64-bit ARM support for Windows is in the works

This should come as no surprise, but it’s noteworthy nonetheless. The folks at PC World have gotten confirmation that ARM is working with Microsoft on Windows support for its 64-bit architecture. Details are few and far between right now, but ARM Program Manager Ian Forsyth told the site, "ARM works with all its OS and ecosystem partners to inform them on next generation technologies and enable their support."

As we learned earlier this week, ARM’s first 64-bit cores, the Cortex-A50 series, are scheduled to ship inside next-generation system-on-a-chip devices in 2014. The cores are based on the ARMv8 architecture, which introduces 64-bit memory addressing and a new instruction set.

Licensees of the Cortex-A50 series include AMD, which announced on Monday that it’s going to offer ARM-based Opterons in 2014. The company hasn’t said anything about ARM-based consumer chips yet, but if Windows support is in the cards, I expect we may see some ARM-powered APUs eventually. Such APUs could fuel not just tablets, but also full-sized notebooks and desktops.

That said, even with newer and faster ARM-based processors, Windows software compatibility could remain a thorny issue. Windows RT lacks support for x86 software, and as Pund-IT analyst Charles King tells PC World, "From a purely technical perspective, porting many common x86 applications to ARM is problematic." That could impede adoption regardless of how appealing the hardware turns out to be.

Comments closed
    • rephlex
    • 7 years ago

    [quote<]Windows RT lacks support for x86 software, and as Pund-IT analyst Charles King tells PC World, "From a purely technical perspective, porting many common x86 applications to ARM is problematic."[/quote<] Why? And is it even necessary? What about the [url=http://en.wikipedia.org/wiki/FX!32<]FX!32[/url<] approach?

      • bhtooefr
      • 7 years ago

      Performance.

      Cortex-A15 is just now catching up to Core 2 at the same clocks when running native code.

      Even in a best-case scenario, FX!32 was about half as fast as the native code on an Alpha. That was fine, because an Alpha was twice as fast as a contemporary x86. (Worst-case scenario, FX!32 was a tenth of the speed, or so. Still, there were plenty of x86s that fast in daily use, so it was acceptable.)

      So, asking an ARM processor to run x86 code would be quite disastrous for performance. We’re talking Pentium II or early Pentium III level performance on a good day, and early Pentium I performance or maybe even 486 performance on a bad day.

    • chuckula
    • 7 years ago

    n/m

      • oldog
      • 7 years ago

      ???

        • chuckula
        • 7 years ago

        It was a rant about the Nexus 4 vs. the Motorola I. I need to think on it some more.

    • HisDivineOrder
    • 7 years ago

    [quote<]That said, even with newer and faster ARM-based processors, Windows software compatibility could remain a thorny issue. Windows RT lacks support for x86 software, and as Pund-IT analyst Charles King tells PC World, "From a purely technical perspective, porting many common x86 applications to ARM is problematic." That could impede adoption regardless of how appealing the hardware turns out to be.[/quote<] Plus, there's the little fact that no one can buy Windows RT separately from a device and it cannot be installed separately, either. That might have some impact on the future of ARM on desktop computers like we're used to.

    • ronch
    • 7 years ago

    I wish chip design companies would disclose more information today just as they did in the past. This would include not just the internal workings of the processor die, but the cores themselves. Intel has been the least secretive lately, ironic considering they have the performance crown. Come on ARM, AMD, MIPS, IBM…

      • chuckula
      • 7 years ago

      ARM: we’ll see (fine for performance in my smartphone, but don’t expect miracles beyond that).
      AMD: If you want performance, hope for improvements in x86 kuz otherwise it’s the exact same “we’ll see” as ARM.
      MIPS: You know what’s funny? Talk to a MIPS guy and he’ll tell you what a bloated power hungry architecture [s<]x86[/s<] [u<]ARM[/u<] is.... (MIPS is actually attacking ARM from the *bottom*) IBM: Doesn't care. Sure there's Power7, but the price differential for that hardware makes the difference between a high-end Xeon server and your smartphone look like a rounding error.

      • willmore
      • 7 years ago

      It’s pretty easy to follow along with ARM chips–at least the processor core is pretty well documented. Check out the ARM website. AMD and Intel provide a lot of fairly low level info, but you really have to wade through 1K+ page tomes. Generally multi-volume sets of them.

    • bthylafh
    • 7 years ago

    My question: Assuming WinRT is successful enough not to kill, will current devices be able to upgrade to the next version, for free or otherwise?

    If they can upgrade, how long will Microsoft produce both 32- and 64-bit versions for backwards compatibility?

    edit: and how easy will it be for devs to offer binaries of both types in the Store? Will 64-bit WinRT be inherently capable of running 32-bit code?

      • willmore
      • 7 years ago

      Some windows devs will have to come to my rescue here, but aren’t apps from the Windows Store not really ‘binaries’ in the sense that they’re directly executable x86 code, but some sort of java bytecode like thing?

        • Klimax
        • 7 years ago

        Only those written in .Net language using bytecode. C++(with or without CX) is directly compiled into machine code. Store will then filter apps by supported arch.

          • willmore
          • 7 years ago

          Thank you Klimax!

      • JustAnEngineer
      • 7 years ago

      Old tech hangs around for a long long time, mostly due to laziness and inertia.

      Microsoft is still making 32-bit Windows for x86, almost a full decade after AMD’s Opteron brought us the far superior AMD64 option.

      Intel is still selling 32-bit x86 processors and even developing new ones, too. 🙁

    • jjj
    • 7 years ago

    Doubt we’ll see AMD go ARM in consumer soon. I got the feeling the first gen will be more of a pilot program and maybe they push their own custom ARM core in 2015-2016 and attack more markets with it.They can’t afford to invest too much too early.

    • Arag0n
    • 7 years ago

    I can see an ARM AMD CPU at almost no cost and crazy low power being a challenge for Intel pretty soon. Remember guys, ARM CPU’s usually cost less than 20$ because development costs are almost non existent for the seller and the designer, ARM Holdings gets money from every single CPU they sale in several thousand million devices and it offsets the inexpensive price.

      • chuckula
      • 7 years ago

      [quote<]ARM Holdings gets money from every single CPU they sale in several thousand million devices and it offsets the inexpensive price.[/quote<] Yeah... ARM Holdings gets a small licensing fee. Where is AMD's money going to come from though? P.S. --> If these chips are at almost no cost, it also means that AMD can't spend any money on differentiating themselves from their competitors with unique designs. How is AMD going to compete against Samsung, who owns its own fabs and can afford to make the chips at a loss while making up for it in profit margins for full products like smartphones/tablets/etc?

      • sunner
      • 7 years ago

      “I can see an ARM AMD CPU at almost no cost and crazy low power being a challenge for Intel pretty soon. …”

      It would be nice to see Godzilla Intel running scared for a change.

      • NeelyCam
      • 7 years ago

      [quote<]I can see an ARM AMD CPU at almost no cost and crazy low power being a challenge for Intel pretty soon.[/quote<] I fact, they will be $2 MINUS a $10 mail-in rebate for your next purchase. And they generate energy. [quote<]ARM CPU's usually cost less than 20$ because development costs are almost non existent for the seller and the designer, ARM Holdings gets money from every single CPU they sale in several thousand million devices and it offsets the inexpensive price.[/quote<] Magically, ARM+AMD has figured out how to develop chips for practically free. It's just too bad AMD didn't know this five years ago.. for years they [i<]wasted[/i<] money developing CPUs. They could've crushed Intel by now. Nonetheless, things are good now, and it also helps that they have a deal with TSMC where they get silicon at -50% below cost. And, AND.. the Taiwanese government will pay for the process development! This is sweet! Rory and Warren for President!

        • chuckula
        • 7 years ago

        So what you are saying is that I can by about 10,000 of these chips for the low-low price of free and power my home with them? ARM is so great at freeing us from those stupid “legacy” laws of physics that Intel has shackled us with with their 1970’s Pinto chips!

          • NeelyCam
          • 7 years ago

          IT’S THE AGE OF NEW PHYSICS, AND ARM IS BLUE ENERGY

          WHAT DO YOU THINK POWERS ALL THOSE WARP DRIVES IN STAR TREK..? FISSION? PATHETIC!

            • chuckula
            • 7 years ago

            [quote<]AND ARM IS BLUE ENERGY[/quote<] I don't know what's worse... that you watched that gawd-awful remake of V, or that I know that you watched it because I watched it too. Oh wait.. neither one is worse. YOUR ABUSE OF CAPS-LOCK IS WORSE.

      • khands
      • 7 years ago

      We probably won’t see these in anything other than Seamicro products for a while, so the cost will be hidden, but I doubt they’ll be less than $200 when they launch retail.

    • sunner
    • 7 years ago

    PC is changing so dang fast its almost scary.

      • Jigar
      • 7 years ago

      If only Intel can kick those watts down… Just remember if by any chance that happens, ARM will face their worst nightmare come true, performance wise Intel can beat ARM any time any where.

        • Goty
        • 7 years ago

        All evidence to the contrary notwithstanding (e.g. mobile Atom vs A15 performance numbers).

          • chuckula
          • 7 years ago

          Can you show us some mobile Atom vs. A15 numbers? The only ones I’ve seen for the A15 are in a Chromebook vs. an old 45nm Atom. The A15 wins at performance, but the power envelope is much much greater than a phone and even larger than is acceptable for most tablets.

          Do you have a link for performance/power benchmarks between the A15 in a lower mobile power envelope vs. Clovertrail or Medfield? I’m assuming the A15 will win, but I’m not sure what the margin of victory will be or if the A15 has a mixed profile (i.e. wins at performance but burns more power or vice-versa).

          Edit: For example, in recent benchmarks of the dual-core Krait Motorola Razer M vs. the Medfield Razer I, the Krait wins most performance benchmarks by ~10-20%, but the Intel phone had longer battery life. The phones are pretty much identical except for the SoC.

            • designerfx
            • 7 years ago

            “the power envelope is much much greater than a phone and even larger than is acceptable for most tablets. ”

            Where do you come up with that? I see nothing that supports that somehow a: the power envelope is too big for a phone.

            the A15 in a chromebook is 2 A15’s, thus it’s twice the power (components notwithstanding). That’s not “substantial”, it’s half of the power draw of what intel can do and at high load around 3/4 of intel’s power draw with atom.

            go look at anandtech on that.

            • chuckula
            • 7 years ago

            Anandtech: [url<]http://www.anandtech.com/show/6422/samsung-chromebook-xe303-review-testing-arms-cortex-a15/7[/url<] And I quote: "Once again however we see a much larger increase in power consumption once the Cortex A15 is under heavy load. Active power consumption more than doubles on the new Chromebook, while we see around a 30% increase on the Atom based system. I do wonder what will have to be done to get the Exynos 5 Dual into a smartphone as an increase of ~4W under load just won't cut it in a phone. The Atom platform shows a 2.6W increase in power under load, which sounds about right for a high clocked 45nm part." If you read my original post, I never said that the power consumption of a non-SoC 45nm Atom is better than the A-15. I *did* say that the A-15 in the chromebook has an active power that is way too high for smartphones and likely too high for tablets as well. Consider the 4 watt power delta: The entire TDP of the dual-core Clovertrail SoC used in tablets is 1.7 Watts, which is less than half the observed power delta between idle and active for the A-15. Most measurements of the iPhone 5 indicate that the A6 SoC is drawing about 1 watt at heavy load. A 4 watt increase under load does not cut it for these applications. Edit: LMAO, the more facts I post the more fanboys are afraid. Guess what: At CPU even a single-core Medfield phone is able beat an Exynos 5 tablet at several CPU benchmarks, oh and the Medfield phone is running an outdated version of Android and could likely do even better with software updates: [url<]http://www.anandtech.com/show/6425/google-nexus-4-and-nexus-10-review[/url<] That review means one thing: Intel doesn't have an x86 problem, it has a PowerVR problem and once it licenses faster GPU designs the "miracle" A-15 won't look so miraculous anymore.

      • Chrispy_
      • 7 years ago

      Remember, the real world runs on businesses who;[list<][*<]took an extra decade to upgrade from XP[/*<][*<]run x86 software that in many cases isn't even available as 64-bit yet[/*<][*<]require x86 features that may not exist for ARM for at least five years[/*<][/list<] Tablets are cool, but they're just like smartphones in that they will supplement (but not replace) the PC as we know it today.

        • oldog
        • 7 years ago

        The consumer market drives the business market today as opposed to the past. I’ve heard this from developers and the moves by MS appear to bear this out.

Pin It on Pinterest

Share This