Hacker News Reader

Comment by imiric | Hacker News Reader

Godspeed to the Asahi team, but as much as I envy the performance and efficiency of Apple silicon, I could never depend on a small group of hackers to reverse engineer every part of a closed system and to maintain it in perpetuity so that I can run free software on it. As brilliant as this team is, and as much progress as they've made, fighting against a trillion-dollar corporation that can decide to shut it down at any moment is a sisyphean endeavor. Spending thousands of dollars on that bet is a hard sell, even for tech nerds.

Not to mention that you'd be supporting a corporation that has this hostile stance towards their customers to begin with.

Meanwhile, other x86 and ARM manufacturers are making substantial improvements that are shortening Apple's lead. You're not losing much by buying a new CPU from them in 2024 or 2025, but you gain much more in return. Most importantly, the freedom to run any software you choose.

View on HN

Aren't tons of Linux drivers for x86 laptops based entirely on reverse engineering? Maybe even most of them? I haven't used Linux seriously in almost two decades, but that's my memory.

Most of the x86 platform (ACPI) is well defined and openly accessible (not free but open).

There’s still some offenders (Surface, HP, Broadcom) that introduce quirks that break sleep and some HID accessories but most of it works out of the box.

ARM has been the Wild West for a while but they’re going in the right direction with device trees et al. Apple however doesn’t have to care about the “wider” ecosystem since they left x86 for their own silicon and tighter integration from bottom up allows for some really nice end user experience.

I still think it’s much better to use the VM infrastructure and just run Linux as a guest. Mac as a platform is very end user friendly as-is unlike Windows.

Device Trees are becoming an old thing now. With ARM SystemReady, most devices need to have UEFI, SMBIOS and ACPI. Only the SystemReady IR (IoT) variant is defined as using Device Trees.

https://github.com/ArmDeveloperEcosystem/systemready-guides?...

Microsoft is the only one pushing for ACPI on ARM, but as you said they have hefty weight in that area. I don't think it is right for the platform, but if it works who am I to say.

You've been out of the game for too long; almost every major hardware vendor has at least one or two people that are regularly submitting patches to the Linux kernel. My ThinkPad work computer running Linux is a major thing of joy; in many ways it performs more reliably on Linux than it does on Windows.

Even though, there are no lack of issues that are simply not cared about. E.g. thermal throttling for ThinkPads is a very annoying problem the best solution to is simply a python script that just periodically overwrites a memory location.

It isn't nearly so dire. The modern solution for this is https://gitlab.freedesktop.org/upower/power-profiles-daemon

it's using dbus how modern can it be? :-)

We are well beyond the days of NDISWrapper, most of the kernel contributions come from hardware manufacturers or integrators

I'm not sure if "tons" is accurate, but some of them are, yes. And most of them are not great IME. Not discrediting the talented programmers who work on them, it's just the nature of supporting hardware in the dark, without any support from the manufacturer. Though this was more common during the early days of Linux, and nowadays many manufacturers offer some kind of support.

OpenChrome far exceeded Unichrome Windows drivers in performance. But things have changed. Modern engineers prefer “official” software. I understand why. Systems are more complex now.

While I don't feel I have enough information to comment about the likelihood that Apple would try to stop the Asahi project, those who are knowledgable are of the opinion that they would not.

However, as a Mac Studio M1 owner that has used Asahi as a daily driver for software development since the first release (originally Arch, later Fedora), I can confidently say that I could care less. By running the software I want to run far faster than macOS could on the same hardware, Asahi has saved me countless hours and made me far more productive. And I'm incredibly grateful for this tangible benefit, regardless of what happens in the future.

You claim that your apps are faster on Linux on an M1 than macOS on the M1; can you add more detail? Which apps, and did you run benchmarks? I find it hard to believe that Apple hasn't optimized apps to be faster on their own OS and hardware.

In short, all apps run faster - I have some more detail in this blog post (which also includes a link to my original blog post with additional performance details): https://jasoneckert.github.io/myblog/fedora-asahi-remix/

I suspect this is primarily due to Linux being a more performance-optimized OS compared to macOS, which seems to have introduced a great deal of bloat over the years.

They won't stop Asahi with a frontal assault, they'll stop it by churning out new chips every year until the work to support them all is unsustainable.

Not sure why you are being down-voted. We're already seeing this with the team saying they won't work on M3 yet because they aren't even close to done with M2...

actually they're saying they're close to getting M3 support and the big thing that has prevented work on it as the lack of a M3 Mac Minis.

How is that apple's fault, nor any form of "deliberate attack"? Like, come on, neither of the parties are malicious, especially not for the sake of it.

I didn’t say it was anyone’s fault. It’s the reality of a closed system.

Why don't new Intel or AMD chips require painstaking reverse engineering with each release?

Historically they do, and currently they don't because their biggest market is server machines where Linux is the de facto choice. It's a life or death question for them to properly support the Linux kernel.

In one search I'm able to pull up Pentium documentation from 1995 as well as recent documentation from 2023. Show me the equivalent Apple Silicon docs that a compiler engineer could use.

> As brilliant as this team is, and as much progress as they've made, fighting against a trillion-dollar corporation that can decide to shut it down at any moment is a sisyphean endeavor

Apple historically cares very little about Linux on Mac whereas it seems like you’re talking about the non-Mac product lines. Indeed, they go out of their way, if I recall correctly, to make it possible and easier in the first place.

I wouldn't describe leaving the bootloader unlocked as "going out of their way" to make all of this possible. Clearly, if just booting another kernel would be sufficient, running Linux on their machines should be easy. Yet none of this is. "Going out of their way" would at the very least be providing documentation and support so that reverse engineering their hardware wouldn't be necessary.

Also, what's not to say that they will decide to lock the bootloader just as they do on all their other devices? What does Apple practically gain by leaving it unlocked? If they're doing it as a gesture of good faith to their users, they're doing an awful job at it. Doing it so they can sell a negligible amount of machines to a niche group of hackers also doesn't make business sense.

Depending on the good will of a corporation that historically hasn't shown much of it to the hacker community is not a great idea.

The changes made were not as simple as not-setting a lock fuse bit. Making the bootloader unlockable in a way that didn't compromise their existing security model did require going out of their way. The status-quo for previous "apple silicon" bootchains (iphone, ipad, etc.) was not built this way.

Even T2 macs had no way to boot custom firmware on the T2 chip, without exploits.

Sure, they could've done way more, but evidently they'd rather not lock down the platform completely.

Apple does new hardware bring up using the Linux kernel.

If they don't release that code to the public, what good does it do? (Also, if they are only doing a temporary in-house version for initial hardware work, they can do all kinds of ugly hacks that wouldn't really be good for upstream use any anyway.)

I am genuinely looking forward to a Dell XPS 13 or Lenovo X1 Carbon which is fanless and have the battery duration and performance of the Apple Macbook Air.

What is the difference in battery life between linux on and macos on the Apple M1?

That is, I would be surprised if linux on the M1 had close to macos levels of battery life. My theory being the better battery life on the M1 is more due to the tight integration between the OS and the hardware power profiles than the power profiles themself.

I’m sure Apple has some unique tricks when it comes to energy efficiency, but I haven’t seen the same level of optimization in other operating systems. Apple’s energy management is just another competitive advantage, offering a level of sophistication that sets it apart technically and strategically. Just add the Mx chips to the equation.

Honestly, I don't think this is true. I had identical efficiency on my Intel MacBook Air running Linux, compared to OS X. Both ran out of battery +/- 10 minutes of each other on the same hardware.

The only distinct advantage is Safari, which is heavily optimized for efficiency. But a lightweight Linux desktop, with fewer active services, can compensate for that.

And, do you primarily use Google Chrome? You lost me a bit with "...but a lightweight Linux desktop" as it seems to counter your point. When discussing battery efficiency, I think it’s key to consider real-world tools like VSCode, which many of us use regularly. Chrome, especially, can drastically alter power usage across different operating systems.

I use Firefox on Linux, but used Safari on OS X. My point was that this gave some efficiency advantage to my MacBook Air running OS X, compared to the same machine running Linux (and Firefox). Safari is a little marvel of energy efficiency. This was countered, to some extent, by a simple barebones X11 setup on the Linux side. Otherwise, my setup was identical on both sides.

Linux tends to be regarded as energy inefficient because it ships with defaults that prioritize performance on desktops, workstations and servers. Some simple tweaks with udev rules can make a big difference. For example, adjusting the energy policy of your SSD can reduce consumption by 0.5-7 W. Wake-on-lan is also something that is typically enabled and drains a lot of battery when the machine is suspended.

I'd be surprised if MacOS could match the efficiency of Linux. MacOS relies on a hybrid kernel architecture that emulates a variety of different APIs that aren't used or integrated fully. The simple act of running software on MacOS is deliberately somewhat inefficient, which is a perfectly fine tradeoff for a desktop OS that's not intended for server or edge applications.

The fundamental hardware of Apple Silicon is very efficient but I don't think MacOS is inherently optimized any better than the others. My experience installing Linux on Intel and PowerPC Macbooks tended to increase their battery quite noticeably.

Well, if in 2021 you took your MacBook Air M1 (8GB) out on a Friday, downloaded movies, watched them, browsed the internet, did some casual development, and came back late Sunday without needing to charge it, I’d be impressed.

Is the Linux support not there yet? How close is it?

It's the hardware that's the problem, not Linux support. Simply, the hardware manufacturers don't make fanless, thin, light, performant, power efficient laptops.

What about fanless chromebooks?

Yup that's a good start! It proves that a company other than apple can do something fanless. Probably they're plastic-y, but they are thin, light, and fanless. Power efficiency and performance are likely not good, but, at least google doesn't deliberately obfuscate their hardware like Apple does. Instead, they just let everything that's not ChromeOS fester, since they're trying to make money. But anyone who wants to start a business selling Alpine on Chromebooks can ;)

Don't let worries about future hardware get in the way of using gear that works in the present. If future Macs aren't supported for some reason, that doesn't break your current hardware, and you can buy different hardware next time.

There are people running Linux on abandoned hardware from companies that went out of business, and that's okay.

No, because from your perspective you are supporting the wrong company. Competition is good for consumers, especially if you're a minority group.

I don't think boycotting Apple because they don't support open source enough is likely to have much effect?

Maybe there are other companies you'd prefer to support, but it's still only going to work if lots of other people buy stuff from them, too.

It's not boycotting, just voting with your wallet ...

I don't see any difference. This "vote" is a rounding error unless lots of people do it, which is not going to happen without organization. The effect on you matters and the effect on the vendor is ignorable.

> Not to mention that you'd be supporting a corporation that has this hostile stance towards their customers to begin with.

Is the "this" in that sentence your previous paragraph of concern that Apple will purposefully break AsahiLinux?

Exactly this. All that work, for reverse-engineering a handful of laptops made by a company whose only desire is to lock its users into its ecosystem and ergonomics. Even more demotivating is that the M1 and the M2 are already superseded.

Similarly, I completely do not understand the popularity of Apple's laptops in this community. Endemic mindless fanboyism.

Indeed, the Apple Silicon hype stemmed from the terrible Intel gen at the time (unlike AMD), and because Apple monopolized the TSMC N5 node for months.

Most of that lead is gone, x86 is cheaper, open, and as battery friendly now.

From https://news.ycombinator.com/item?id=42016931, it looks like M4 is very much still in the lead.

It's also not just the CPU: the laptops themselves are simply phenomenal. I have never had a x86 laptop that's this much of a genuine pleasure to use (ignoring macOS, which is my least-favourite operating system). Some of the newer Windows laptops are ahead, but they're still worse in at least one metric.

Not to mention Mac laptops are some of the only laptops with non 1080p or non 16:9 aspect ratio.

Microsoft also has forced makers to drop the old S3 sleep modes which set x86 laptops back decades regarding sleep and power management.

That was true 5 years ago. Today, many flagship Ultrabooks have 3k+ screens and 16:10 ratios. I'm typing this from my Lenovo yoga 7 that does.

That's good to hear. Last time I had to get a new PC laptop was probably ~3 years ago so that sounds reasonable.

My AMD Ryzen Framework Laptop 13 running Ubuntu 22.04 is a joy to use. Huge difference between that laptop and the first generation Intel Framework Laptop 13.

I don’t notice the fan running on it much at all.

That's an unconfirmed leak. I only compare chips that are available

> You're not losing much by buying a new CPU from them in 2024 or 2025, but you gain much more in return

That's just false. Performance-per-watt, AMD, Intel, Nvidia and Qualcomm still get smoked by comparable M-series chips.

On desktops the equation is different because CPUs and GPUs can go ham with the wattage.

I’m with you in spirit, but most of the work has already been done. Purchased hardware won’t change. Firmware updates could be held if needed as well. Another team could take a crack at it.

Worse case you restore macos and possibly sell at a medium loss. That said I’m still waiting.

Sadly, this is the exact reason why I hold back trying Asahi and run the chance of liking any of it :-(

Recently saw someone wondering why no one has tried building a laptop with as much quality as an Apple? A special version of Linux to run on such a laptop would offer more long-term commitment and maybe pull in more adoption.

> other x86 and ARM manufacturers are making substantial improvements that are shortening Apple's lead

x86 has fundamental issues that I believe prevent it from ever achieving the MIPS per watt efficiency of anything from ARM. I mean... the newest M4 laptop will have a 24 hour battery life. That exceeds anything remotely possible in the same laptop form factor but with x86 by nearly an order of magnitude.

So now you're talking just ARM. Linux has been compilable on ARM for a while now, so where are the competing ARM laptops that are anywhere close to the power of Apple's version of ARM?

I do get what you're saying though (I'm definitely a Linux fan and have a Linux Framework laptop), but I wish it wasn't an x86 laptop because its battery life is crap (and that is sometimes important).

> so where are the competing ARM laptops that are anywhere close to the power of Apple's version of ARM?

Better question: where are the incentives for them to make it? Apple is pretty much the only company with an outstanding architectural license to design ARM cores, and the best off-the-shelf ARM core designs don't even compete with 5-year-old x86 ones. If you're a company that has Apple-level capital and Apple-tier core design chops, you might as well embrace RISC-V and save yourself the ARM licensing fee. That's what Nvidia does for many of their GPU SOCs.

If SoftBank offered ARM licenses under more attractive terms, there would be genuine competition for good ARM CPUs. Given that Apple has a controlling stake in SoftBank, I wouldn't hold out faith.

> Apple is pretty much the only company with an outstanding architectural license to design ARM cores

Many other companies have done this to great effect in the past, but in recent years it has become more common to just license one of ARM's own stock cores, instead of designing your own from scratch.

This follows a period where companies like Qualcomm and Samsung were still trying to roll their own core designs from scratch, but ending up with designs that were slower and less power efficient than the cheaper stock cores you could license from ARM.

> the newest M4 laptop will have a 24 hour battery life. That exceeds anything remotely possible in the same laptop form factor but with x86 by nearly an order of magnitude.

Honestly, why is this such an appealing feature? Are you often away from an outlet for 20+ hours?

I use 6+ year old laptops that last 4 hours at most on a single charge, even after a battery replacement. Plugging them in every few hours is not a big inconvenience for me. If I'm traveling, I can usually find an outlet anywhere. It's not like I'm working in remote places where this could even be an issue.

Then there's the concern about fan noise and the appeal of completely silent computers. Sure, it's a bit annoying when the fan ramps up, but again, this is not something that makes my computers unbearable to use.

And finally, the performance gap is closing. Qualcomm's, AMD's and Intel's latest chips might not be "M4 killers" (or M3, for that matter), but they're certainly competitive. This will only keep improving in 2025.

It's not that these are must-haves: it's that it removes any such anxiety about these to begin with. I can take the laptop with me wherever I'm going, and not have to worry about charging it, or that the heat will make it unbearable to use on my lap, or that the fan will be noticeable. It means I can work at a cafe, in a car, on a bus, on a train, on a flight without power, and not have to worry.

And these things compound, as the other poster mentioned: 24 hours of light use means longer heavy use, which actually does matter to me. I often move around while using the MacBook because it's good to change up my (physical) perspective - and it means I can do that without the charger while hammering every core with rustc.

Once you see that better things are possible, it's very hard to go back to the comparatively terrible performance-per-watt and heat generation of equally powerful x86 laptops.

I'm not saying that these things are not appealing. I would certainly like to enjoy them as well. But there are always compromises.

With Apple, you first have to pay whatever exorbitant price they want to charge you. The Air is relatively affordable, but bump up the specs on a MBP and your eyes will start to water.

That's fine. You get what you pay for, right?

Except now you're stuck in their walled garden and have to pay the same tax for overpriced accessories, software, and anything else they decide to charge you for.

OR you go the Asahi route and live with the stress and uncertainty of depending on a small team of contributors to keep your machine running. I already experience this stress with Linux on supported hardware, and couldn't imagine what it would be like if the hardware had to be reverse engineered for this to work.

That said, I'll probably end up buying a used MacBook Air just to follow the progress of Asahi. If nothing else I'll have a more informed opinion the next time this discussion is brought up.

> Once you see that better things are possible, it’s very hard to go back

Yeah, there’s something a bit freeing about being able to go all day or more without charging. Just not needing to think about power or charging when you’re busy focusing on other things.

I’m glad other manufacturers got a bit of pressure to catch up as well. Now people come to expect laptops to just run for days at a time without charging.

With all due respect I think this is a "640k is enough for everyone" problem, in the sense that you don't realize what something enables because you're simply so used to not having it:

1) Internet cafes that removed outlets to encourage laptop people not to squat :)

2) Airports where you can sit anywhere you want instead of just near an outlet

3) Airplanes when the power doesn't work (has happened more than once in my experience)

4) Cars, trains, subways, buses

5) My boat, sometimes I like to work from it for a change of pace

6) Don't have to hunt for an outlet when my fam is at the grandparents' house

7) I can work on my deck without dragging a power cord out to the table

8) I can go to a pretty overlook and watch the sunset while working

9) Conference rooms, don't have to deal with the hassle

10) Libraries, same thing. I can sit outside or in the stacks (quieter there) instead of only in the reading room (those tables are the only ones with power, in my local library)

11) Power outs and just other situations where you lose power for a time

12) It's extra juice to power your phone off of (or anything else)

I'm certainly forgetting more examples.

Some of those things would be nice, sure, but it comes down to the question of whether they're worth trading for using a computing platform tightly controlled by single corporation.

macOS is a deal-breaker for me, whether I use it in a cubicle or on a boat.

So the only alternative is Asahi Linux, which has its own set of drawbacks.

Clearly we must have different priorities, which is fine. Just don't think that yours are somehow superior to mine.

I own a Framework NixOS laptop (the newest one), a System76 Thelio NixOS beast (ryzen threadripper), and various Apple stuff. (And a Prometheus XVII Windows gaming laptop and an M1 MBP that I'm selling on eBay right now.)

So I completely get where you're coming from, because it's not a mutually-exclusive thing with me. As a software dev, I love open-source stuff!

I did just order the M4 Max Macbook Pro with maxed-out specs though, because I like to work on locally-running machine-learning stuff

Over years, I've worn laptops (with sealed-in batteries) down to three-ish reliable hours. There are never enough power outlets (or AC vents or even seats at the table) for a big meeting. That's a problem for a very long meeting format like a war room or a promo committee.

That's why you throw a couple of multi outlets in the drawer or a cabinet in that room.

Tech companies wire rows of desks for laptops and big monitors, but I think it'd be hard to find a meeting room where you could safely plug in more than a dozen 140 W chargers.

You probably can, as that's the max power, not the used power. To consume that much power you'd have to compile something with all the cores, and do something with the gpu (some rendering?), while the battery is being charged. And even then, I think there's a margin still on the power rating. (Running on solar, so I keep close eyes on my devices consumption). I believe most outlets are rated for 15+ amps, which is 1.8kW per receptacle (120v).

24hrs of web/office is a work day of something more intense, or a few hours of something crazy.

Apparently Qualcomm's new ARM laptops are pretty close.

However I think ARM platforms tend to be way less open source friendly than x86, at least on mobile. Maybe the laptops are better because they have to run Windows and Microsoft probably wouldn't put up with the shit vendors pull on Android. I don't know.

Except that ARM just cancelled Qualcomm's chip design license, so... oopsy

We might not know if it's a valid cancellation for another few months and many millions of lawyer dollars.

I wouldn't be surprised if Arm is strongarmed into playing hardball with Qualcomm by Apple itself.

Reply on news.ycombinator.com