Become a fan of Slashdot on Facebook

 


Forgot your password?
Close
typodupeerror
×
AI Microsoft Operating Systems Windows Hardware

Microsoft To Take On Apple Silicon With Custom ARM Chips 50

Posted by BeauHD from the what-to-expect dept.
According to Windows Latest, Microsoft is working on new ARM chips to compete against Apple Silicon. "I have also spotted some job listings that suggest the company is building its own Silicon-based ARM chips for client devices" writes Mayank Parmar. "Additionally, I understand that Microsoft is optimizing Windows 12 for Silicon-ARM architecture." From the report: These developments coincide with the upcoming launch of Windows 12, which has a special version optimized for silicon and designed to leverage AI capabilities. The job listings (most of them have now been taken down) describe positions related to custom silicon accelerators, System on Chips (SoCs), and high-performance, high-bandwidth designs. This suggests that Microsoft is building its own ARM-based chips, aiming to compete with Apple's M chips lineup in terms of performance and efficiency.
This discussion has been archived. No new comments can be posted.

Microsoft To Take On Apple Silicon With Custom ARM Chips More

Microsoft To Take On Apple Silicon With Custom ARM Chips

Comments Filter:

  • Microsoft extends its ARM (Score:5, Funny)

    by awwshit ( 6214476 ) on Monday May 01, 2023 @05:45PM (#63489986)

    Gives Intel the finger.

    • Waiting with bated breath for Intel Horse Creek and newer to finally get out of delay hell, so Microsoft can be given the finger it deserves.

      Sifive and Starfive boxen are slow RPi-class dev boards, Horse Creek is a fast dev one, let's see how soon we can get machines for regular desktop users.

      There's no reason whatsover to pick Microsoft for anything ARM or RISC-V, so it's a fat opportunity to get rid of the biggest obstacle computing ever had.

      • Wait for it, wait for it...

        ARM will begin working on RISC-V. At the least they will issue good performing examples and let the market choose or not.

        More likely they will catch up quickly and burn the competition. Again.

        Intel closed their RISC-V team, big mistake. Microsoft is not yet admitting to even evaluating RISC-V. Watch that space.

    • Re: (Score:2)

      by Pieroxy ( 222434 )

      They are not giving Intel anything. Until Microsoft releases something like Rosetta, to allow x86 apps to run on those ARM machines, this will go nowhere. Who will buy a computer with zero apps available ? It turns out we already have the answer to that question: it's not the first time a Windows machine will run on ARM. There were tablets back in the days. Nobody bought any.

      Rosetta is the reason Apple's transitions from PPC to Intel, and Intel to ARM went so smoothly. Even when they migrated from 68k to PP

      • Users can run x86 and x64 apps without any issues on Windows 11 ARM. Meanwhile, developers can incrementally port using Arm64EC which allows for mixed x64/arm64 binaries within the same application. Once all x64 code successfully recompiles, they can opt to release a fully native arm64 app. It is actually a lot more flexible than Rosetta. Shame that Windows itself is still⦠well⦠Windows.

      • They've had 64-bit emulation [windows.com] for over a year. They supported 32-bit apps from launch, back in 2017.

        They also released a new ABI called ARM64EC [windows.com] to aid developers in transitioning x86 applications to native ARM versions.

  • What were the previous versions optimized for... wood?

  • Console Computer? (Score:3)

    by zenlessyank ( 748553 ) on Monday May 01, 2023 @05:51PM (#63489996)

    No thanks. I guess that proprietary bug has bitten them again.

    • Re: (Score:2)

      by AmiMoJo ( 196126 )

      That seems to be an issue with the ARM architecture in general. If you look at all the high end SoCs like Apple M1 and Qualcomm's offerings, they require very tight integration to get a fraction of the performance of x86 discrete CPUs.

      It's due to the ARM instruction set being relatively low density compared to x86, or rather AMD64 as it mostly is now. Modern AMD64 CPUs do a massive amount of on-the-fly optimization, including out-of-order execution, speculative execution, and register renaming. To do that s

      • A modern ARM CPU has between 1000 and 2000 instructions.
        A modern AMD64 CPU has between 1000 and 2000 instructions.

        Where were you going with this?

        • No idea about x86 or AMD64, but an ARM has only roughly 40 instructions.
          Everything else is encoded n modifier bits to those instructions. E.g. additional shifts, or conditional execution.

          Perhaps you refer to extensions like THUMB, or JBC?

          See: https://iitd-plos.github.io/co... [github.io]

          • I'm referring to ARMv8, the 64bit version of ARM.

            • Yes. So?

              Hardly more than 40 instructions.

              And there is hardly any other processor anyway that has much more.

              We have:
              add, sub, mult, div, mod
              double that for floating point
              load, store
              compare
              branch - on about 16 bits, so 16 instructions
              pop/push stack
              and, or, xor, not
              jump / branch subroutine
              switch privilege - aka go into hyper/super modus
              return from hyper / super modus
              return from subroutine
              jump - unconditional
              shift left and right
              rotate left and right

              Did I miss anything important? I'm pretty sure ARM is about 40,

  • Why assume it's consumer? (Score:4, Interesting)

    by Pinky's Brain ( 1158667 ) on Monday May 01, 2023 @05:53PM (#63489998)

    Seems equally likely it is meant for Azure.

    For the consumer market, processors are hardly the most pressing need for competitiveness. They need mobile phones, finance, home automation etc etc. Ecosystem or perish.

    • because that doesn't make for a high volume clickbait article.

    • Re: (Score:2)

      by Xenx ( 2211586 )
      I think they assume it's consumer because that is Microsoft's current plan. They've released multiple ARM based consumer PCs, and 3rd party SOCs cannot stand up to Apple. They very much are trying to compete. That doesn't, however, mean they don't recognize it's use in other markets down the line.

      • Apple uses ridiculously large amounts of silicon, while buying out the most advanced nodes from TSMC for 2 years ... it's not really an ISA issue, it's a profit margin issue. If you can't sell high margin products at Apple volumes, you just can't afford to design silicon the way they do.

        They should just pay AMD for a version of Phoenix with a wider memory bus and LPDDR on interposer, that would do more to close the gap than ARM.

        • Apple have run away with the performance envelope in certain types of processing, especially with the AI and Metal extension on their silicon, and with the added benefit of the secure element. It isn't perfect, of course, but sure makes x86 procs look shitty by comparison. Apple also has the advantage of making profit on all the other components for 100% of their install base. MS does not. Apple's model would be attractive for Microsoft to ape, if not for the profits, for the competitiveness. MS will not be

          • Re: (Score:2)

            by gtall ( 79522 )

            If MS is attempting to move everyone to the cloud, does it matter that they are designing their own chips?

          • A laptop with a say a RTX3060 has 100+ TFlops from the Tensor cores ... Apple doesn't really compare.

    • Re: (Score:2)

      by Trongy ( 64652 )

      Microsoft already has an arm64 server offering in Azure [microsoft.com] using Ampere Altra CPUs. I don't think their market share is enough at this stage to justify developing their own silicon.

      This article is based on job descriptions for System on Chip (SoC) development. That terminology generally referrers to the chips used in portable devices.

  • Took them long enough (Score:3)

    by rnmartinez ( 968929 ) on Monday May 01, 2023 @05:53PM (#63490002)
    I thought they had a deal with Intel that they wouldnâ(TM)t do this sort of thing. Maybe it expired?

    • Intel is undertaking the difficult journey of turning itself into a fab. Just as Apple doesn't manufacture their own silicon, neither does or will Microsoft. Intel should see this as an opportunity.

      • well, yes true, neither MS or Apple will own the fab, but they will have a direct hand in the design which I thought was a no-no for MS (can't remember if it was due to a deal with Intel or anti trust?)
        • Probably contracts with OEMs. But designing and rolling out silicon is a BIG undertaking, which will take years of development on the silicon side, and years of OS revamping. Apple are the experts in this area and it takes them years to do transitions like this (POWER >> Intel >> Apple Silicon). Each >> was a few years, with transition assistance for developers and emulation layers in between. Microsoft is just planning for the day those contracts sunset.
  • The custom silicon race has gone far beyond any possible profit margin scale at this point.

    Back when Intel got lazy over having a monopoly the field was wide open. Arm came in with mobile stuff, AMD took the opportunity for server chips. Etc.

    But now we have Apple, Qualcomm, ARM, AMD, Intel, Tenstorrent, Microsoft, some other ARM ISA server producer I can't even remember the name of, and ought else. This amount of competition means margins will be squeezed down until producing your own custom chip is f

    • > The custom silicon race has gone far beyond any possible profit margin scale at this point.

      Well, that depends.

      For companies making commodity components, or trying to sell devices made from commodity components, then yes, trying to make a profit from them is very hard. And in general, investing in R+D to make incremental improvements is not likely to see a good return on that money.

      But if you are a vertically integrated company, and can design upcoming hardware around new hardware, and can sell those de

      • Re: (Score:2)

        by Xenx ( 2211586 )

        Yes, there's lots of very cheap alternative devices out there, but there's a distinct lack of "almost as good as apple" products.

        Pretty much what I was going to say, until I got sidetracked with actual work. They want/need something to directly compete with Apple, and nobody can. Qualcomm is well positioned at the top end for mobile devices, but their SOCs for PC are underwhelming.

    • Apple was an investor in ARM when the âoeAâ still stood for Acorn, but then sold their investment.

      The initial ARM chip was so efficient that even the creators were surprised. They were some smart people. A good video on the history:

      https://youtu.be/nIwdhPOVOUk [youtu.be]

  • MS has amazing record of supporting non-x86 hardware. Look at MIPS, PowerPC, Itanium, Dec-alpha and so on.

  • The Microsoft SQ2 (and SQ1) processors in the Surface Pro X were developed in partnership with Qualcomm.

    Is this a revolution or an evolution?

  • The amount of time it takes before your current hardware no longer supports the latest version of Windows halves every 18 months.

  • > Windows 12, which has a special version optimized for silicon

    The normal version isn't optimized? I wouldn't be surprised...

  • twice in the summary/article there's a reference to "Silicon-ARM". Is there a kind of ARM processor that isn't based in silicon? (I know there are other semiconductor technologies out there; none are used for complex processors.)

    Or is "Silicon-ARM", with a capital S, some sort of nonsensical branding, like "Wet-Water".

  • The ARMs race (har har) is going to be which company is going to dominate the chip fab plants production capacity. Apple already soaks up a lot of TSMCs capacity with chips destined for iphones, ipads, and Macs. What's Microsoft going to make that would warrant being able to demand priority in the fabs?

  • I wouldn't touch this. Not in a million years.

    Here's my prediction: Microsoft will create a chip, spend hundreds of millions adapting Windows and Office to that CPU, only to kill it a few years later.

    This is going to be like the many other hardware projects that were born at Microsoft (and later died horrible deaths).

Slashdot Top Deals

Some people manage by the book, even though they don't know who wrote the book or even what book.

Close