m speaking, of course, of ARM64EC, which lets developers recompile parts of their x64 apps to ARM64 for native performance, always a good thing, while running other parts of their apps in emulated mode. This benefits all kinds of apps, but consider the case of Microsoft Office, which is being made available in native ARM64 for the first time this year, and will now support both native ARM64 performance and full compatibility with x64-based add-ons. Smart.

Between x64 emulation and ARM64EC, which, yes, will require developer buy-in to be successful, the WOA compatibility story is about to change dramatically. This is big news for fans of this platform, and for anyone who believes that ARM is the future. It’s big enough to support the notion that that latter belief can actually happen.

But again, there is one piece missing. Native ARM performance today is OK at best: Just using Windows—10 or 11—on an ARM-based system is generally fine, though there are some resource-intensive tasks like updating Windows, installing apps, and so on that instantly bog down the system. Running emulated apps, again, is even slower, but at least it works. What’s needed, of course, is more raw performance.

And for that to happen, Qualcomm and Microsoft need to dramatically improve their respective chipsets, and do so without killing the battery life that is also key to this platform’s promise. To date, the impact of performance improvements has been obvious: A Snapdragon 8cx-based PC does perform better than one based on previous generation Qualcomm chipsets, but it also delivers worse battery life. For example, Microsoft claims “up to 15 hours” of battery on its Surface Pro X, which is powered by an SQ-series processor derived from the 8cx. But claimed battery of first-generation WOA PCs was 25 hours, and I saw over 22 hours of real-world battery life in my own testing.

Today, many modern Intel and AMD-based PCs can achieve 10-15 hours of real-world battery life, some even higher. So one naturally wonders whether we simply end up with a situation in which WOA-based PCs offer no real battery life advantages but are still encumbered by some performance and compatibility issues. I certainly do.

But it’s important to remember the other advantages of ARM and of WOA-based PCs, which have no fans and are silent, and were designed from the beginning to provide superior power management and cellular data connectivity. For Intel and AMD to achieve the same, they will need to rearchitect their own chipsets with ARM-style big/little core designs. Which, of course they are.

I should also mention the dark underbelly of WOA, which is the cost.

WOA-based PCs are incredibly expensive. That HP Elite Folio that I love so much starts at an incredible $1700 and can be configured to over $2300. That’s over $1000 more than the typical premium portable PC, so even before factoring in its performance and compatibility issues, the price makes no sense at all. The cheapest Surface Pro X, including the cheapest (and very necessary) Type Cover starts at $1069, but that has the older SQ-1 processor, and a more reasonable starting configuration costs about $1500, or over $1600 with a Surface Slim Pen. Good God. These prices need to move down, and do so without sacrificing performance. Otherwise, the platform will remain a non-starter.

Whether WOA wins out in the end is still quite unclear, and the most obvious possibility is that it coexists alongside Intel- and AMD-based PCs and that all three see success in the PC market. But I’m guessing that the slow but steady progress that Qualcomm and Microsoft have made will finally bear some fruit. And that the next-generation Qualcomm chipset will be what finally puts Windows on ARM over the top.