Why Windows on ARM is far from dead

[Redaktion]

Some people compare Windows on ARM with Windows RT, declaring its death. But such proclamations may be a tad early. Arguably, Windows on ARM is more alive now than it has ever been since the 2016 Qualcomm/Microsoft partnership announcement, as it is getting ready for a busy year 2024.

https://www.notebookcheck.net/Why-Windows-on-ARM-is-far-from-dead.768037.0.html

[A]

X Elite benchmarks are a bit sus, they showed 15K multi-core Geekbench for 80W and 14K for 23W initially (still show that on NBC too), it can make some people raise an eyebrow.

[The Werewolf]

ARM fans seem to love playing alternating games of strawman arguments and moving goalposts.

Windows has been available for RISC processors like MIPS, SH3 and ARM for decades. And sorry, like it or not, Windows RT has to be taken into consideration because it was the first head on attempt to make a consumer ARM-based Windows portable.

The problem is twofold.

First is the need the tech press (and even industry) has to panic everytime Apple does something unexpected, and second, to consistently fail to grasp why you can't map Apple to the entire computer industy.

Apple is a single company. They make their own OS and hardware. They do not for the most part license any of it to anyone else and they work very hard to make sure that once you're in the system, it's as hard as possible to ever get back out.

That means they can do something no one else can do: they can make huge changes with little fear of losing customers. If you want to stay on Macs or iPhones, you're going to do it their way. When Apple decided they wanted to own their own CPU, you had no choice but to switch.

That's simply not reality for PC makers. In Microsoft decided "ARM-only," people would either refuse to upgrade or switch to Linux or MacOS. This isn't even hypothetical - we saw it with Vista and Win8 and we're seeing it again with Win11.

There already are several ARM laptops out there, Samsung makes a couple (I have a Galaxy Go), so does Microsoft - but they don't sell well because in the end, they don't offer anything more than an Intel or AMD laptop or desktop does for most consumers while breaking existing software that has to run in emulation.

And as a developer, I have no real reason to port to ARM, easy as it is, without a customer base to justify it because it's still more cost to me.

It's a classic chicken-and-egg problem. Customers won't migrate to ARM if the software they use doesn't run on it or runs badly, and devs won't port to ARM without customers.

Apple doesn't have this problem because they can basically mandate that devs support ARM or their software gets removed or blocked. Microsoft cannot do that.

In the end, the hardware doesn't matter - it's software and workflow. Whatever hardware can run it will always be the first choice, but people won't change their workflows "just because"... well, unless they're Apple fans.

[ArsLoginName]

1 - A brings up a great point regarding the power draw. But if the leaks about Zen 5 8950HS from RGT are correct, AMD is in a similar situation with only a 10% boost when going from 35 W to 95 W or so if I remember the scores correctly. They went wide (similar to Apple and Qualcomm) and run into a performance-power wall pretty quickly. Also, if the info from Charlie at Semiaccurate is right (he generally is on target), the power draw of QC Elite X is due to QC mandating the use of certain electrical components and design choices. That is, OEMs could do it better, but QC says "This is the way."

2 - The Werewolf is spot on regarding software-hardware ecosystem choices between Apple and Windows. There is a different. It's Apple's way or the highway for MacOS and iOS.

[A]

And as a developer, I have no real reason to port to ARM

Smh... As a developer you should know that most of so called "porting" within the same platform goes as far as recompiling to different target. And support for arm targets is established long ago.

No one has to go or will go "ARM only", and Apple never did (yet). Apple apps are mostly Universal, containing builds for both platforms in them. The same app runs natively on Intel and M macs.

[MQ.Yang]

No one needs ARM on windows. period.

[Anonymousgg]

ARM's "advantages" over x86 are often overrated, and Qualcomm/Microsoft botched Windows on ARM hard the last time.

Let them take another shot, and compete on price. Maybe it can be different.

[A]

Unless we get removable memory and storage, I'd rather not.

And as a developer, I have no real reason to port to ARM

Smh... As a developer you should know that most of so called "porting" within the same platform goes as far as recompiling to different target. And support for arm targets is established long ago.

No one has to go or will go "ARM only", and Apple never did (yet). Apple apps are mostly Universal, containing builds for both platforms in them. The same app runs natively on Intel and M macs.

That's what you think. generally that is the case, you set the compile target and it compiles for that target. But that is assuming all the libraries exist for that compilation target. And it isn't uncommon to get bugs when compiling to a different target. Because while it may be fine 99% of times, that 1% can be a PITA requiring writing nasty workarounds.

On top of that, many developers won't even bother. This is why you see even electron apps not compiled for all operating systems even if all it would take the developer is changing 1 setting.

[A]

That's what you think. generally that is the case, you set the compile target and it compiles for that target. But that is assuming all the libraries exist for that compilation target. And it isn't uncommon to get bugs when compiling to a different target. Because while it may be fine 99% of times, that 1% can be a PITA requiring writing nasty workarounds.
On top of that, many developers won't even bother. This is why you see even electron apps not compiled for all operating systems even if all it would take the developer is changing 1 setting.

I think you mess up a bit cross-platform porting and cross-architecture porting. If libs are not present for your arch - recompile them. If you have compilation issues - dev tools should be fixed, not your app. The only real deal is porting x86 SIMD instructions (that is, if you were using them low-level), you will have to spend some time moving to Neon... But even that can be semi-bypassed with solutions like simd-everywhere etc.

Apple's rapid platform switch was achieved mostly because porting is "single click". And it's probably the reason why Visual Studio is among the first apps ported to ARM by MS, so this time MS is doing it right - "emulation of old architecture + dev tools for new one" is the king.

And yeah, f electron apps. )

[A]

P.S. ARM switch on Windows is actually not about app porting at all. It's all about external GPU support. So the ball is on nVidia's side - in the end they will be the ones to decide if Windows moves on or stays with outdated architecture from 80s with a huge burden of backwards compatibility and semi-faked performance in benchmarks by using more and more complex sets of SIMD instructions.

In the end Windows economy is about gaming PCs/laptops and those juicy 4090s.

[YUKI93]

Looking at NBC's Robo & Kala review shows the real advantage of ARM over x86. The only thing that Windows on ARM need to grow bigger is more developers' interest. Qualcomm can throw in Apple Silicon-beating SoCs, but it won't make much difference if desktop software developers still go to macOS more than Windows.

[NikoB]

That's simply not reality for PC makers. In Microsoft decided "ARM-only," people would either refuse to upgrade or switch to Linux or MacOS. This isn't even hypothetical - we saw it with Vista and Win8 and we're seeing it again with Win11.

So they moved on, so they moved on that the share of users of Linux distributions does not exceed 2-3%, approximately the same as 20 years ago. There wasn't really any problem with W8.1 other than not being stupid and changing the shell to something else. From the point of view of compatibility with x86 software, problems began in Vista-W7.

And as a developer, I have no real reason to port to ARM, easy as it is, without a customer base to justify it because it's still more cost to me.

This is a problem as old as time. It can only be solved by a real breakthrough in performance (quality) several times greater than on x86, which we do not see. There is no "silver bullet" that would completely destroy the x86 market.

Users would immediately switch to Arm if it were several times faster. And the operating system is friendly and comfortable for the average person. But it is unpleasant and uncomfortable even for IT professionals on the x86 platform, so what can we say about ordinary users? Constantly lagging behind in hardware support and real performance.

pple doesn't have this problem because they can basically mandate that devs support ARM or their software gets removed or blocked. Microsoft cannot do that.

This is one big problem for Apple. Developers on its platform will only do something if they get a greater benefit than if they exist in the x86 ecosystem (where, for obvious reasons, there is much more competition). As soon as Apple destroys this difference, developers will immediately leave the Apple platform and it will be left with a bare bottom. Nothing personal - this is business...

In the end, the hardware doesn't matter - it's software and workflow.

Yes, it does. After all, the x86 share is more than 80% of the market and for good reason. Because even now x86 processors are 1.5-2 times faster than the fastest Apple solutions, as proven by tests. At the same time, having 1000 times more high-quality commercial and free code.

[A]

Users would immediately switch to Arm if it were several times faster.

People will switch to ARM when they see laptop can carry them through transatlantic flight while doing their normal workflow and still being compact and silent. Almost no one gives a f*ck about "faster" in 2023. CPU speed is plenty today, and it already was 2 years ago.

At the same time, having 1000 times more high-quality commercial and free code.

Yeah, typical bs from someone who doesn't know macos is UNIX and everything is cross-compilable. IT building janitor still thinks there's a thing like "x86 code" in 2023.

Because even now x86 processors are 1.5-2 times faster than the fastest Apple solutions, as proven by tests.

*Desktop* x86 processors faster than *laptop* Apple solutions.

[NikoB]

Users would immediately switch to Arm if it were several times faster.

People will switch to ARM when they see laptop can carry them through transatlantic flight while doing their normal workflow and still being compact and silent. Almost no one gives a f*ck about "faster" in 2023. CPU speed is plenty today, and it already was 2 years ago.

At the same time, having 1000 times more high-quality commercial and free code.

Yeah, typical bs from someone who doesn't know macos is UNIX and everything is cross-compilable. IT building janitor still thinks there's a thing like "x86 code" in 2023.

Because even now x86 processors are 1.5-2 times faster than the fastest Apple solutions, as proven by tests.

*Desktop* x86 processors faster than *laptop* Apple solutions.

Again troll A gets in with his stupid attitudes. Why do companies buy "AI" accelerators if processor speed is already enough for everything?

I don't need a standalone laptop, like many millions of people who don't fly anywhere and don't carry laptops on the streets, but I do need a laptop with the most powerful processor, but at the same time quiet, easy to upgrade and as comfortable as possible in other aspects. And in the 2023 market there are simply no such models with an optimal ratio of quality and price. And we don't need a desktop.

Yeah, typical bs from someone who doesn't know macos is UNIX and everything is cross-compilable. IT building janitor still thinks there's a thing like "x86 code" in 2023.

Stupid troll, right now I'm using software that has always been absent on MacOS and the developer doesn't care about Apple's products, because development on this platform is a priori unprofitable. And this software is extremely dependent on hardware and its performance.
Stupid troll, right now I'm using software that has always been absent on MacOS and the developer doesn't care about Apple's products, because development on this platform is a priori unprofitable. And this software is extremely dependent on hardware and its performance.

You are a janitor at some startup, judging by the fact that even when answering your favorite questions about "AI" (not even from me), you quietly disappeared from the forum for almost a week. =)

[A]

Why do companies buy "AI" accelerators if processor speed is already enough for everything?

Lol, so you really have no idea. They buy AI hardware to _train_ AIs, dummy. _Training_ takes a lot of resources, and _inferencing_ can be ran even on Raspberry Pi (for some models). You probably don't know what inferencing is too though, lol, I don't care, google it up.

but I do need a laptop with the most powerful processor,

Says the guy with cheap a** hardware.

Stupid troll, right now I'm using software that has always been absent on MacOS and the developer doesn't care about Apple's products, because development on this platform is a priori unprofitable. And this software is extremely dependent on hardware and its performance.

Yeah, right, secret software on your secret S-PVA display.

You are a janitor at some startup, judging by the fact that even when answering your favorite questions about "AI" (not even from me), you quietly disappeared from the forum for almost a week. =)

Adventure into the mind of a local clown. Thanks for worrying about me babe. I've been sick, my world doesn't rotate around you or this forum, you know.