Windows 8 ARMED!

Bleh, old news man, you're about 2 weeks late to the party. :D

Now, if someone would just leak a video we'd be all set...
 
heres to hoping that along with the new multiplatform release, the next windows version breaks compatibility/dumps legacy cruft even in the x86 release....

ARM compatibility for the next version of windows really seems to gel well with the rumors of MS utilizing heavy virtualization in the next version of windows... color me intrigued ....
 
Joe Average said:
Bleh, old news man, you're about 2 weeks late to the party. :D

Now, if someone would just leak a video we'd be all set...
Click to expand...

Well it was just officially announced, the press release was today from Microsoft so no not late, just official. And I wouldn't be shocked if Windom 8 Beta just happens to be released around the Honeycomb time frame.
 
brucedeluxe169 said:
heres to hoping that along with the new multiplatform release, the next windows version breaks compatibility/dumps legacy cruft even in the x86 release....

ARM compatibility for the next version of windows really seems to gel well with the rumors of MS utilizing heavy virtualization in the next version of windows... color me intrigued ....
Click to expand...

I do think that this is the end of the Win32 API with .Net FINALLY taking its place as the API for Windows but Microsoft will be forced to maintain a high level of compatibility, while that can be a drag it is one of Windows' great strengths.
 
heatlesssun said:
Well it was just officially announced, the press release was today from Microsoft so no not late, just official. And I wouldn't be shocked if Windom 8 Beta just happens to be released around the Honeycomb time frame.
Click to expand...

Within the next 3 months? That's gonna be interesting... the Motorola Xoom, the Asus Eee Pad, all those are supposed to be shipping with Android 3.0 from the factory and they're all supposed to be out by April, per the current estimates based on what was announced at CES this week.

Honeycomb looks pretty nice from that one leaked video (if it's "official" and actually produced by Google and I still have my doubts there), and with most every one of these new devices being "dual core" <whatever, Tegra 2, etc> that should be more than adequate to handle most anything people can throw at 'em.
 
Thanks, almost forgot about this, this is his biggest CES presentation ever in terms of his carrer. Looks like Microsoft's roadmap is complete in the mobile space and Windows 8 will blow the doors off of even Windows 7 in terms of deployment numbers if the ARM stragetgy works like it should.

This might not be a bad time to buy Microsoft stock.
 
We'll if you want video it's there. Ballmer and friends showed the ARM version Windows running on the big three major ARM platforms using the Windows 7 interface and even Office 2010 running on it. They didn't show ANY next gen UI or touch stuff but you can bet that's going to borrow HEAVILY from Windows Phone.

This is just fucking a game changer and is the first real threat that Microsoft has put up to Android and boy is it a big one.
 
That was very interesting seeing Power Point and Windows media player showing 1080p movie trailers running on ARM processors.

I know the release of windows 8 is likely 2 years away but I have a few questions which I think I would love ask the windows 8 development team. What are you thoughts?

Given the huge variety of hardware devices windows 8 can run on, I wonder how this Windows 8 performance will scale with the phones, tablets, and PC's? i.e. could you run photoshop slowly from your phone by hooking it up a large monitor (your TV) through HDMI, given appropriate accessories and enough RAM on the phone, that is.

Will there be a Windows Application store that will have both x86 and ARM versions of software? If you buy an x86 version, will it cost extra to get the ARM compiled version?

Hopefully drivers can be worked out by release of Windows 8 for ARM and x86 versions better than they were for Vista. I will keep my fingers crossed.
 
As the tier-1 defense line is breached against the incoming XXX,

1. Windows on ARM is not about absolute performance.

2. It is about keeping RDF and Google honest. ARM, x86-32/64, now everyone is even.

3. To guard Office, both to protect and expand, a little development effort is nothing compare to the implication of Office.
 
El Barto said:
That was very interesting seeing Power Point and Windows media player showing 1080p movie trailers running on ARM processors.

I know the release of windows 8 is likely 2 years away but I have a few questions which I think I would love ask the windows 8 development team. What are you thoughts?

Given the huge variety of hardware devices windows 8 can run on, I wonder how this Windows 8 performance will scale with the phones, tablets, and PC's? i.e. could you run photoshop slowly from your phone by hooking it up a large monitor (your TV) through HDMI, given appropriate accessories and enough RAM on the phone, that is.

Will there be a Windows Application store that will have both x86 and ARM versions of software? If you buy an x86 version, will it cost extra to get the ARM compiled version?

Hopefully drivers can be worked out by release of Windows 8 for ARM and x86 versions better than they were for Vista. I will keep my fingers crossed.
Click to expand...

Excellent questions and I was just thinking about the ARM and x86 app thing. I have little doubt that most apps will make you pay twice, definately not Office. Something like Photoshop, which I really don't see happening on ARM soon most definately.
 
Apparently what they demoed wasn't truly ARM, it was still x86 based code being interpreted by an ARM processor so... boo on them (this according to something that Engadget posted after the presentation as an answer to an engineer at the show).

But, even so, the fact that they're going to move a desktop OS towards a mobile CPU and hopefully do it properly could be a whole new ballgame. They can't just port the whole damned OS over - they have to do it from scratch, designed for the platform it's going on, or it'll be a waste of time.

Yes, personal opinion again but, dammit, I haven't been proven wrong yet. :D
 
Joe Average said:
Apparently what they demoed wasn't truly ARM, it was still x86 based code being interpreted by an ARM processor so... boo on them (this according to something that Engadget posted after the presentation as an answer to an engineer at the show).
Click to expand...

This is NOT what the guy doing the demo said. He CLEARLY stated that this was a native ARM client more than once, they even when out of the way to print a document saying that the printer drivers were recompiled for ARM, this would be pointless with a translation layer. Microsoft isn't stupid, they've probably had this working for a LONG time and decided that this would be their tablet strategy and probably the best one they could bring to bear.
 
Joe Average said:
Apparently what they demoed wasn't truly ARM, it was still x86 based code being interpreted by an ARM processor so... boo on them
Click to expand...


If that was true, it would be HUGE. A real time x86 to ARM emulator? No way.
Boo on you my friend.
 
Stoly said:
If that was true, it would be HUGE. A real time x86 to ARM emulator? No way.
Boo on you my friend.
Click to expand...

Hmm, why? x86 on alpha worked, powerpc on x86 worked, so why can't x86 on arm work?
 
Stoly said:
If that was true, it would be HUGE. A real time x86 to ARM emulator? No way.
Boo on you my friend.
Click to expand...

You do realize that Android runs on top of a VM right? Java code otherwise cannot execute on ARM hardware.
 
I wonder if MS would implement a remotefx client so you could run x86 applications (and games) from a server or the cloud.
 
01-05Sinofsky_page.jpg


Does that remind anyone of Picard?
 
That's pretty amazing (Stoly's youtube link). Good to see MS being 'hungry', that's when they do their best.
 
Stoly said:
realtime?
Click to expand...

Given that you can emulate a complete x86 PC on ARM (or indeed anything else with a linux port) with bochs (admittedly slow, but it works), running a single executable and handling the syscalls it makes should be well within the realm of possibility (and much faster). Not that you need to do it that brutally; you can do on-the-fly translating, or pre-translating, or something in between.

Looking at NX!32 (the x86-on-alpha tool for NT 4) , it actually did an interesting combination: It had an x86 emulator fast enough to be usable, and used that the first time you started something. While running, it captured profiling data, and when the app exited, it used that to generate translated code for the parts that had been executed; those were then used on next launch, falling back on emulation for the rest (if any). Assuming I'm reading my source right, of course. The idea was that the translation was fairly slow, so it should be done when it wouldn't interfere with the running program.

I don't know too much about Rosetta (PowerPC-on-x86 for OS X), but it seems to do on-the-fly translating (with caching - no need to translate the same code twice); there's more CPU power (and cores) around in current hardware, so it's not such an overhead any more.

However, it's worth noting that both of these let you run software on a faster platform - PentiumPro -> Alpha, and PowerPC G4/G5 -> Core2 were both steps up in speed, which helped the non-native software feel decently fast even though it was slowed down by the translation/emulation. Going from an i7 to some ARM CPU won't necessarily have the same buffer to fall back on. Which is why recompiling is more sensible, if possible - and it usually is. If you've got decently clean code it might even be easy.

Or, alternatively, you can write java or .net, and if you don't call any native libraries (JNI and whatever it's called in .net) it should Just Work as soon as the runtime has been ported.
 
Last edited:
While WinARM is what is capturing headlines, IMO that is incidental and of no more Consequence than Windows NT on PowerPC was. It can be seen like those efforts as hedging, or lighting a fire under Intel, but ultimately it is not the critical work.

The real story here is Windows on SoC, that includes ARM, but it also includes Intel x86 SoC, which will be the real important one.

Before I started digging into this news, I thought that Windows would just install/run on Intel x86 SoC. It doesn't. The Intel SoC makes a significant divergence from typical PC architecture and requires a change in the lower level interfaces. These are must do changes for Microsoft and will be similar to what is required to support ARM SoCs (so a decent time to hedge).

From the CES presentation, MS was touting a big deal feature of the Intel SoCs, compatibility with the Windows Legacy and this is why these SoCs will be of major importance and ARM SoCs more of an afterthought/hedge.

Other consequences. Windows Embeded Compact (WinCE) for tablets is DEAD. In the summer there were prototypes of this making the rounds. This effort is completely shut down. It is also likely that at some point WinCE on phones will also be dead (AKA WP7). I don't mean the end of WinPhone, but that at some point, perhaps WP9 will be NT kernel based, rather than WinCE based.

So WinNT everywhere is really the message here and the end of winCE on consumer devices.

NT on ARM will likely only be a force in phones branded as WinPhone where all the code is managed VM code and can run easily on ARM/x86.

Laptops/Netbooks/Tablets will all remain essentially x86 for the massive legacy advantage.
 
Stoly said:
If case you haven't seen it

http://www.youtube.com/watch?v=xKc_XGuvNIk

Its a demo of windows 8 on intel, OMAP, Snapdragon and Tegra.

The presenter clearly says its NATIVE ARM OS. running NATIVE ARM apps like word, and power point (both were recompiled.).
Click to expand...

Reminds me of a devkit unit. They recreated Windows on arm using an x86 as a dev unit, and then executed on the ARM units. So it's basically their version os "Hello World" :D

What interests me is the Word application and how he said "re-compiled", which implies that he took an existing source code and simply reused it. That significantly lowers the learning curve.
 
Sly said:
What interests me is the Word application and how he said "re-compiled", which implies that he took an existing source code and simply reused it. That significantly lowers the learning curve.
Click to expand...

Well, if the API is essentially the same (and I would expect it to be), porting is a case of recompiling and checking if you've made any unsafe assumptions about the architecture (byte ordering, alignment, that sort of thing). It's the same issue mac developers had going PPC to x86, or anyone going i386 to amd64, or anyone compiling linux apps for different architectures: I would have been deeply surprised if they couldn't simply reuse the code, apart from any small bugfixes they had to make.
 
HHunt said:
Well, if the API is essentially the same (and I would expect it to be), porting is a case of recompiling and checking if you've made any unsafe assumptions about the architecture (byte ordering, alignment, that sort of thing). It's the same issue mac developers had going PPC to x86, or anyone going i386 to amd64, or anyone compiling linux apps for different architectures: I would have been deeply surprised if they couldn't simply reuse the code, apart from any small bugfixes they had to make.
Click to expand...

You shouldn't trivialize the effort though. I work as a programmer and we sink a lot of man hours even when merely changing compilers, and that is when hitting the same targets.

Changing CPU Architecture on very large, very old codebase will be a very significant effort, revealing subtle bugs, and requiring extensive testing. It is seldom re-compile and go except for small, trivial code. IMO when they said recompiled, they didn't imply it was trivial, they just wanted to make clear it wasn't emulation, that it was ARM native.

WinArm for desktop/laptop/netbook/tablets is unlikely to rise beyond the level of curiosity because it has the Massive disadvantage of not working with the huge Legacy library of Windows apps, most of which will NEVER be ported.

By the time Win8 (First verison to run on SoCs), Intel SoCs should be perf/watt competitive with ARM SoCs.

Then your choice is Win8Intel, that runs all old applications and Win8Arm that doesn't. Why would you buy a Win8Arm machine?
 
Snowdog said:
You shouldn't trivialize the effort though. I work as a programmer and we sink a lot of man hours even when merely changing compilers, and that is when hitting the same targets.

Changing CPU Architecture on very large, very old codebase will be a very significant effort, revealing subtle bugs, and requiring extensive testing. It is seldom re-compile and go except for small, trivial code. IMO when they said recompiled, they didn't imply it was trivial, they just wanted to make clear it wasn't emulation, that it was ARM native.

WinArm for desktop/laptop/netbook/tablets is unlikely to rise beyond the level of curiosity because it has the Massive disadvantage of not working with the huge Legacy library of Windows apps, most of which will NEVER be ported.

By the time Win8 (First verison to run on SoCs), Intel SoCs should be perf/watt competitive with ARM SoCs.

Then your choice is Win8Intel, that runs all old applications and Win8Arm that doesn't. Why would you buy a Win8Arm machine?
Click to expand...

Mmh, it does of course depend on the code. I suspect they shook out a bunch of the unexpected x86-dependencies getting office ready for amd64, but I'm still glad I didn't have anything to do with it. I slightly overreacted to how incredulous the guy I quoted seemed ("What, they can compile a windows app for another architecture? The shock!") :)

I'm personally sceptical to intel-SoC being as good as they claim - intel has had a very long time to make ARM-beating mobile cores, and their hype has so far not matched up to the results. Still, I don't think ARM will take over the desktop/laptop role, for much the same reasons as you. It might happen, if the ARM systems are so fast and attractive that dynamic translation (Rosetta-style) is worth it ... but I wouldn't bet on it.

However, I think Win8ARM will stick around; there are a few possible niches for it. It would work fine on a HTPC (something like a more open Apple TV), or perhaps as a replacement for WinCE and the new WinMobile, if they want to go the "one OS to rule them all"-route. Oh, and it could lead to some rather neat WHS hardware.

We'll see, I guess.
 
HHunt said:
Mmh, it does of course depend on the code. I suspect they shook out a bunch of the unexpected x86-dependencies getting office ready for amd64, but I'm still glad I didn't have anything to do with it. I slightly overreacted to how incredulous the guy I quoted seemed ("What, they can compile a windows app for another architecture? The shock!") :)
Click to expand...

I do think amd64 has a lot more commonalities, and remains backward compatible. ARM is a RISC chip of a completely different family. More work seems inevitable.

I'm personally sceptical to intel-SoC being as good as they claim - intel has had a very long time to make ARM-beating mobile cores, and their hype has so far not matched up to the results.
Click to expand...

I wouldn't bet against Intel. They stumble from time to time, but then they get serious and devote more resources and change direction if necessary. They are also tops on the process side. The will probably have a 22 nm SoC before ARM processors make that Jump. Last I saw Intel was aiming for 32 nm Medfield by late 2011 on cell phones. While it might not be the best cell phone chip, it does show a significant drop in power if they can get an x86 cellphone out there.
They are reported at the customer Sampling stage with Medfield:
http://www.xbitlabs.com/news/mobile...s_to_Sample_Next_Gen_SoC_for_Smartphones.html
"Intel believes that with lower-power Medfield it will register design wins and after that will be able to offer both lower power consumption and greater performance than ARM-powered SoCs. "

Naturally Intel wants to paint it's own products in a good light, but I don't think they are going to miss that wide. While they may not be quite enough to really gain cellphone wins, I think they will be more than competetive enough on other portables(netbooks/tablets), that the Legacy windows code advantage will ensure the win. Medfield will certainly ship before Win8Arm.

Still, I don't think ARM will take over the desktop/laptop role, for much the same reasons as you. It might happen, if the ARM systems are so fast and attractive that dynamic translation (Rosetta-style) is worth it ... but I wouldn't bet on it.
Click to expand...

Yep. I certainly wouldn't. This effort is clearlly aimed at SoCs, they don't have power to spare for emulation and even Atom is more powerful than ARM SoCs and this isn't a transition, it is parallel. Apple was jumping to an (arguably) more powerful chip, and it was a clearly defined transition, so emulation quickly improved to the point that it was almost as fast as the old PPC HW and that old PPC HW was effectively frozen. Here you want to emulate already faster HW that is a moving target and there is no sign that MS wants to do any transition or emulation work. I think this will be a stark choice. Intel SoC with legacy support, ARM SoC with none. Anywhere Legacy is important, Intel will win.

However, I think Win8ARM will stick around; there are a few possible niches for it. It would work fine on a HTPC (something like a more open Apple TV), or perhaps as a replacement for WinCE and the new WinMobile, if they want to go the "one OS to rule them all"-route. Oh, and it could lead to some rather neat WHS hardware.
Click to expand...

I do think this announcement is definitely the death knell for WinCE. I think WinOnArm will only succeed in the one place where the NT legacy doesn't matter. Phones.

Yet I can just as easily imagine a new class of WinPhones, that run standard WinPhone managed code software, and are based on x86 with some support for full legacy windows applications. Your phone could be your only computer. Just plug it into a docking station and use full sized applications.
Something like this, but when you plug it in, you get desktop windows, desktop Office.
http://www.anandtech.com/show/4112/anand-goes-hands-on-with-motorolas-atrix-4g-webtop
 
Keep in mind nvidia's "we will make an ARM CPU fast enough for computers, workstations, high-performance servers" - announcement, though: They claim they'll make something fast enough to be competitive. I'm mildly sceptical, but even if they fall short it should be well faster than the current SoCs. :)
 
HHunt said:
Keep in mind nvidia's "we will make an ARM CPU fast enough for computers, workstations, high-performance servers" - announcement, though: They claim they'll make something fast enough to be competitive. I'm mildly sceptical, but even if they fall short it should be well faster than the current SoCs. :)
Click to expand...

Then they won't be competing with Atom, they will be competing with i7 and beyond.

They might get some odd niche Wins, but Nvidia ARM on the destkop against i7?

Good luck with that. AMD has been at it for decades and can't match i7.

Back on WinNT everywhere. I still have big concerns on the low end. WinNT on Atom with a 1GB of Ram is a sluggish and slow, iOS/Android on ARM with 512 MB flies. Atoms are more powerful than current ARM SoCs. WinNT just doesn't seem efficient enough to really scale to the low end and replace WinCE.

It will definitely be interesting to watch this unfold.
 
Last edited:
Snowdog said:
Then they won't be competing with Atom, they will be competing with i7 and beyond.

They might get some odd niche Wins, but Nvidia ARM on the destkop against i7?

Good luck with that. AMD has been at it for decades and can't match i7.

Back on WinNT everywhere. I still have big concerns on the low end. WinNT on Atom with a 1GB of Ram is a sluggish and slow, iOS/Android on ARM with 512 MB flies. Atoms are more powerful than current ARM SoCs. WinNT just doesn't seem efficient enough to really scale to the low end and replace WinCE.

It will definitely be interesting to watch this unfold.
Click to expand...

It's definitely going to depend on the price and performance of whatever nvidia launches - there's no theoretical reason they can't launch something a bit faster than atom with comparable power draw, or for that matter something much faster with a power draw on the i5/i7 scale - and that would be downright fun.

As for competing in speed - ARM isn't x86/amd64, and it has had a lot of optimization pumped into it to make it run fast on very little energy. No one has really tried scaling ARM to mainstream performance since Acorn in the mid 90s, so it's in the blue how it'll perform if you really throw silicon and power at it. Maybe it'll be easier to scale up than amd64? Maybe it'll be nigh-impossible? I'm actually excited; it would be hilarious if the end of the x86-kills-everything period was ARM of all possible things. (It's the last legacy of the British computer industry, too - for added weirdness. :D )

Before we start comparing NT and Android I'd really love to see them tested on the same hardware; from the looks of things that'll have to wait until someone gets hold of both an NT-supported ARM device and an android port to the same ...
(NT itself is probably fine, but it drags along more subsystems and a larger feature set than what you can use in an android app - as for exactly how much it matters, well.)
 
Last edited:
Snowdog said:
By the time Win8 (First verison to run on SoCs), Intel SoCs should be perf/watt competitive with ARM SoCs.
Click to expand...

Doubtful. ARM was built specifically for low-power use, x86 was not. x86 has been moving towards low-power in the last few years, but it is still nowhere near ARM.

Example: ARM's (currently available) dual-core power-optimized Cortex-A9 SoC consumes ~0.6mW per MHz, or about 8 DMIPS / mW.

Intel's (recently announced) single-core, dual-threaded E660T SoC uses ~2.5mW per MHz, and is the neighborhood of 0.9 DMIPS / mW.

Yes, you can't directly compare MHz to MHz and IPS to IPS between ARM and x86, but this should illustrate the distance Intel has yet to go to catch up with ARM when it comes to power efficiency. (If anyone has a more appropriate method of cross-platform power-efficiency comparison, please let me know.)

I should also mention that ARM SoCs are a fraction of the size of the Intel's... try under 5mm^2 for the whole Cortex-A9 SoC, while Intel's E6xx are 22mm^2, and that's not counting the 23mm^2 PHC.

ARM isn't slowing down either; Cortex-A15 isn't too far away and boasts further reductions in feature size and energy consumption.

EDIT - Also worth noting, Cortex-A15 will provide full hardware virtualization support. x86 VM anyone?
 
Last edited:
phide said:
Is this the optimal approach, in your opinion? The same operating system on every platform regardless of the platform's performance profile?
Click to expand...

I think that modern software engineering teaches us that it SHOULD be the optimal approach. A good OS should be reusable, scalable, platform indepedent and have clear separation of concerns. Of course I know that many wouldn't say that Windows isn't a good OS. Just a little joke, I'm not saying that's you.;)

The power of this is clear. Say I'm working on my desktop in Word or Excel, I need to go but continue my work, I hit a button to instantly transfer what I was doing to Word or Excel Tablet. The same functionality at its core, using much of the same code base but with a different UI, the good old model-viewer-controller pattern. This is even better than the typical cloud solution as that's just running the same client in a browser.

I don't see why this isn't possible and unless I really missed the boat on software design this actually a highly desirable approach though it's definately not easy.
 
Electrofreak said:
Doubtful. ARM was built specifically for low-power use, x86 was not. x86 has been moving towards low-power in the last few years, but it is still nowhere near ARM.
Click to expand...

I think that this is a big part of why Microsoft had to do this or have something ARM based though they could have done WP7 though I imagine in time Windows Phone will become Windows on a phone.

There's simply too many ARM processors out there and having no Windows on them simply isn't acceptable if you're Microsoft. One of the great things about Windows is it incrediable hardware support. This isn't Microsoft being bold as much as it's Microsoft doing what it's good at but there's a lot of risk involved. Now the next risky thing is the UI for Windows ARM but it's starting to sound like that's going to be very much WP7 Metro based which makes a lot of sense as that UI is very good for touch as seems to be well regarded.
 
Electrofreak said:
Doubtful. ARM was built specifically for low-power use, x86 was not. x86 has been moving towards low-power in the last few years, but it is still nowhere near ARM.

EDIT - Also worth noting, Cortex-A15 will provide full hardware virtualization support. x86 VM anyone?
Click to expand...

No doubt the current solution is far behind in Perf/Watt. AFAIK, Intels SoC is still 45nm for CPU and 65nm on the integrated GPU. Medfield is where they close the gap significantly. Note that neither I nor Intel is claiming Medfield will beat ARM just that the gap will close significantly. IMO that will be enough to keep ARM out of windows netbooks/laptops. Is an hour of battery life worth giving up all the Windows Legacy software?

BTW Virtualization isn't instruction set emulation. ARM as optimized interger machine (Also why DMIPS is problematic) and would barf running emulation of Intel floating point code.

This may seem like a big deal, but ARM windows will pretty much be on Phones. This is more about getting NT kernel in phones, than ARM Windows desktops.

Edit:
Note Intel has been delivering on announced targets, they realize they aren't going to catch up overnight(AFAICT Intels first x86 SoC was in 2009). If you look back at IDF 2009, they were predicting Moorestowne, with a drastic cut in idle and Medfield on 32nm giving cuts all around in 2011. Given they are already sampling, it looks like they are still on target:
http://www.techradar.com/news/world-of-tech/bullish-intel-bucks-the-recession-637735
 
Last edited:
You must log in or register to reply here.
Back
Top