The needed prerequisite for staying is arriving. Unlike M1, I have yet to see all these new Qualcomm ARM AI buzzword wunderwaffen. We're at a point where a 8yo computer works exceedingly well and a real breaktrough is needed to signify any real generational shift. Not to mention that I want my laptop to run things other than Windows. 20 Hours runtime is easy when you're incapable of running real programs.
Qualcomm already upstreamed a lot into the 6.8/6.9 kernel, more including GPU have been submitted for 6.10/6.11 and existing Microsoft branded devices like Surface ARM run Linux. The programs run in Windows on ARM are no longer restricted/special and can be x86 32/64 bit as well.
It's really not a bad direction, the question is more is it the most interesting direction for most consumers when you include competition like Chromebooks on the low end and MacBooks on the high end.
In the future two companies will be heavily studied in business school for how badly they blew their technical lead and proficiency to the point that it may be the daeth of them. Those companies are Boeing and Intel.
I would posit in both cases the problem is extreme financialization. Both companies had such a large market share (and even a monopoly in certain segments) that the much-quoted but apropos Steve Jobs quote [1] applies. All these modern CEOs seem capable of doing is stock buybacks, cutting costs and raising prices where they can.
In the 2000s, Intel faced a serious challenge from AMD with the Athlon/Opteron architecture. Intel's failures were a continuation of their Pentium 4 architecture and the company was only saved by the Pentium 3, which became the Pentium-M and the Centrino platform then the Core Duo and the modern CPUs we have now. They're all Pentium 3 descendants.
What we've seen over the last decade is most people have enough CPU power so the move has been towards smaller, cheaper and more energy-efficient CPUs. We've seen this in phones and then tablets and now CPUs. Apple's move to ARM was really the nail in the coffin for Intel as it culminated a decade of not being able to respond to the market.
Intel is seemingly unable to respond to this trend. All Intel CPUs still have super-high TDPs and haven't adopted any silicon to boost AI performance, which is increasingly going to matter. I fully expect this to be further integrated into the M4 and later CPUs.
I, too, expect Windows on ARM to be here to stay because there are simpnly too many advantages to the platform and it's going to eat away at the consumer market in a big way because ARM laptops and tablets will give people what they want: smaller, cheaper and more energy-efficient computers.
Meanwhile Intel just goes from disaster to disaster. The move to 10nm became a running joke almost to the level of Duke Nukem Forever. Intel's once-legendary fab ability is seemingly being eclipsed by TSMC and ARM chips seem to be the primary beneficiary of that.
I rushed out and bought one of the new Snapdragon laptops – the Lenovo Yoga Slim with 32GB RAM. I figured that since Microsoft marketed this as an "AI" laptop, it would be a perfect platform for ML development. However, my experience has been the complete opposite. Python for Windows ARM64 is a complete mess. You have to build almost every module from scratch, and it’s a nightmare of broken and buggy dependencies.
After 10+ years of Windows ARM, I find these basic toolchain gaps inexcusable.
For what it’s worth, WSL on Windows ARM64 is pretty much great.
Windows ARM is 10+ years old. Performant Windows ARM is not. There hasn't been an incentive to make all that stuff work until now because there wasn't much to do with it. I expect that will change soon.
WSL is alright. The real MVPs are the linux distros which have been building and shipping everything for arm64 for ages, and which happen to run on WSL.
I dunno. Microsoft is pretty great at backpedaling. Even when they don't abandon a product entirely[0], they'll fold like a cheap suit and start making compromises at the first whiff of customer pushback.
> Snapdragon chips really shine in multicore benchmarks, overtaking all the other CPUs aside from Apple’s M2 Max and M3 Max.. Snapdragon laptops are generally cheaper than their Intel or AMD counterparts.. Qualcomm Dell XPS 13 is $200 cheaper and the Samsung Galaxy Book4 Edge is $150 cheaper.. [upcoming Intel] Lunar Lake will be closer in design to Apple Silicon, with RAM incorporated into the chip itself, and Intel claims that getting rid of hyperthreading will decrease power consumption, so laptops with these chips will supposedly get better battery life
> on-device AI features will eventually receive widespread adoption, but it will take several years for that to happen. However, the simple fact that we have another CPU option over Intel and AMD is already a big win among consumers.. the improved battery life with comparable performance that Snapdragon laptops bring will force x86 chip manufacturers to innovate on power efficiency or be left in the dust.
It’s amazing to me how little folks dig into the actual data of the test results used to make these “headline” claims. As is always the case, there’s some serious cherry picking of stats and a ton of apples to oranges comparisons.
> Snapdragon-powered laptops have still precipitated some good sales numbers, with Copilot+ AI PCs raking in 20% of global PC sales during launch week. However, industry analyst Avi Greengart said that most users purchased these AI laptops for their better battery life, rather than their AI capabilities.
I was sceptical at first as I very much remember Windows 8 RT, but it seems pretty good. Now I'm curious if ARM can beat x86 in the high-performance desktop market as well. More efficiency is great for less power usage in laptops, but can it also mean more scalable performance in desktops?
It's unfortunate that there is so much proprietary software in the wild that is forever stuck targeting x86. If we move past x86 on the desktop in our lifetimes, I bet a large percentage of people using desktops will continue to carry around x86 emulators for the rest of our lives.
The amount of assembly code is getting smaller in absolute and relative terms everyday. We've been moving code up the stack where it is mostly platform neutral for decades. Each transition makes the next one easier.
Take the Mac, for instance.
- 68k -> PPC: terribly painful, lots of ASM, emulation.
So far the only ARM vendor that bothered with hardware accelerated x86 emulation is... Apple. Microsoft somehow still can't figure out how to do the same, and Google for the most part can just ignore it, since neither ChromeOS nor Android are fighting in the same market.
The crucial difference being that Apples developer ecosystem is accustomed to jumping through whatever hoops Apple demands when they deprecate something, while Microsoft has built their brand on keeping Windows software working forever if at all possible. There's ups and downs to both approaches, but for an ISA transition the Apple way of dictating that everyone will switch or get left behind is obviously a huge advantage.
Apple is so aggressive about deprecation that it wouldn't be surprising if they remove Rosetta 2 within in a few years and kill any software which didn't get updated to ARM, just like they killed 32bit Intel software, and PPC software before that. Microsoft is never going to be able to do that.
Apple is a poor comparison since they switched when Intel's offerings had been uncompetitive compared to AMD's for several years. The last gen of x86 Macs was also in many ways the worst.
The amount of shilling for apple that happens online is so massive that I'd want to try a machine in person before deciding if to buy it or not, because I can't trust anything that I read online.
There are people whose opinions I trust who like them, so I don't doubt they work well. The big complaints are the obscene prices for memory and storage. People buy what they can afford, and it's a fraction of the memory and space on an equivalent Windows system using the same chips from the same factories. And often what they can afford in Apple's world isn't enough.
I quite despise Apple but... you can. I think every Apple store that I have seen or been in has display models that you can just walk up to and start doing stuff on.
Even non-Apple stores that sell Apple devices like BestBuy or Walmart tend to have display devices you can try out.
Much of the stuff which Mac users brag about, like long battery life or efficiency under heavy workloads, is stuff that you can't easily test with a display model. You need to make it your daily driver for a week to get the whole picture.
Right. I just pictured myself walking into an Apple store, unplugging the machine so I can see the battery effects, setting up my dev environment, and building the linux kernel or something on it to see how it performs. It's legitimately what I would want to do to test it, but I'm willing to bet the employees don't let me get that far before they stop me...
if you need to compile linux kernel on the go often then i would advise you to have an always on workstation\server that you use to do the heavy lifting that you then just access remotely form the macbook..
doing heavy compiling work on any device will eat away your battery faster then when you do general use..
but even then, with that workflow while macbook might not give you the many hours of battery that apple brag about, it will still give you more time on battery then then most devices..
you don't necessary need to be logged to a workstation to do all your work..
You could develop de code locally in your macbook in a nice interface, when you are ready to compile then you push the code to a remote git repo in your workstation that also would trigger the compile job, tests, etc..
Do the CPU light work locally and the CPU intense jobs remotely in a system that can have more powerful CPU proper to the task while also freeing your laptop from the burden this CPU intensive job that will make it hot and burn trough the battery..
but yeah, this workflow could work in whatever laptop\os you want, from a chromebook to a macbook..
Apple offer a return period. At a certain point I have to wonder exactly what you're expecting Apple to offer... free extended rentals of Apple hardware so that you can perform an exhaustive analysis of their offering? What manufacturer offers that?
Sometimes things are just objectively good. I have many, many complaints about Apple but their hardware, laptops in particular, are first in class.
edit: c'mon, don't complain about downvotes. You're getting downvoted because you're not offering a useful contribution, not because everyone in the world is an Apple fanboy.
You seem very determined for MacBooks to be bad. You're allowed to hold that opinion. Different people have different requirements from their hardware and I can believe that a MacBook isn't a good fit for everyone. But IMO when assessed as a general purpose machine the MacBook is objectively the best device available.
(you can certainly debate value for money, though)
Macbook is objectively one of the best general purpose devices out there.. and there is a lot of data and reviews out there to confirm..
But if you have specific needs, and that seems to be your case, then you need to test if your workflow will work fine on MacOS\Macbook, but then that will also be true for any OS or device..
Mac can be great for many specific use cases, video editing, photo editing are some examples..
For development it can go both ways, it could be great or terrible depending on many factors..
Well if ARM laptops more power efficient, cheaper, faster which makes consumer choose it (either by going to iMac or installing linux) Microsoft might lose market share. It's also a great excuse to get new eye balls on legacy code and fix all king of issues
On another note: Windows have (by default) tons of legacy to support a lot of different hardware
Am I wrong to assume that that most drivers on ARM cpu running through virtualization? Does Microsoft actually build them or get them from partners to install on the system
How come Windows works on ARM but there's no way to run it bare metal on Apple Silicon? Asahi works so Microsoft should have the resources to port it shouldn't they?
> Asahi works so Microsoft should have the resources to port it shouldn't they?
Asahi works because open source devs working for free made it work by reverse engineering stuff.
But why should Microsoft spend the effort to make Windows work on Apple Silicon, when by far the biggest benficiary of that would be Apple who would sell more hardware? Apple should be the one to invest resources to do it. Not to mention that if Microsoft did it, Apple could throw a wrench in it at any time, leaving Microsoft with users that expect support that it can't support because Apple won't let it (or makes it more difficult than it needs to be).
They already have plenty of hardware makers building ARM laptops running Windows or planning to. Windows is still the dominant operating system outside mobile, and mobile hasn't yet killed laptops and desktops despite perennial announcements of their demise.
They have the resources to do all kinds of things they don't do for the same reason: why would they? What would they get out of it? Apple is nothing next to the smallest Windows OEM.
Probably either an unwillingness from Apple to share HW documentation with MS or MS viewing Apple as a direct competitor now that they have their own machines so they prefer not to undermine their own business.
The greatest thing about Intel/AMD chips is they just work always.
I’ve got zero tolerance for crashes and weird behavior and edge cases and whenever I tried ARM servers that’s exactly what I got.
Computing is unreliable enough without throwing arm into the windows world… it’ll be painful for sure. I can’t see any benefit in arm so compelling that I’d want it over the standard cpu architecture.
Compatibility is everything. It was different for Apple because they control the entire ecosystem and frankly their software engineering on cross compatibility was startlingly good. Microsoft doesn’t have that level of control plus it’s got windows cruft plus drivers and third party hardware and software and it’s hard to imagine a seamless experience. Why would I buy that…. what’s the benefit?
The fact is that hardware, operating systems, CPU architectures - all these things are boring. We use computers as tools for doing the interesting tasks.
Users only start to care about these things when they start to get in their way. If a new ARM PC fails to run some legacy app that's critical to the workflow, then they'll discover that it's because their shiny new laptop has some weird new technology inside it. And then they'll hate it because the computer is no longer fit for purpose. On the other hand, if everything works, then they'll carry on with their lives as if nothing happened.
That's not my experience, we've moved many cloud workloads over to ARM and haven't suffered stability issues.
Maybe you mean the Windows implementation on ARM, if you're using that, but that's not the chip's fault.
Of course, Apple's gone through major chip transitions a number of times and there were issues. Those issues get fixed (on the whole) over time, so just don't be an early adopter if you don't want that pain. I think at this point Windows has already been running on ARM long enough to justify the title of this post.
The biggest advantage is that Microsoft somehow was able to bring a proper regular Windows to ARM devices. Now I don't if it still has the same level of BC support as x86 Windows, but still it is impressive.
Windows on ARM was already very decent on virtualized Apple Silicon, running x86 binaries without much fuzz, they sure have been working on it in the background ever since Windows RT.
Interesting, I was seriously wondering who would buy Windows on Arm on purpose. But in a way, you answered my question.
At work, there are some SAP development functions that required Windows, so people with MACs would beg, borrow and steal to get Windows on their MAC inside a VM. Apple's 'M*' chips ended that.
That was R/3, I do not know about the new version of SAP. I would think they would fully support Apple with that. But you never know.
I use this daily, but it does seem rather absurd that the only way to run Windows ARM on a high-end laptop is... to buy an Apple product, boot an Apple OS, fire up a third-party VM, and then run Windows inside that. (You can't even get a Surface Pro X with more than 16GB of memory.)
Meanwhile Microsoft ships apps that don't work on ARM, like Flight Simulator.
Flight Simulator is a Microsoft app. It doesn't work on Windows Arm. If Microsoft is serious about Arm as a platform, the least they can do is make sure one of their most iconic game titles works on Windows Arm.
Getting sued doesn't stop you from doing things, injunctive relief or losing the case does. At this point neither has happened and it'll probably be a while before it does (it's been 2 years already at this point).
> Qualcomm has most of the critical functions working inside.. Linux 6.9.. critical functions include UEFI-based boot support along with all the standard bootloaders like Grub and systemd.. Functions that are being worked on for future iterations of Linux include additions to battery support, on-board display connectivity, external DisplayPort connectivity, sleep and wake-up functions on the GPU, camera support, video support, better CPU frequency support, and speaker/mic/headset enablement. These functions are expected to arrive in Linux 6.10 and 6.11.
How? The Raspberry Pi and company run on ARM just fine? Assuming all the hardware components are the same or similar, I'm not seeing how it could be drastically worse. I guess missing drivers?
A lot of what makes x86 machines so interchangeable is ACPI, which automates things like device enumeration - having drivers is one thing, not knowing how to even find what hardware is installed is a much bigger hurdle.
It's why ARM support is such a vendor-sensitive roller coaster. MacOS on Apple is easy, since they know their hardware, the RasPi foundation puts serious effort into making Linux etc. work on RasPis, but just look at how much Linux struggles on Macbooks, or how hard it is to run a current kernel on a 3 year old mystery Raspi clone that had no documentation and shipped with a custom Linux kernel binary.
The situation is slooowly improving, but we're far from there yet.
situation will only improve when arm is made acessible to the masses..
i think this will go a bit like secure boot.. it existed for ages, it only became a real available in every computer when MS got behind it and made it happen..
now you can install even most linux distros without a problem and it just work..
arm will be a pain in the beginning but having microsoft backing it might the best way to make it happen..
i am curious to see how microsoft, that historically aim to keep backward compatibility as much as possible, will handle such a braking change like this move to arm..
but one could argue that this breaking change might be what they need to just get rid of some legacy stuff..
> The Raspberry Pi and company run on ARM just fine?
Not in my experience. Maybe things are better now, but as of a year or two ago you still couldn't use a generic kernel on there. You had to have special patches. There's also a fair amount of packages that won't build on ARM. Raspbian is a pretty good experience and it shows they have put serious effort into it, but venture away from that and you're rolling some dice.
