Windows for ARM CPUs...

[Silent_Buddha]

Well, looks like it may finally be happening.

http://online.wsj.com/article/SB10001424052748704851204576034051605593000.html

The company next month plans to demonstrate a new version of its widely used Windows operating system that targets low-power devices and adds support for chips based on designs from ARM Holdings PLC as well as the x86 chip technology offered by Intel and Advanced Micro Devices Inc., these people said. Microsoft will discuss the software at the Consumer Electronics Show in early January, though it isn't expected to be available for two years, they added.

The article seems to strongly suggest this will be a derivative of the Windows OS and not WP7.

I wonder if this will just be Windows CE or a more capable but stripped down version of the desktop OS. I'd prefer the later, but doubt Intel based programs would easily run the ARM based version of Windows.

I wonder if I'd be happy with a Windows CE derivative? I'm still not sure I could do everything with CE that I want to do on a tablet.

Regards,
SB

 

[Ike Turner]

Well, looks like it may finally be happening.

http://online.wsj.com/article/SB10001424052748704851204576034051605593000.html



The article seems to strongly suggest this will be a derivative of the Windows OS and not WP7.

I wonder if this will just be Windows CE or a more capable but stripped down version of the desktop OS. I'd prefer the later, but doubt Intel based programs would easily run the ARM based version of Windows.

I wonder if I'd be happy with a Windows CE derivative? I'm still not sure I could do everything with CE that I want to do on a tablet.

Regards,
SB

All this stuff is still speculation at this point. Microsoft still has to release Windows Embedded Compact 7 (CE7: http://www.microsoft.com/windowsembedded/en-us/products/windowsce/compact7.mspx remember that WP7 is still based on the CE6 Kernel) which many in the general press confuse as Windows 7 for ARM. And there'salso Windows 7 Embedded Enterprise, Automotive etc which is the Embedded version of the regular X86/X64 OS etc.. more info here http://www.microsoft.com/windowsembedded/en-us/default.mspx
I would be surprised if MS was actually porting the regular Win7 to ARM giveng that they have CE7 in the starting blocks .

 

[Arun]

Windows 8 will be released in early 2012. ARM has said that they expect ARM-based notebooks (not just netbooks!) shipping in the second half of 2012.

Put two and two together and you get something like my comment on David Manner's blog:

I am willing to bet this is true, and I am willing to bet it will be an ARM version of Windows 8 - to be released very slightly after the x86 version. A couple of points:

1) The last slide from this ARM presentation indicates we will see ARM Notebooks in 2H12: http://www.jp.arm.com/event/pdf/forum2010/t4-2.pdf
2) Steve Ballmer said Windows 8 was Microsoft's "riskiest product bet".
3) There are reliable rumours that Windows 8 will have an App Store, and they'll probably make an ARM version of .NET so all existing managed applications will work.
4) Cortex-A15 adds native hardware virtualisation support and one major focus of Windows 8 will be virtualisation: http://www.zdnet.com/blog/microsoft...microsoft-is-the-next-release-of-windows/7786

The real question is whether ARM chips will be able to run native x86 applications. One very interesting possibility is that this will be handled by a combination of the new virtualisation mechanism and a Mac OS-like Universal Binary system. It's also very interesting that Microsoft took an ARM ISA/Architecture license in July although I don't see what they'd need it for here...

 

[Silent_Buddha]

That could certainly shake up the netbook and slate markets if a slimmed down version of Win8 will be made available for ARM and Intel/AMD CPUs.

Regards,
SB

 

[willardjuice]

Ugh does this mean we are again going to have 32-bit and 64-bit versions of Windows?

 

[Exophase]

Compared to entirely different ISAs it'd seem like 32-bit vs 64-bit would be the least of your concerns..

 

[Blazkowicz]

with ARM CPUs getting powerful I would bet on a full, regular version of windows if this (as a desktop OS) were to come - only issue is that an overwhelming number of people would try to run x86 software, unaware of what a CPU ISA is.

 

[Silent_Buddha]

Ugh does this mean we are again going to have 32-bit and 64-bit versions of Windows?

Gah! Didn't even think of those possible ramifications. Then again if this is limited to the netbook/slate market, it shouldn't impact my beloved PC gaming side of things. Still, I was hoping Win8 would be the death of 32 bit Windows.

Regards,
SB

 

[willardjuice]

Compared to entirely different ISAs it'd seem like 32-bit vs 64-bit would be the least of your concerns..

Yes, of course. I should clarify though, I was speaking more to Arun's question of "if the ARM version of Windows would run native x86 applications". I assume it couldn't run native x64 applications. Thus a x86 version of Windows would seem likely (or at the very least, a focus on x86 applications).

 

[frogblast]

How widespread are applications distributed purely in MSIL form?

I imagine lots new software is written in .Net, but I have no idea if a majority of developers choose to distribute MSIL or x86 to end users.

 

[Blazkowicz]

on a tangent subject, I would like amd64 windows to be able to run all x86 applications. namely 16bit ones, which would require a CPU emulator as much as an x86 application trying to run on ARM.

this is one of the big reasons for the 32bit version of windows 7 : people or business actually caring about running some DOS or Win16 applications.
surely this would not require much effort - update the ntvdm so it runs an emulator. I can't understand why it wasn't implemented in the first place years ago.

 

[rpg.314]

My guess is that they will include dynamic binary translation to run x86-win apps. With all the warts of x86, it should be easy peasy.

 

[Squilliam]

Yes, of course. I should clarify though, I was speaking more to Arun's question of "if the ARM version of Windows would run native x86 applications". I assume it couldn't run native x64 applications. Thus a x86 version of Windows would seem likely (or at the very least, a focus on x86 applications).

It'd be a good justification for a native application store for Windows if they did that. Maybe they would simply direct people down that route rather than have them search the dregs of the interwebs. Hell it'd probably be safer for your mom and dad browser than using google plus internet explorer.

 

[wco81]

with ARM CPUs getting powerful I would bet on a full, regular version of windows if this (as a desktop OS) were to come - only issue is that an overwhelming number of people would try to run x86 software, unaware of what a CPU ISA is.

Yeah the advantage of Windows is the legacy software support.

So what is the advantage of Windows on ARM exactly?

And are ARM CPUs going to be more powerful than x86 CPUs or just more power efficient on full-fledged laptops?

Or is it more of a tablet play?

 

[Exophase]

MS has an ARM arch license so who knows what market segment they're shooting for with a new CPU design, but obviously they don't have the experience to compete with Intel or AMD in raw peak performance.

The advantage of Windows to us is software support, but I would fully expect that for the average user just being able to stay within a familiar environment does actually mean something. Especially given the sort of things many users spend the most time doing, like browsing the web and watching videos/music; things that they'd likely use standard packaged software for.

It also gives MS one less new OS to market and maintain, for the ARM devices that don't fit under Windows Phone. This could be a strategic play for the tablet market which still hasn't really embraced anything outside of iOS yet.

Binary translation is doable, and by controlling the operating environment they gain a huge advantage in terms of stuff they don't have to emulate, very important for efficiency (think qemu system vs qemu user mode.. and so far no one has done qemu user mode + wine on ARM, at least not my knowledge, although it'll happen sooner or later). MS's acquisition of Connectix, who did some really good recompilers, could pay off here. MS also has a huge advantage over the WINE team in getting faster DirectX support, not least of all because they'd need actual DirectX GPU drivers (IMG at least is intent on supporting this, probably others). You definitely shouldn't expect 90% native performance on most real world apps though, that's way too optimistic.

 

[Blazkowicz]

a good point I've seen on Slashdot is that a port of windows would get used on ARM servers.
this makes a lot of sense, as you mostly run 1st party stuff (active directory, dhcp, dns etc.), java, web or .net stuff (those are CPU agnostic in theory), or some open source or big name software.

so you're much less likely to wish to run some software that was compiled 13 years ago, unlike on desktop windows - plus it's easy to buy or rent an additional x86 server if you need to.

PC hardware is also damn overkill - any Xeon or Opteron is ridiculous when you want a file server, print server, domain for a dozen or two users, etc.

 

[Deleted member 13524]

I fear so much the kind of fragmentation this could cause..

That and Mr. John Doe wondering why his good old office 2007/Civ IV game doesn't work with his 4-core Cortex A15 1.5GHz w/ Windows 8 laptop.

 

[rpg.314]

I fear so much the kind of fragmentation this could cause..

That and Mr. John Doe wondering why his good old office 2007/Civ IV game doesn't work with his 4-core Cortex A15 1.5GHz w/ Windows 8 laptop.

I think MS is more aware of this than anyone else. Whatever they are going to do, I am sure they will keep this in mind.

In this light, the ARM arch license takes on quite a different meaning.

 

[Exophase]

I agree, this could explain better why MS wanted an arch license.

It'd be cool if they designed an ARMv7+ CPU with coprocessors extensions that are useful for x86 binary translation. Something like the extensions Godson 3 added. This could potentially offer uses for emulation outside of x86 as well, for instance if it has hash lookup instructions.

 

[rpg.314]

I agree, this could explain better why MS wanted an arch license.

It'd be cool if they designed an ARMv7+ CPU with coprocessors extensions that are useful for x86 binary translation. Something like the extensions Godson 3 added. This could potentially offer uses for emulation outside of x86 as well, for instance if it has hash lookup instructions.

The co processor approach, while attractive on surface, cannot explain the need for an arch license imo.