1. 20

  2. 5

    I think a lot of these articles are based around the idea that “everybody” wants a certain type of PC, and suggests Windows on ARM is languishing due to the absence of that device.

    But I think a better way to look at it is PC users are hugely varied, and the Intel PC ecosystem consists of a wide variety of devices. The ARM ecosystem consists of a handful. The probability that the best device for a given user is an ARM is therefore very small. Adding one or two devices at any price/performance point is unlikely to change things dramatically.

    The article mentions Powershell being slow to load, although this isn’t just hardware: this is because (for compatibility) .NET ends up running under x64 emulation, which makes managed code look much worse than native code in these comparisons (which probably also explains the strange Firefox comment.)

    I own a 7c Windows laptop. It’s definitely not fast, but it does run Windows 11, has great battery life, got a lot more useful with the recent release of native ARM hosted compilers, and costs $250. For the record, Powershell launches in 2 seconds, which still seems (to me) incredibly painful, although it’s not as bad as the article suggests out of the box.

    1. 4

      The PC marketplace is as you say incredibly varied and broad, but, and you’ll forgive me if I’m falling prey to the very fallacy you’re signposting, I’m still confused as to why the PC industry and Microsoft haven’t at least focused on competing against the Macbook Pro.

      Apple had a really awful run for a few years there when it comes to laptop hardware, but the new M1 models are really quite excellent.

      After overdosing on pain and frustration with my work issued 2017 butterfly keyboard bearing Macbook Pro, and feeling like MacOS X was becoming more and more locked down and painful to develop for, a couple of years back I bought a Thinkpad T15 gen2 which I’m generally very happy with, but the difference in both battery life and compute between it and one of the new M1 MBPs is very very notice-able, and as pleased as I am with my Thinkpad, it falls prey to the usual PC laptop foibles - a 300 nit display of not amazing quality, god awful built in speakers an a battery that drains REALLY quickly if I do significant work like running an IDE on the machine.

      So I’d love it if despite the challenges you cite either the PC industry could come up with a compelling ARM laptop or if the Intel side somehow broke out of its funk and caught up :)

      1. 3

        Apple had a really awful run for a few years there when it comes to laptop hardware,

        Yeah, between ~2016 and ~2020, there was pretty widespread discontent, but I can’t say that I saw any serious attempt to bump off Apple as the “premium” laptop manufacturer. And it’s not as though Apple is invincible: for a long time, Apple was the “education” leader, but then Chromebook undercut them and now they’re essentially shut out of that market.

    2. 3

      It’s all about sales volume. It costs around $1 billion to tape out a 7nm chip. Apple sells 30 million Macs a year, so assuming one new chip a year, that’s $33 per Mac. Microsoft sells about 1 million Surfaces a year, so if they were to do their own chip, it would cost over $900 per unit just for the tape out cost.

      1. 3

        I seem to remember Apple’s computer chips sharing a lot of technology with their phone chips, which would spread out the cost even more.

        1. 3

          Yes, and the newer iPads uses the M1 as well. They’ve done a phenomenal job achieving massive economies of scale, which is hard for any one competitor to beat.

        2. 1

          ISTM that while this is true, it is only partial.

          Yes, Apple’s SoCs are in-house and proprietary.

          But that does not mean that everyone’s have to be. There is no particular reason that I can see why MS has to do the same thing and design its own SoC in-house, but this is an unstated assumption of your argument.

          MS is free to buy Arm SoCs from whoever it wants, and port to them. MS is a rich company, and it also has deep specialisation in OS design. One of the factors limiting portability of Arm OSes is that there’s no standard firmware.

          MS has Arm firmware. It has sold both Arm-based devices running both Windows (the Surface RT) and Android.

          I wrote about the Surface RT nearly a decade ago: https://www.theregister.com/2013/11/14/microsoft_surface_rt_stockpile/

          Now there is also the Surface Duo: https://www.microsoft.com/en-us/surface/devices/surface-duo#overview

          MS has Arm chops. It has or had the skills, the expertise, the design nous. (If that’s gone it’s the company’s own fault.)

          If there are no other Arm SoCs that are truly competitive with Apple’s ones, then that is not MS’s fault. It the the wider Arm industry’s fault.

          But MS can write its own firmware for various Arm SoCs and port its Arm Windows to various hardware from various vendors and cherry-pick the best open-market Arm hardware.

          You imply it must do it in-house. It doesn’t.

          1. 1

            MS is free to buy Arm SoCs from whoever it wants, and port to them. MS is a rich company, and it also has deep specialisation in OS design. One of the factors limiting portability of Arm OSes is that there’s no standard firmware.

            The problem is that no one else is producing M1-competitive cores. The Arm SoCs that you can buy are either:

            • Mobile phone chips from companies like Qualcomm or Samsung. These aim for a very low power draw.
            • Server chips from companies like Ampere. These focus on performance and power efficiency in SoCs on the order of 100W.

            No one is producing laptop-grade chips. A laptop SoC has a lot in common with a mobile phone chip these days (which is why Apple can share costs between their phones, tablets, and laptops) but it’s a separate SoC and so incurs a load of NRE that you need to then amortise over a large number of units. If someone starts making a successful high-end Android tablet, then Windows on Arm laptops could benefit from that, but until then there’s the problem that few people buy Arm Windows laptops because they run phone chips and no one is building laptop chips for them because the market is too small.

            This might change if it becomes possible to install Windows on Apple Arm laptops and people do in large quantities: that would show a company like Qualcomm that there’s a big market for a high-performance Arm Windows laptop.

            1. 1

              Well, yes, I agree entirely.

              That is the problem, and the responsibility, of Arm licensees and vendors, though.

              AFAICT: nobody took Arm seriously for ~30 years now, and considered them as only being important for embedded stuff and battery-powered devices without active cooling. (As a former Acorn Archimedes owner, this pains me.)

              So: nobody has invested in performance-optimised Arm SoCs.

              However, the message to which I replied blamed MS for this and seemed to expect MS to fix it by paying for the development of high-performance Arm SoCs for its Surface devices.

              That isn’t MS’s responsibility, and I think it would be actively harmful to the industry if it did so.

        3. 3

          I’m mostly running macOS, but for some work things I need to use Windows. So I run a virtual Windows, which worked ok on x86 - but now I’m on M1 and things are worse. I mean, there’s not even a native version of Visual Studio for ARM (but apparently they’re in beta now?). Microsoft has been selling Windows-on-ARM for… 10 years or something? And they haven’t even bothered to get VS running on it - you’re supposed to develop on x86 and only run tests on the target platform. It really shows that they haven’t been even remotely serious about ARM.

          1. 2

            I keep seeing claims like “Qualcomm announced they would have a chip to rival Apple’s M1 in 2023… except that Apple’s already moved on to M2, so Qualcomm’s targeting performance metrics already three years out of date!” As far as I can tell, Qualcomm never said M1-class, they said competitive with Apple M-series. The rest looks a lot like a media inference, and if so, bringing it up isn’t a great look.

            Happy to eat crow if I’m wrong.

            1. 5

              I wonder if it’s about money and market share.

              Qualcomm is making money hand over fist selling chips for mobile platforms. I wonder if, as a result of the fragmentation @malxau cites above, they’re feeling like it’s not a good investment to put enough gas in the engineering tank to really compete with the m1.

              1. 1

                You’re right, they said as much last November: https://www.windowscentral.com/qualcomms-nuvia-based-advanced-arm-chip-pc-rival-apple-2023

                So I guess the uncharitable take is they’re aiming where Apple M-series was at that time (M1), which seems unlikely.

                Most charitable take, they’ve estimated where Apple will be by 2023 (with aid of their Nuvia acquisition and probably other intelligence gathering), and are aiming for there.

                I think most likely is they’re gonna spend whatever their budget is, do their best, and see where they land. The marketing department will surely find at least one metric where they compete with Apple.