Windows On ARM — Again
by Jean-Louis Gassée
The ARM architecture has proven itself in many places, from IoT systems, to Raspberry Pi homebrew kits, to Apple Silicon chips. Today, we look at the ins and outs of Microsoft moving from Wintel to WinARM.
When Apple began their ARM journey by acquiring P. A. Semi (née Palo Alto Semiconductor) in 2008, the reasons were obvious: ARM chips are simple, open, and flexible; and the architecture provides an inherent gain in thermal design power (in layman’s terms, it runs cool). Thirteen years later, the investment has resulted in industry-leading M1 Pro and M1 Max Apple Silicon chips.
ARM’s advantages haven’t escaped Microsoft’s attention. Just before leaving Microsoft in 2012, Windows President Steven Sinofsky published an extensive, erudite paper detailing his team’s efforts to create Windows on ARM (WOA)
WOA builds on the foundation of Windows, has a very high degree of commonality and very significant shared code with Windows 8, and will be developed for, sold, and supported as part of the largest computing ecosystem in the world.
Unfortunately, the initial WOA endeavor failed to make a dent in the marketplace. I bought a Surface laptop running Windows RT and saw what it could and couldn’t do…and promptly resold it to a Polish software developer who was anxious to bring it back to Warsaw.
Today, things have improved…somewhat. Go to The Best Windows on ARM Laptops You Can Buy and you’ll find models from Acer, HP, Samsung, and Microsoft itself that offer variants of ARM processors made by Qualcomm. But there are problems: A number of popular Windows apps are yet to be ported to ARM, some run (slowly) in an emulator, but very few are fully-native ARM apps. Worse, you pay more than you would for an Intel-based laptop from the same vendor. In some cases, you pay more than you would for a MacBook Air that runs ARM software much more skillfully.
As a Be engineer liked to say: It costs more, but it does less
(Of note, Microsoft Surface revenue went down by 17% year-to-year for the quarter ended last September. Apple Mac sales increasing by 1%.)
This leaves the future of WOA less than clear. In the perfect world of Theory, where every thing simply works, moving Windows to ARM is a no-brainer. ARM-native versions of Windows apps running on cheaper, more efficient chips inside lighter, environment-friendlier machines…What more could you want?
First, we must take care with the “ARM” label. There’s no standard for ARM chips the way there’s a de facto agreement for what we call x86 CPUs. While x86 is called (by Wikipedia, at least) a “family” of architectures, the agreement is that most software built for the family will work well enough on all x86 chips. This was due, at least in part, to the well-established library of low-level software called “drivers” that intervene between applications and the CPU, smoothing out the bumps.
The “works well enough” agreement was enforced by Intel and Microsoft. There were occasional clashes, but the x86 world was policed, nonetheless.
Today, there’s no broad culture, no Wintel, that can regulate custom-built ARM chips. Chip A could legitimately be called ARM, but it won’t necessarily support software that was built for chip B. For example, an ARM chip designer could build extended instructions for gaming software that causes trouble when the game is run on a different ARM-based CPU that doesn’t have the correct drivers.
Samsung would like to be considered a prospect for the Intel mantel, but the company probably isn’t competitive enough. The most likely candidate to play the de facto standard-setter for ARM in PC applications is Qualcomm. After their Snapdragon chip was found wanting when compared to Apple Silicon, the San Diego company acquired NUVIA, a company started by former Apple Silicon engineers who wanted to build chips for no-holds-barred server applications. As I wrote in a March Monday Note, “Qualcomm Buys Unintentional Apple Silicon Spinoff”, the acquisition was lauded by Microsoft:
When [Microsoft] says it sees “an incredible opportunity to empower our customers across the Windows ecosystem”, does that mean more Windows On Arm developments, more ARM-native Microsoft applications?
This sounds like a good plan — until you look at the transition between Wintel and WinArm worlds. There would be a long period during which one would be faced with the following choices: a Wintel laptop, a WinARM device, or an Apple MacBook. Here, I can’t help but imagine the Apple marketeers smiling and modestly calling their product the Real ARM Choice.
But, wait, there’s more. As Intel watches this unfold, how would we expect them to react? Would they continue to place their bets on x86 with the hope that the new-fangled Extreme UV Lithography (EUV) process would fill the complexity gap? Or would Intel CEO Pat Gelsinger push one of his new foundries, currently being built in Arizona, into the ARM design and manufacturing business? After all, he has publicly vowed to regain Apple’s business and he clearly understands it won’t be with an x86 design. Of course, such public vows have a way of going nowhere, as we saw with a decade of “Just You Wait!” Intel imprecations.
Whatever happens, it’s not going to be harmonious. For a while, the lives of decision-makers and users are going to be more complicated.
As it turns out, Microsoft just scored a higher market capitalization than Apple’s: $2.523T vs. $2.482T this past Friday. Perhaps Microsoft CEO Satya Nadella, a strong leader who turned the company around after Steve Ballmer was let go, can make the transition to ARM more manageable. It would behoove Nadella to decree a new kind of Pax Microsoftia by stating what the WinARM standard is going to be, with whom, and when.
In the meantime, I’ll try my hand at installing Windows 11 on the Raspberry Pi 4 I ordered a couple of days ago. No warranties expressed or implied regarding the effects of my klutzy exploration of ARM-based hardware.
— JLG@mondaynote.com
