Time to get excited about the growing Windows on ARM Ecosystem | Hacker News Reader

Time to get excited about the growing Windows on ARM Ecosystem

community.arm.com

71 pointspjmlp43 comments

I'm going to disagree with ARM here: it's not quite yet time to get excited about Windows on ARM. The app support just isn't there yet, and the emulation is not nearly as good as the emulation on macOS. I've been tracking major app and development platform updates [1], and while there is forward momentum, all the pieces aren't quite there for many apps to offer ARM versions. Many Windows apps use Windows Presentation Framework, and ARM support isn't available yet (currently scheduled for the first half of next year). Apps like 1Password and Affinity Photo are waiting for that update to enable them to offer ARM versions.

Hopefully by the end of next year developers will have no excuse not to offer ARM versions.

[1] https://iswindowsonarmready.netlify.app/

It's kind of sad that Apple will just pull the emulation in a year or two, so it's kind of beside the point how good it is. They offer it just long enough to induce people across the gap, not long enough to provide meaningful support for older software. They don't really do it for the users' benefit, per se.

If it makes it into Windows, it'll probably stick around for a decade.

I really hope Rosetta 2 lasts longer than Rosetta 1. Rosetta 1 was licensed tech that Apple presumably didn’t want to keep paying the license fee for, particularly when they were less wealthy than they are today.

Furthermore, Rosetta 1’s performance was too poor to provide a truly good user experience, whereas Rosetta 2 performs really well. And, there’s just a larger library of Intel Mac apps than there ever were PPC apps, because of how the platform has grown.

Apple most definitely won’t keep Rosetta 2 around forever because they want developers to release native apps. Apple doesn’t care if you can’t run an app that was last updated in 2010.

Rosetta 1 lasted 5 years as far as I can tell; there’s a huge gap between 5 years and forever.

Rosetta was first publicly available with Mac OS X "Tiger" 10.4.4, released in on January 10, 2006; the initial release of "Tiger", 10.4, was released nearly a year prior on April 29, 2005.

Mac OS X "Leopard" 10.5.8 was the last version of Mac OS X that supported booting on PowerPC-based Macs, released August 5, 2009. Apple's final update for the operating system was released in June 2011. An update for Safari and iTunes was released in 2012.

Mac OS X "Snow Leopard" 10.6.8 was the last supported version that contained Rosetta, released August 28, 2009. It became unsupported in late 2014, though it received an update in 2016. For all intents and purposes, let's end support in September 2014 with the final iTunes update.

So from 2006 to 2014, roughly eight years of official Apple support for versions of Mac OS X containing PowerPC-support. Eight years across three versions.

This was before the era of major releases being made available for free every year. If Rosetta 1 was anything to go by, and given that the Rosetta 2 technology is Apple-built, not licenced, we should see Intel support for quite some time — it depends on how quickly software is ported to Apple Silicon, iOS apps are ported to AppKit or Catalyst, and high-speed emulation of an Intel-based release of macOS becomes viable.

We may see Intel releases of macOS for a while longer than with the previous transition, given that the most popular Macs are MacBooks, and they run on latest gen Intel processors; compare that to the iBook and PowerBook, stuck on the G4 for years after the G5 came out.

But either way, 8 years is nothing to sniff at. :-)

Edit: mistook 10.5.8 as the latest Rosetta-supported Mac OS X release; it was actually 10.6.8.

> Mac OS X "Leopard" 10.5.8 contained the last supported version of Rosetta, released August 5, 2009.

Nope. Snow Leopard supports Rosetta, just not real PPC hardware (although even that appears to have been pulled at the last minute—PPC support is present in the kernels of developer previews, and people have successfully gotten retail versions to work.)

Anyway, the point of my post was I hope Rosetta 2 lasts longer, for the reasons I detailed.

Oh crap, you're right. It includes Rosetta but isn't universal, and I mixed those up in my head. Edit incoming!

Those who wait until it gets "good enough" for them might miss the wave

Though yes, if WPF is not working (not sure why) then it's harder to make the case for it

> Those who wait until it gets "good enough" for them might miss the wave

What do you mean? Are you saying that application developers should go out of their way to make their apps available for Windows on ARM? Do you think they'll miss out on users if they don't?

Yes (of course, it really depends on how much of "out if their way" is needed and the relevance of the platform to the developer) and yes

If your app runs fine on emulation, then it's ok. But you probably don't want to get caught with an underperforming app while your competition is already native

Perhaps not surprisingly, they didn’t communicate the most obvious reason: Apple Silicon has shown that it’s not only possible, but that there is potentially huge value in desktop Arm.

I think it's in ARM's self-interest that Apple doesn't run away with this market. Apple doesn't generate any revenue for ARM or it's a fixed amount, right?

That would mean ARM has a strong incentive to promote non-Apple ARM platforms.

I like your Best of CL post, too bad it's gone.

ARM has always had a strong incentive to promote more ARM platforms. Apple doesn't really change that, and they still pay royalties per chip, even with a perpetual license on the ISA AFAIK.

Simply having less x86/Intel around can't hurt. Competitors are going to want similar battery life. This effectively means ARM. Also, there's now ARM chips being used in very capable laptop, desktop, mobile, and server applications. All of this reduces cross-architecture friction (cross-compiling, different memory models, etc).

Honestly, consumers want similar battery life too.

My current laptop is the latest XPS. What a fucking power-hog. Dell's done well at giving me OK battery life, but I pay for the power with pretty appalling thermals. I LOVE the idea of a fanless laptop that also lasts well into double digit hours. So much so that this is a non-negotiable requirement for me. Just like the Surface lineup instantly disqualified all 16:9 laptops from my life for eternity, the M1 has instantly disqualified shitty thermals for me.

> they still pay royalties per chip, even with a perpetual license on the ISA

Why would they have to pay royalties per chip? I would've thought the whole point of a perpetual license on the ISA was that they didn't have to do that. What am I missing?

> I think it's in ARM's self-interest that Apple doesn't run away with this market.

Apple is a high-profile demonstration that ARM is perfectly suitable for desktops (well, laptops). Also, Apple usually doesn't corner the market - instead it blazes the trail and others follow (see smartphones, tablets, retina displays, watches and lousy keyboards). If history is any guide, others will soon follow.

All of this sounds like a great thing for ARM. Now, if only it survives nVidia...

Do Windows on ARM devices tend to have hardware support for x86-style Total Store Order (as does M1), or is it all done through software tricks? If the latter, it's hard to see how it can be both performant and reliable.

It's amazing how quickly ARM has become adopted by so many mainstream companies (for ordinary "desktop" operating systems).

My pet theory is that we can thank Chromebooks and SBCs like the Raspberry Pi. About 5 years ago, my preferred development machine was a 10" Chromebook with Crouton. The project gave you a full Linux chroot environment with shells and X windows surfaced as ChromeOS browser tabs.

I loved it. I could run EDA tools while sitting in the middle of the woods, and the battery would always outlast the sun. It was tiny and light enough to fit in an inside pocket. And I never felt nervous taking them anywhere, because I could write a few of them off for the price of a "real" laptop.

The only hurdle was the ARM SoCs. You could only run software that you could build from source, because almost nobody built proprietary applications for ARM ISAs. But the relevant tooling saw a lot of development and improvement during that time period, because suddenly a lot of people were giving them a workout.

So 5 years ago, you still needed a backup x86/64 machine for proprietary applications. But the myriad benefits were clear, and I'm glad to see that these power-sipping RISC chips might finally get their day.

And if you think this is exciting, just wait until more people start making RISC-V SoCs on modern process nodes!

> It's amazing how quickly ARM has become adopted by so many mainstream companies

ARM is 35 years old and first appeared in mobiles in 1997.

And now you can run proprietary EDA tools natively on Arm systems.

Since you talked about ARM Chromebooks, I wonder why so many current Chromebooks, including Google's flagships, are still Intel-based.

I would prefer Arm, but it is basically Intel competing on its own turf in terms of the laptop format. As Arm improved Intel could bring it's celeron lines closer to i3, then manufacturers could switch to i3, i5, etc, for premium. A laptop maker that did an Arm was also doing a unique system as a whole instead of modifying the ROM of a windows laptop and they all know that format with Intel very well. Further, things like performance were even more mysterious to consumers when looking at some manufacturer brand and ghz.

Really, Arm was more influential in that a second bid is influential, whatever an Arm could do became the floor. I think most vendors who did an Arm model were doing it for experience in working with Arm in a new format and consider if they want to make Android TV, etc, down the road.

It’s mobile phones. They helped push the perf/watt needs and mass manufacture drove the price of this stuff down a lot. Without it we’d be stuck with m4 and that class which is where ARM is particularly well suited. they’ve always won on lowest absolute wattage and have only started being competitive for perf/watt in the past few years. then being able to scale that up has only been more recently with Apple being the clear standout on what they’ve achieved.

Visual Studio needs to be on ARM. Otherwise you have to cross-compile, and the developer cost and effort to try to debug your code across multiple devices just isn't worth it these days.

As someone who doesn’t do this, why? GDB supports remote debugging for starters. So what makes it hard?

If anything, remote debugging is annoyingly slow - at least from my experience.

Even if the machine is right next to you?

There isn't any good hardware available today with Windows on ARM. So, while you CAN do it, it's not the best experience for doing it. By comparison, XCode shipped natively with the M1 release. The native debugging experience on Visual Studio is one of the best in the business. Take that away, and it's a crummy experience.

Please explain mobile apps then. That's all remote debugging, even if using the device simulator.

Visual Studio already runs on arm64 Windows.

VSCode (based on Electron) runs on arm64, not the full-featured Visual Studio 2019.

Nope, the fully featured VS runs on Arm64 Windows too. It's just not a native app.

Is there a "real" Windows on ARM yet, or is it still Windows RT? Can I build a proper native MFC app targeting ARM? Does it produce PE executables?

Windows 10 on ARM is real Windows, with Win32, MFC, and all. You can build and distribute native applications using any of those APIs. Some ARM-based Windows PCs come configured with Windows 10 S, which only accepts apps from the Microsoft Store, but those can be upgraded to Windows 10 Pro, which has full support for side-loading just like x86/x64.

Disclosure: I used to work on the Windows accessibility team at Microsoft.

How do I test my stuff on Windows on Arm?

Is there any virtual machine available for testing or buying a new machine is the only available option?

Buying a device is currently the only legal option considering that Microsoft is only licensing Windows on ARM to OEMs. That being said you can find older Snapdragon 835 laptops on eBay and such for pretty cheap. I was able to buy one for $260 USD.

Looks like QEMU has some ARM emulation support: https://en.wikipedia.org/wiki/QEMU#ARM

I've never actually run Windows on ARM in any context, so I have no idea how things are supposed to work once the VM exists. Maybe Microsoft will offer VM images...?

That wikipedia page is pretty out of date (notably it doesn't mention the 64-bit arm support at all...)

A few days ago I downloaded Parallels for Apple M1, then downloaded a free VM image from Microsoft for Windows 10 ARM. It all installed and sort of worked, but kept freezing randomly (for me), so I deleted it. In any case, it was actually all free, despite verbiage to the contrary in various places, including Microsoft's own website.

Test on wine on linux on arm in qemu?

The only fast enough WOA device right now is a virtualized M1, that’s kinda sad.

Reply on news.ycombinator.com