Yeah, no. Maybe with old laptops, but newer laptops still have their fair share of issues. When I bought my thinkpad A485 kernels wouldn't boot without additional parameters, the graphics would freeze at times and cause a hardlock, sleep and hibernation have been fixed and broken again intermittently over several kernel versions, the wifi card's AP mode started causing segfaults in kernel 5.2 due to the driver's rewrite but has since been fixed, the fnlock key LED didn't update properly, which I spent a while debugging and submitted a kernel patch for, and while over the years the fingerprint scanner has been implemented, it's a pain to install and support for fingerprint scanning in linux is still in a very sorry state. Oh and bluetooth still can't connect more than one device at a time, so I had to buy a dongle to connect two joycon controllers.
Granted, I've always had these kinds of issues with new laptops, especially when it came to proprietary nvidia or AMD graphics (before AMDGPU) and I agree it's improved a lot, but I still need to tell people that there's caveats with some (especially newer) laptops.
With Wayland, Gnome and KDE have no way to adjust the scroll speed on a laptop trackpad. Not the pointer speed, the scroll speed.
In 2022.
That is the kind of basic thing that does not work.
In addition to that, if you have a high-DPI laptop display and you want to plug it into a low-DPI desktop monitor (or vice-versa), good luck getting the scaling to work in a usable way.
Like just give me a big text file with hundreds of tweakables and tunables like X had...
They hide behind 'you just need to get your client to make the right API calls'... but that just means most wayland compositors don't support most of the available options...
If we are talking about desktop Linux, a lack of config files is not a problem. If you expect people to edit files to get their desktops to work properly, you have already lost.
The same config pane where I adjust my pointer speed should let me adjust my scroll speed.
Config files are far preferable to some random database dug somewhere in registry of DE blob of stuff. Because you can make a simple program that just "does the right thing" for user then include it in distro
The storage mechanism is not the interaction mechanism, and Linux config files are not user friendly. All other desktop operating systems have a control pane for this stuff.
Configuration files are a developer-friendly and user-friendly way of supporting customization. It allows more conventional Linux users to do their customization using their favorite text editor, in the familiar interface that they already like. And, if somebody decides they want to write a GUI for configuration (maybe to chase users who are more familiar with consumer OSes), all the GUI has to do is write to a file (an easy task in most languages).
I'm a user and I don't find being made to directly edit config files user-friendly at all. Why on earth would I want to use a text editor to configure my printer? I don't use Powerpoint to edit photos either, even though in theory I probably could, but what a hassle. That sort of paradigm is fine for applications intended for professional use to be deployed to servers etc., but config management systems exist for those uses and the people managing these applications tend to be paid quite well, which is no coincidence. Targeting ordinary users? If there is no (decent) config UI, then it may be developer-friendly, but it's definitely not user-friendly. Yes, that can be a valid trade-off for resource-strapped hobby projects, and yes, it's fine if there is a config file in the back, but that shouldn't be the only way to configure an application. No other platform tolerates this sort of thing, and with good reason, given that most people find configuring software to be both intimidating and annoying even under the best of circumstances. I think devs actually believing that all users are just like themselves for decidedly non-dev applications plays a sizable part in why desktop Linux fails to break out of its tiny hardcore technical people and sometimes their parents niche year after year. Most people's favorite (i.e. only) text editor likely will be Microsoft Word, and I'm not sure how many of those people would say they like it.
You can check into git so you have a history of changes?
So you can copy the config to another machine?
There are lots of reasons why text files are the preferred format to store configuration in.
Other than perhaps a slight performance boost, why do we want settings in a non-human readable database?
Hell, even Microsoft are starting to use json config files for stuff like Windows terminal because they know people like to be able to quickly copy and edit settings.
You are talking about the configs being stored in text files. The comment you are responding to was talking about being forced to edit text files to configure.
Yours is about the format of data representation and theirs is about UX.
The first step of not forcing users to edit text files is having sensible well thought out defaults. If I have to think about configs the designers of the app failed me.
The second way to not force the users to edit text files is by having a well thought out gui for the kind of changes you might want.
The format of how the config settings are stored is almost orthogonal to this questions. And yes, you are right, a text based format is preferable over a properitary binary one.
It is a community-developed project, so it only really needs to appeal to developers. What motivation is there to attract non-technical users? Particularly ones who require lots of effort doing uninteresting polishing related tasks to keep them happy. Other platforms do this sort of thing because their entire reason for existing is to satisfy customers.
> What motivation is there to attract non-technical users? Particularly ones who require lots of effort doing uninteresting polishing related tasks to keep them happy.
First of all, adjusting scrolling speed is not an "uninteresting polishing related task," it is a basic standard of usability.
Secondly, if you don't think Linux on laptops should be broadly usable by the general population, you are in the wrong thread. The central point of the HN post we are all commenting on is the usability of the Linux desktop ecosystem on commodity laptops.
Except on Linux usually most environments stop at the config file part. And if a GUI tool happens to exist, it isn't kept in sync with the file format.
A convention for writing JSON, YAML, or whatever config files that standardized types of options, their default, and something about their types (is it a toggleable boolean, a bounded continuous value, an integer, etc etc) could be nice. It might be possible to provide enough info automatically generate the GUI (draw the first as a button, the second as a slider, and the third as a number box), and that sort of info would be helpful to have if you were writing the file by hand anyway.
Adjustable scrolling speed is useful (I mean, I haven't used it, but I can see why one might want it). Creating a GUI to adjust it is an uninteresting polishing task.
Swaywm has the ability to set this (you have to edit the config file). It seems weird that gnome or whatever you use lacks this option. Although, gnome has a lot of t's to cross and i's to dot, maybe they just haven't gotten around to it.
To add in the fancy slider you want so much, there needs to be a corresponding tunable in Wayland. MacOS does this with plists, it's not some radical or hated design pattern.
If you think that is a "fancy slider" you lack perspective. It is a basic, expected feature that is supported by Windows and macOS for laptop users. And I'm pretty sure it used to work on Linux too, before Wayland.
> In addition to that, if you have a high-DPI laptop display and you want to plug it into a low-DPI desktop monitor (or vice-versa), good luck getting the scaling to work in a usable way.
Sure? This is exactly the thing that Wayland was supposed to solve. Only X has one DPI for all screens.
I still use X because I'm on FreeBSD and I even got multi-screen multi-dpi scaling to work there, with xrandr settings but indeed it was not fun. In Wayland it should be click & play though.
Wayland supports it quite well but any app that uses the legacy XWayland compatibility layer does not work. Which is basically everything using Electron.
It's not a fault of Wayland but it is reflective of the whole Linux laptop experience.
It doesn't even require multi dpi. High dpi with XWayland is simply Broken in every compositor today: you have to choose between not scaling you Wayland windows and accepting super pixelated X windows (due to applying bitmap based scaling that almost no one would want with no option to disable it).
KDE's upcoming release in October should hopefully be addressing this by allowing you to disable the bitmap-based scaling.
You'll find they're basically identical to what you'd use to enable/force Wayland on Chrome. Also VAAPI {en,de}coding and pipewire based sharing
You can also replace --ozone-platform=wayland with --ozone-platform-hint=auto for less strong-handed encouragement
I use quite a few different Electron-driven things on Wayland. Discord is the only one seemingly refusing to update their Electron base... and getting free Wayland support
If not for them I'd remove XWayland support entirely from my Sway configuration
Well, it took a long time for all Windows programs to work somewhat decently, and even now the OS still does not have great multi-DPI support.
For example when I move my mouse from my 192 DPI screen to my 96 DPI screen, the mouse position translates in physical pixel, not in physical location. So at the bottom it matches but near the middle of the 192 DPI screen it stops going to the left (it already ends up on the top of the 96 DPI screen going from the middle of the 192 DPI screen) and becomes an 'invisible wall'. Even in Windows 11 they didn't bother to fix this :(
The only OS that had a good transition to multi-DPI capabilities was macOS and that's really because Apple doesn't care about legacy and forces app devs to update their stuff. But it's not just Linux that's having a hard time with this.
But I didn't know this was a specific problem. I'm not using Wayland yet and won't for the foreseeable future. I'm on FreeBSD and KDE on Wayland has been broken a long time. When I hear this it sounds like a good decision anyway :)
Windows' HiDPI does have some glitches (like the mouse positioning thing), but most apps work seamlessly (including browsers, Electron, and IDEs) or reasonably well (Microsoft Office). Also, Windows supports any scaling level, whereas macOS only supports 100% and 200%, which makes it easier to mask any sort of issues or bugs (pixel-doubling looks less blurry than scaling things up by 125%). Many (most?) 13-15 inch Windows laptops ship with a 1080p screen, which works best with a 150%-125% scale.
MacOS supports intermediary DPIs, it has multiple "looks like resolutions". You may have to press ALT while clicking "scaled". But you can't (easily?) set any arbitrary scaling, like 123%. You can on Windows, but it recommends against it. And I've noticed that most apps work fine at 150%, but many feel weird at 125% (which is also "standard").
The Windows mouse thing has been somewhat fixed in Win11 22H2, where you can now even move your mouse to the side "above" the other screen and it will still move there.
As for apps working seamlessly, I'm really not convinced. Not even the taskbar works well. If you change the DPI while it's running, the taskbar icons become blurry. The initial start menu (on first click) adapts fine, but then if you start typing to search something, the results are a blurry mess. Edge has weird artefacts in the tab animation after a DPI change, where half of the icon moves at a different speed. IntelliJ has funny fonts, with some of them huge, others tiny.
To me, the killer feature of MacOS when it comes to multi-DPI setups is that it remembers the per-screen-per-setup DPI. In my case, my PC has a 14" 1920x1080 screen. When I use it alone, it's much closer than with an external screen. I like it in 100% mode. When I plug in the screen, a 32" 4k, they're both much further away. They have roughly the same DPI (by design - I mostly use Linux) so there's no "matching" to do, but I'd like both of them to be at say 125%. Tough luck. If I change the laptop's screen to 125% while the external screen is plugged in, it will stay at 125% when on its own, too. MacOS would remember that with this screen it's 125%, alone it's 100.
> MacOS supports intermediary DPIs, it has multiple "looks like resolutions". You may have to press ALT while clicking "scaled".
The “looks like resolutions” work by setting your screen to the resolution it claims to be multiplied by two, and then downsampling the image to your native size. Depending on the resolutions involved, the screen might feel a bit blurry. On Windows, setting an intermediary scale changes the way the UI is drawn, while keeping your native resolution.
Do you honestly expect users to swap out the window system to solve a simple problem like a lack of scroll speed adjustment?
Most users won't even know the difference between Wayland and X.org and X11 unless they are already the kind of tinkerers who used Linux on the desktop despite its drawbacks. Normal people have no idea what any of it means, and they should not need to know.
I'm talking specifically about OPs complaint - they've identified it's a Wayland specific issue, so yes? If using Gnome, it's a single click from GDM, and your preference will be remembered.
X is purely worse for this case. Wayland supports live switching window scaling as they drag across monitor boundaries. X does not. The issue is that electron apps do not support Wayland yet so they run in the X compatibility layer which can not live switch. So reverting to X means you go from some of your windows not working to all of them not working.
Supposedly upstream electron has fixed this but I'm yet to see a single electron app that works. Maybe they just haven't updated electron.
Though it has nothing to do with Wayland before the flamewar starts, it’s just libinput and gtk maintainers not agreeing upon whose responsibility is it to handle scroll events (it is gtk’s though, libinput doesn’t have enough context to implement kinetic scrolling, so it really should be the framework that adds semantic meaning to an event stream)
>it has nothing to do with Wayland before the flamewar starts, it’s just libinput and gtk maintainers not agreeing upon whose responsibility is it to handle scroll events
Sure, but for me as an end user, it's irrelevant who's fault of this bazaar engineering endeavor it is that very basic quality of life features from Windows/MacOS do not work on Linux.
As a dev I understand the struggle why this and many other stuff doesn't work right on Linux, but as a consumer/end user I don't care about their internal feud and I expect the product I use to have basic stuff like this working out of the box.
Sure, I target it more specifically to the anti-Wayland group. I don’t really care whether they continue to use X indefinitely, or use another OS or whatever, but don’t attack an open-source project, and at least not on false claims.
Also, unfortunately the bazaar style of development sort of begets this kind end-user experience. Some people like it, others don’t. I change between OSX and Linux quite often nowadays, what I prefer in the latter is that I actually have a chance of fixing problems, not just wait around and pray to the Apple/Microsoft gods that they may have fixed the issue in the next multi-GB update. Also, piece-by-piece, free software often beats out proprietary offerings’ alternatives, it is usually the experience together with the whole stack that is lacking. E.g. pipewire may well be a better sound stack than that of the other two OS’s.
>but don’t attack an open-source project, and at least not on false claims
What did I attack and which false claims did I make?
>what I prefer in the latter is that I actually have a chance of fixing problems
What I and most consumers want is a product that does not require fixing or learning how to fix things. I and most other people don't want to play sys-admin at home despite having cut my teeth in it and making it a career.
I work in cybersecurity so all our workforce is fluent in linux which we daily drive at work and yet at home everyone of us only uses Windows and/or MacOS on our personal machines with only one guy using Linux religiously at home.
When even experienced linux users don't want it in their personal lives that says something. Even though we know how to fix things but our free time is much more valuable. Nobody likes a desktop that stutters and ruins your immersion and productivity, especially if you're running a system that costs several grand.[1]
Maybe when the hardware manufacturers can work with the bazaar engineers and finally agree on something and work together with the desktop environment devs on how to make Wayland a fully feature complete drop in replacement for X11 with no rough edges, quirks or issues and have feature parity, smoothness and polish to Windows/MacOS, we can finally have the "year of the (polished) Linux desktop". Until then, I and most consumers will continue to use whichever OS provides the best experience with least amount of friction.
You certainly know what you're talking about because from a lot of experimentation, my take on the state of 2022:
If you want the "works so well it's boring", go with X11. The one exception, as you note, is multi-DPI, which has native support in Wayland.
For Wayland, there are (depending on DE/compositor) some specific issues or inconsistencies, like the scroll speed you are mentioning. Personally, I also have qt5 apps being all over the place with window placement under wlroots. There are times when you'll need to look up some environment variable to make an application or toolkit behave properly.
So if you're in the high-DPI+low-DPI scenario, yeah, it still takes some effort. For anyone else, I think OP holds.
My pick for a "boring stable desktop" stack:
* Dist: Your preference of Fedora/Debian/Arch. (Mint, Pop, and Endeavour acceptable derivatives)
I can’t imagine not being in the “multi DPI” scenario though? It’s not some edge feature that affects a minority of power users, it’s just plugging an external display into a laptop. The DPIs almost never line up - not since high DPI displays became mainstream about 10 years ago.
I think this is more niche to care about than you might imagine. First off, a huge % of the population barely knows what an HDMI cable is and just doesn't plug their laptop into anything. Second, office workers might have a nice docking station setup, but in that case you are probably basically turning the laptop into a desktop (with a keyboard, mouse, and nice monitor) -- why bother with the laptop screen at all? Finally, perfect DPI matching isn't really required -- a 1080p laptop with a 1440p monitor should probably be fairly reasonable looking for most normal screen sizes.
Arch should not be listed as a boring stable desktop. In fact, just the other month it had a black screen outage involving grub that affected every user that did an update. If the distro requires the users to monitor an email list to avoid this kind of gotcha before doing an update, then it is neither boring nor stable.
It's not. I've managed to do it in xrandr. I have my main screen (and thus all screens) rendered at 200% and then scale down the other 2 back down to 100. Basically how macOS does fractional scaling.
It's not real multi-DPI no but effectively it does work. Does require a pretty decent GPU to render all screens at 200% before it scales them down though.
At one point I was using two 24" screens, one 1920x1080, one 3840x2160. I've tried messing around with settings, until I ended up on xrandr scaling as being, basically, the only solution. Five minutes later, the low-dpi display was in the closet, because I couldn't stand the blurry fonts.
It's ok for me but I used macOS before which also uses similar font anti-aliasing (macOS tries to preserve the look of the font in contrast to Windows which tries to align to pixel boundaries).
But indeed YMMV here.. There is no way to use non-anti aliased fonts at small sizes this way. For me it is fine, I would set it up the same way anyway but I forgot it's not for everyone.
For the past 20 years I have been only using Thinkpad from the T and the X series. The only one with an issue was I think the X220 with the SD card reader not being stable. All the other ones are working perfectly well. My current one is a T480.
But I always take some time to look if somebody succeed in installing Linux on the laptop I want to buy before. If it means I need to wait an extra 6 months, then I wait a bit.
I kind of expected the A485 to have similar issues to the T480, since it's the same laptop except for the cpu and graphics. I did some preliminary searching and the listing of the laptop as "ubuntu certified" gave me too much confidence, I guess.
I forgot to mention in the parent post that the SD card reader can't detect insertion/removal at times, yeah, so I have a script to reload the rtsx_pci_sdmmc kernel module to force it to recheck.
The A485 has really terrible support
! To the point I'm wondering if Lenovo didn't do it on purpose ... Very disappointed with mine to this day. Time definitely did not improve support.
same. it's been rock solid on thinkpads because thinkpads are some of the strongest pc laptops and as such have been popular (and well supported by) oss developers.
old IBM Thinkpads were pretty solid, but now they're made by Lenovo who has a long history of shipping devices pre-infested with malware and backdoors, usually in exchange for money. They've even been caught stuffing malware into UEFI so that users reinstalling their OS would be infected over and over again. However nice their laptops are, I could never trust them. They have already proved that they are perfectly willing to compromise your security and privacy to line their own pockets.
All the builtin radios, cameras, microphones, and sensors in modern laptops make them ideal for stealing your private data. I already have an untrusted cell phone, I want my personal laptop to be something I can feel comfortable keeping my data on. Because I can't personally audit every chip, that means I need some level of trust, and Lenovo has demonstrated over and over and over again that they cannot be trusted.
every once in a while there's a lenovo default windows image/hardware security controversy, but never one that has affected me directly.
i don't care what they put on the default windows partition (i replace it on arrival) and the uefi issue was a production mistake where they imaged with a nonproduction image.
they're still used widely by serious people in academia, open source and security sensitive industry.
i suspect a lot of the bad press they get comes from the fact that there's a lot of very sharp eyes making use of their gear and that similar issues happen in other lines but just go unnoticed.
if you're truly paranoid, a pine arm machine or fully open source risc-v may be your jam. everything else is going to be loaded up with proprietary blobs everywhere along with overcomplicated supply chains and overzealous marketing departments cross selling adware onto that default image you should be tossing anyway.
> i suspect a lot of the bad press they get comes from the fact that there's a lot of very sharp eyes making use of their gear and that similar issues happen in other lines but just go unnoticed.
No, it's really just them. They worked hard to earn that bad press. It's not even that they keep pre-installing malware, but how they've handled it when they're caught speaks volumes.
When the truth about superfish came out first they fiercely denied there was any security risk to anyone ("we have thoroughly investigated this technology and do not find any evidence to substantiate security concerns”), then eventually they admitted it was a problem and said they'd stop shipping devices infected by it, but continued to anyway more than a month later (https://arstechnica.netblogpro.com/information-technology/20...) and the instructions they gave users for removing the offending software still left systems vulnerable while giving people a false sense of security. When they were caught doing that they issued new instructions and those still left users vulnerable!! (https://www.theguardian.com/technology/2015/feb/20/lenovo-ap...)
I had a Framework and really liked it. Unfortunately the components just degraded really quickly for me, and now can't work but 20 minutes at a time until it freezes completely. Swapped out for a Thinkpad last week.
Even swapped out the Framework mainboard after a long back and forth with support. Just some poor battery unloading or similar causing shorts. I was very close to committing my company to using them until this started happening to my tester unit and my lead engineer's tester unit.
I hope the best for Framework -- I really love their repairability promise -- but before I can commit my company to them I need them to not be lemons.
Right now I'm liking System76. Expensive, but I like that they come with linux working out of the box. They're specifically designed/tested to work well with linux so no worries about the hardware not being supported. Like many other companies they are basically selling re-branded laptops made by the Taiwanese company Clevo. I still can't audit every chip in them, but at least I don't have clear documentation of repeated abuses (so far).
System76 is looking into making their own hardware now too so I'm really looking forward to seeing what they come up with in-house.
I got a Meerkat mini desktop 6 months ago, and I love it.
They do make their own desktops and minis now. I think they use Clevo for laptops, and those do get more complaints here on HN than the desktops (but I think the consensus is they are getting better). They have more laptop models, so making their own would be a huge task.
> a long history of shipping devices pre-infested with malware and backdoors,
Lol thats rich. They did it like 2 or 3 times for the windows laptops they sold most of the time not part of the thinkpad line. So yeah. Long history it is. You also have a long history of making bad comments then?
HP Elitebooks and Thinkpads are designed and built with Linux in mind. I never came across an HP Elitebook or Thinkpad which failed to run Linux out of the box.
Dell XPS is the latest addition to this group.
Consumer laptops come with a lot of trickery analogous to WinModems of the era, which require Windows specifically. Hence these cost saving measurements create a lot of problems.
They never fail, true, but there is always some special model capabilities that can only be made available after the usual weekend and late nights fine tuning.
This seems to me as a more generic problem with newer hardware, not specific to Linux. Likely devices rushed out the door to meet some idiotic deadline, badly specced and with incompletely implemented drivers.
At the end of 2021 I got an EliteBook 845 g8 (Zen 3) that worked completely fine out of the box on Linux (Arch with up-to-date kernel). Every last bit of kit worked perfectly. Bluetooth, IR webcam, fingerprint sensor, light sensor, mute LEDs, etc. On Windows, to this day, the webcam isn't recognized because of some USB chip along the line. There's also a lot of lag when adjusting the display backlight, for some reason.
I also have its cousin, an EB 840 g8 (intel 11th gen). A few days ago I installed Win11 22h2 on it. I was lucky to have had an external mouse, since neither the touchpad nor the track point could be used for setup. And it absolutely needs the latest Intel GPU drivers to correctly output 4k@60 through its HP dock (DP pass-thtough, not DisplayLink). On Linux, the same display setup has worked well since day one. But the mute LEDs are still broken.
Both laptops don't come with integrated wired network, so I have an HP USB dongle (Realtek chip). This works quite well on Linux. On Windows, it initially works well, but then, for some reason, Windows figures it needs to update the driver. Then it gains some interesting failure modes, where from the terminal I can do whatever I want, but Edge keeps thinking the connection is lost.
> At the end of 2021 I got an EliteBook 845 g8 (Zen 3) that worked completely fine out of the box on Linux (Arch with up-to-date kernel). Every last bit of kit worked perfectly. Bluetooth, IR webcam, fingerprint sensor, light sensor, mute LEDs, etc. On Windows, to this day, the webcam isn't recognized because of some USB chip along the line. There's also a lot of lag when adjusting the display backlight, for some reason.
Had an experience like this several years ago, but with hackintoshing.
On a Dell workstation laptop with a Quadro FX770M GPU (basically a relabeled Geforce 8800M GT), the Nvidia drivers had an issue under XP, Vista, and 7 where if the card downclocked when idle it'd cause Windows to bluescreen. The only fix for this for many years was to disable power saving features on the card, turning the laptop into a furnace even when it was doing nothing.
The proprietary Linux drivers for the card worked better (at least it could idle properly) but occasionally they'd cause your WM to lock up for no apparent reason.
The only thing that ran the card for extended periods without issues, of all things, was hackintoshed OS X. The built-in Nvidia drivers recognized it as an 8800M GT (which had been used in real Macs at some point) and it ran beautifully with power saving and everything. I even used that setup to play WoW on for several years.
The bug in the Windows driver was finally fixed at some point during the Windows 8/10 era, and so now I can run Windows on that laptop without problems, but holy cow it shouldn't have taken a decade (it was manufactured in 2008) for that to happen.
Arch is a rolling release distro. So it gets hardware support faster than other distros that stick to older kernels and userspace. Most users are not on rolling releases.
I personally like the rolling approach, but that doesn't reflect everyone's experience.
Arch/Fedora are fine if you are OK with rolling release, but there are options for newer hardware support in stable distributions. E.g., you can install your kernel and firmware from backports (or even unstable) on Debian, so you get a stable distribution with bleeding edge kernel (currently 5.19 in the repos). Or, you can go the other way and run Debian testing/unstable, and pin some packages from stable. And, Redhat backports new drivers into its frozen shipping kernel version, so some popular new (server) hardware will work with their old kernels.
HWE is not a rolling release, it's just backported kernels from newer non-LTS Ubuntu into an LTS version, releases every 6 months, so still very much a time-based release..., Ubuntu calls them "rolling" because when using HWE you get new kernel versions through regular "apt upgrade" once it gets released. But rolling release usually means that packages get upgrades as soon as possible after they are released upstream.
btw HWE isn't even the best "ubuntu flavored kernel" in terms of hardware support.., there are the OEM kernels designed for Ubuntu certified laptops (such as XPS 13 Developer Edition) which get newer kernel versions and drivers faster than HWE, you can install them on any Ubuntu with regular apt ("apt install linux-oem-22.04" for example) ...
Oh, my mistake; you're quite correct that it's not a proper rolling release. I still think that it's a step forward for drivers, but yeah that's still a significant lag especially when Arch is in the running. Didn't know about the OEM kernels, either; that's a nice touch for what hardware it supports.
I was looking at the new EliteBook g9 that has a i7-1280p and 64GB RAM DDR5 2TB, only $2500 (ouch). I would be surprised if Linux runs great OOTB. It is best to wait a gen before trying.
In all my years I have yet to see HDMI output from a laptop to a monitor work on the first try in Ubuntu. Always need to install the proprietary drivers for that to work at all. If it can't even do that without a headache, after 10+ years of Linux use, I would call that a Linux problem, not a hardware problem. My colleagues seem to run into the same issue frequently as well. This article seems kind of ignorant. I'm glad it worked on the first try on his ancient ThinkPad. That doesn't mean Linux is stable enough for most normal use cases on most hardware for me to recommend it to any of my non-tech proficient family.
What Laptop is this? I never had an issue with the external outputs on any laptop I owned (and I've been running Linux since the 90s). I also don't know anyone who had these issues. The main issues I had were typically docking stations and suspend (but that has been super stable for my last 3 laptops).
Laptops with iGPUs usually work pretty well. The ones with built-in discrete graphics cards can become kind of a mess to configure. A friend had one where, if I remember correctly, he'd gotten the built-in screen working with (I think) the iGPU, but anything into the HDMI port switched it over to the dGPU, which had some of those crappy NVIDIA drivers, causing both screens to shut off or something like that. (I didn't debug it so this is just an outline of the problem).
This was a long time ago, but I had an HP Envy 14-1000. It had an Intel iGPU with a separate AMD card.
It was a muxed setup. The screen was switched back and forth between GPUs and one would power off as needed (assuming everything went well). The HDMI port was only connected to the discrete GPU. T here was no way to get video out on the Intel card. By default, Linux would power on both, but use the Intel.
This was well before any AMD cooperation, and I had the laptop much longer than the FGLRX setup was supported. The open source Intel driver and simply turning off the AMD card was eventually the only way I could get it to run.
Even in Windows it was a strange setup. You had to manually switch, and when you did the screen would turn black, you'd wait a few seconds, and now you were on the other GPU.
I'm sure the situation is better these days, but after that experience I just stick to integrated.
Windows 10 fixed this sometime around the end of 2020, where you have igpu and a discrete - prior to whatever patch they pushed, if you alt tabbed out of a fullscreen game, there was a good chance that the game window would be a 1" square when you alt tabbed back, if it didn't crash in the meantime.
The "black screen for a couple seconds" thing is still there, you just don't notice it, and once a game has "started" the discrete GPU, you can seamlessly switch back and forth.
some people are mentioning that "i can't believe it took 10 years for this to get fixed" - however back in the late 90s this exact scenario was the most common power gaming setup, with 3dfx cards you'd have 3 cards, two 3d cards with SLI, and a 2D card, usually an intel. The same black screen for a couple seconds, and switching between the desktop and a game had the potential to break things.
The "automatic" switching between igpu and discrete was managed on windows before 2011, because i had a laptop with that setup in 2011 and it would detect 3d applications and use the discrete for that, or you could force one gpu or the other, if you wanted.
I've never had any issue with HDMI or DP output. But it's true that my only laptop with a dedicated GPU was an MBP, all the others have or had integrated graphics.
Intel seems to have the best GPU support. My Dell XPS from a few years ago works fine with Thunderbolt dual monitor dock and USB-C to HDMI adapter.
My desktop with a AMD Vega 64 crashes weekly (with occasional stable months) running Fedora (usually about 1 minor version behind mainline) since I've gotten it (maybe 3-4 years ago now)
If you had problems with the HDMI output, I suppose that it must depend on the GPU model.
I have used Linux on many laptops and I never had problems with the video outputs, but most of them had NVIDIA GPUs and a few used the integrated Intel GPU. I have no recent experience with AMD GPUs on laptops.
I do not normally use Ubuntu, so that might matter, but when I bought a Dell Precision, it came with Ubuntu preinstalled and it worked fine until I wiped Ubuntu and I installed another Linux distribution.
I used once a Lenovo on which I had to waste a couple of days until I made the GPU work properly in Linux, because it was an NVIDIA Optimus switchable GPU, but even on that laptop there were no problems with the video outputs, but only with the OpenGL acceleration, until it was configured in the right way.
Your laptops have Nvidia GPUs, I suppose? In my experience, that's the one brand to avoid when shopping for laptops to put Linux on. (Though you can usually get things to work, with some of effort.)
It's not about new vs old but who makes the main parts and chipsets. Intel everything is always a really good bet, even when they're brand new, but there are other safe choices.
It used to be quite hard to find new laptops with hardware combos that worked well with Linux but it's become a lot easier in recent years.
Also my experience with windows has actually gotten quite a bit worse, actually, unless you use the stuffed-full-of-garbage oem installs I've found it way more likely that I get stuck in a catch 22 where there's no network drivers for either the Ethernet or wifi so you wind up downloading some drivers off a sketchy site to put on a USB stick just to get started.
I think in general Linux users develop a tolerance for stuff not working so they say "This works perfect" when the reality is that a lot of the features are not working but they are still able to get day to day work done. Stuff like finger print readers and often web cams not working is borderline expected. On OLED monitors you can expect brightness adjustment and often sleep/wake to not work quite right if at all, etc.
> I think in general Linux users develop a tolerance for stuff not working so they say "This works perfect" when the reality is that a lot of the features are not working but they are still able to get day to day work done.
For me, every OS has rough spots and it's about which ones I can tolerate the most. On Linux I get better window tiling than on Windows, and shortcuts for navigating directly to a virtual desktop, and no shenanigans with WSL2 having a separate memory pool from the rest of the OS. And I don't feel like the entire OS is antithetical to how I use a computer like with macOS.
But a bunch of more mundane things become a lot more fiddly or flaky. E.g., this week openSUSE Tumbleweed pushed out Gnome 43 before any of my extensions got marked as compatible and now they just won't work for a little while. That's easier for me to live with when the OS is well suited for me most of the time.
> For me, every OS has rough spots and it's about which ones I can tolerate the most.
+1 if you're looking for some anecdata. The thing that finally pushed me from Windows to Linux was a privacy setting not actually being persisted (after a long battle to find the relevant settings). The fact that some wireless network cards don't work yet is definitely a rough spot, but I can also just buy a new one or write a driver, whereas getting Windows to care about my privacy or MacOS to care about basic usability with respect to keyboard remapping or window positioning seems unnecessarily daunting.
> I get stuck in a catch 22 where there's no network drivers for either the Ethernet or wifi so you wind up downloading some drivers off a sketchy site to put on a USB stick just to get started.
If you've got an Android phone and a USB cable, you should be able to USB tether to your phone's WiFi connection. This should work out the box on Linux and Windows.
man i remember those days when you had to go and find all the software you use from random sites and same with drivers, then having issues you can't figure out because of a outdated old driver vs Linux where you just get the latest stuff all in once place.
how do people on windows figure out what driver has updates? do you guys check the version installed and go to each manufacture to see if there is a new version>?
Why would anybody expect that new hardware that has to be reverse engineered and for which no Linux drivers are provided to work out of the box as soon is it is available? It's an impossible expectation, and also one that no other OS would have a chance of fulfilling unless hardware vendors specifically catered to them.
It's like asking for a book review of a book that hasn't been published. Yes, other people have published reviews, but they got advance copies and a supplementary synopsis from the publisher six months ago.
>Why would anybody expect that new hardware that has to be reverse engineered and for which no Linux drivers are provided to work out of the box as soon is it is available?
The title of this post is "Linux on the laptop works so damn well that it’s boring".
Because that is there experience and has been my experience too. Except we can't predict when a user wants to go install a old disto with a 7 year old kernel and say "no it doesn't"
Linux users don't have to either. (Linux devs do.) Another thing they don't have to do is searching vendor site for drivers which may not even be installed correctly.
It's very rare to have to search for drivers these days on Windows. Last time I had to do so a few years ago, it was for an old IrDA receiver - and I doubt many people these days even know what that is.
Almost any hardware you get today will have standard components with Linux support. Can you give some more examples of which companies and what components you're referring to?
There's a long list of devices that have problems with linux. The hardware I think I've seen the most complaints about are wifi/ethernet chipsets, printers/scanners, and spotty support for specific features like sensors, LED lights, and fan/cooling controls. Problems with video cards are far less common now than they used to be but I still see people having weird issues from time to time and sadly most of the firmware still contains closed source binary blobs.
RGB control is quite spotty. Finger print readers are almost entirely non functioning. Often webcams on brand new laptops do not work. Brightness control for OLED often doesn't work. Suspend and sleep still seems to be an issue, especially with something relating to "modern sleep" where the OS remains functional for things like internet access while sleeping.
I think largely these days people just plug less stuff in, printers, scanners, other odd gadgets are less common so its really just USB mass storage devices and video outputs getting physically plugged in. Otherwise everything else is controlled over wifi with a phone app.
I don't expect it. Which is why I switched from buying Windows laptops to install Linux and now buy MacBooks where I can expect that absolutely everything will work on day one.
My wife's Macbook can't even go to sleep when it's plugged into a Dell monitor via USB-C. The act of sleeping wakes it up, sounds a ding, then it goes idle and soon will sleep again.
My older Mac Mini can't do the HDMI handshake after booting. If the AV receiver isn't on when the computer boots, it will never use full display resolution or play audio over HDMI.
The idea that Apple products do anything "absolutely" is silly fanboyism.
Not doubting that situation but its pretty rare. Currently use a Macbook for work and home use and can't list a single thing that doesn't work. Which is a first for me from a laptop.
Dell Latitudes have been relatively painless for years with Ubuntu/Kubuntu LTS, probably because Dell sells a version with Ubuntu preinstalled. Still, Dell doesn't have fingerprint reader support in Linux, and the built-in card reader needed additional setup, but other than that it just works on a fresh install. Even my favorite Windows games work on Steam with Proton, if you accept minor texture glitches, which I gladly do to avoid dual booting.
I love it because these days I have less time to fiddle with it every six months.
My main issue with my current laptop is that the synopsis touchpads connect over i2c, and there has been a lot of ongoing work in the kernel that keeps requiring me to change my kernel config (PINCTRL_AMD needing to be selected for the 5.18 to 5.19 kernel update).
My last laptop (an AMD version of the HP Envy 13) was also rough at the beginning. A BIOS update updated the AMD GPU firmware or microcode or something and broke compatibility with the current kernel stable kernel at the time. Had to switch to an -rc kernel to get video to work.
Admittedly, my day job is basically Linux kernel development so I'm intimately familiar with most of this stuff. Not exactly your typical user.
I was going to cry foul since 5.18 and 5.19 aren't considered "stable" by most maintainers, but you mentioned that you do kernel dev, so it makes sense. 5.15 is "stable", at least on gentoo.
> I was going to cry foul since 5.18 and 5.19 aren't considered "stable" by most maintainers
Distro maintainers certainly, unless you're Gentoo, Arch, or one of the other mostly-bleeding-edge rolling release distros. The "stable" kernel is whatever the current release is and "longterm" kernels are typically the last major kernel version released in a given year.
Most distributions pick whatever the latest longterm kernel is when they cut releases. Sometimes they don't and things get strange, such as when Canonical chose kernel 4.15 for Ubuntu 18.04, requiring them to maintain an unsupported kernel themselves. IIRC that was because a bunch of AMD CPU and GPU support was added in 4.15.
For new laptops, if you have the option, go for something that has some official support out of the door. Something like System76 or your local equivalent. Otherwise yes, it can still be a bit of a lottery. Everything could be smooth, or you might become your own Linux support guy. Some people enjoy that.
I've mentioned this in a different subcomment, but I should note that the laptop I mentioned in the parent post is "ubuntu certified"[1]. I realize now that this means much less than having "official" support from the manufacturer, but there's certainly a lot of misleading bits about the way these things are marketed.
Pop has gone from something that was expected to "just work" to deciding it's more of a developer or enthusiast product.
Linux still needs an 'it just works' version. I really thought pop would be it, but the last year of development has been very disappointing with system breaking updates being pushed (I'm on system 76 hardware).
My Ubuntu LTS "just works" until the laptop decides it needs to have an heavier connection to download big data, then its wlan infrastructure dies and needs to be restarted.
It "just works" until I decide to view videos online, and despite all my hacking efforts, keeps using software decoding for videos.
It "just works" until I try some 3D stuff that requires GL 4.1, but the AMD open source driver for the GPU only does hardware acceleration up to GL 3.3.
Sleeping "just works" until I wake it up and is in such a frozen state that only taking the battery out makes the booting process work again.
I am pretty disappointed with System76. On my Galago with PopOS suspend doesn’t work properly , the fan sometimes goes into super speed mode overnight, the screen flickers when it wakes up from sleep so I have to reboot. It’s definitely not a smooth experience. Support couldn’t help either.
Yep, my 2 year old Ryzen laptop still doesn't have properly functioning sleep without tweaking kernel parameter, and that workaround got broken on 5.19 kernel release and I had to find out new parameter to tweak.
Linux worked perfectly on my old laptop from 2015 though.
Out of curiosity, what laptop? My samples in the last 6 years are:
- Thinkpad Carbon X1 14" (i7-5600u). Everything worked out of the box with Arch Linux at the time. Best experience I've ever had.
- HP Envy 13z (R5 2500u) everything works today but the out of the box experience was very poor. Windows update installed an APU microcode update that broke the Linux AMDGPU driver and had to run an -rc kernel for awhile. Took a year to get a touchscreen driver and years to get the driver for the tablet sensors (rotation, etc.). Total wait of 3 years for all features, but I never had the desire to use it as a tablet so I was okay with it. Sleep works but this laptop had awful battery drain issues in sleep (30% per day).
- Dell XPS 15 7590 (i9-9980hk) - Sleep is broke in both Linux and Windows. Everything else works well, including, notably, NVIDIA Optimus / DRI PRIME.
- Asus ZenBook 14 (R7 5800U) - second best out of box experience. Touchpad is connected via i2c and my Gentoo install didn't have it enabled. I'd never bumped into i2c hid devices other than touchscreens.
> Dell XPS 15 7590 (i9-9980hk) - Sleep is broke in both Linux and Windows.
Yes! How can they sell these like that? My XPS 13 will never go to sleep correctly, either the screen stays on or it doesn't shut off correctly, in Windows or Linux. You'd think that this is the basic feature a laptop has to have. And it's not just me, their forums are full of people having problems and their support has no idea. They were sending me guides for latitudes from 2012.
Lenovo Thinkbook 14 G2 (Ryzen 7 4700U), old laptop was Dell Latitude 7xx0 (or something, don't remember exact model), with Intel i5-5300U CPU, Linux worked perfectly on that machine.
These kinds of things probably still depend a lot on the brand and the product line.
The post is really only an anecdote about a ThinkPad, and a relatively old one at that, which is probably as good as it gets in terms of Linux compatibility.
I personally more or less agree with the title, though, assuming a suitable hardware choice. I have a new-ish Ryzen ThinkPad for work and the only issue I've had is Gnome occasionally semi-hanging, and I don't know if that's just because of Ubuntu being a bit flimsy or because of something more general such as an issue with the AMD graphics driver.
Also, the Teams client the post mentions is about to be dropped by MS and it was never really that good to begin with, but having seen about two decades of desktop Linux, I'd rather be surprised that it's been available and worked somewhat reliably at all without hit-and-miss with Wine.
I tend to agree. If you pick a random new laptop, you will probably have a bad time running Linux.
When I decided to switch to Linux as my main OS, I researched well supported models and settled on the X1 Carbon. I bought it a large discount right after a new generation was released and the Linux support has been near perfect. Really only one or two minor issues in the past ~3 years, which is similar to what I have experienced with most Windows and macOS devices.
I had a X1C bought early 2018 and it had great battery life. And I generally was happy with it. Unfortunately it got stolen (luckily I had full disk encryption). Without much investigation I took a X1C again. This time it was a 7th gen. It had HiDPi display which was a bit of pain to have all apps working with a satisfactory resolution and the battery life is much poorer than before. I would take my older one back right away if I could.
Edit: The author uses an 11 year old machine. Not a surprise it works well. With all the new stuff the vendors introduce difficulties are much more common. I hear a lot of complaints from colleagues with Thunderbolt docks, the newest Intel camera generation has no Linux support, not that much has changed. Whether it's 2 steps forward and 1 step back or the other way round is debatable.
> Maybe with old laptops, but newer laptops still have their fair share of issues.
Even "Linux works damn well on your ancient laptop" is a great selling point. Want to run Windows or macOS on an ancient machine? You can run an insecure ancient version, or, if the up-to-date version can even be installed, it'll run at a crawl. Linux makes those machines still usable.
Interesting; I got a A485 two years ago and everything except Bluetooth worked out of the box without problems (BT kind-of works but with a lot of problems, so I just got a USB one for €10). Never needed any kernel parameters, graphics work fine (including for some games), sleep always worked fine. Never tried AP mode or fingerprint scanner. Only reason I got a different laptop was due to hardware issues (some issue with the mainboard).
I only used Void Linux on it; maybe it's different with other distros.
I'll be honest, aside from the FnLock LED issue, most of the issues I've mentioned were probably fixed by 2020. I got this laptop in january 2019, and none of the stable distros (ubuntus and such) would even boot without kernel parameters back then. 5.1 became the golden kernel version for me for a while, where everything worked as later kernels would break suspend/hibernate a few times before stabilizing properly. It was a rocky few years but by kernel 5.10 (december 2020) everything worked fine, probably earlier like you've mentioned.
Graphics always worked fine except for random full system lock-ups/kernel panics in amdgpu which have been fixed at some point I don't remember when. I have no idea what caused them but a kernel option (something with iommu) made them go away until it was properly fixed, and I think that wasn't exclusive to this laptop. Graphics are still scrambled when waking from sleep though, but they take a split second to restore. The rest of the problems (bluetooth, fingerprint), still persist.
My daily driver is a Thinkpad x230 running Linux and I also vehemently disagree with his author's claim. Yes, things are certainly better than they were 10 years ago, yet I still have a lengthy to-do list of things that are broken, but low-impact enough to work around that I don't want to waste the hours of dicking around with poor documentation, noxious forum threads and command line fiddling to sort them out. Circumstance do force me to finally tackle one of these issues regularly, yet the list never seems to get shorter...
Like what? Must be things I do not use I guess, which definitely can be the case as I never plug in external monitors etc, but everything I use works very well for dev work.
Unfortunately this seems inevitable. Hardware manufacturers have to support windows or they won’t survive, so that support is going to be there day 1. Some of them have spotty support for Linux, most have none. So it falls on the community to buy the hardware and iron out the issues.
Can confirm, my god how terrible BT support can be.
So much for things working on older laptops, my 6ish years old Asus as some weird Intel BT chip that has completely broken drivers on Ubuntu. Not as in that they can't be built or installed, but the damn thing keeps fucking disconnecting and reconnecting every few seconds. It literally would've been better if they hadn't bothered.
But also like in general, at least anyone making any new protocol or standard can rest easy knowing that they cannot possibly fuck up worse than IEEE making the bluetooth spec.
This has also been my experience with a new ThinkPad P1 G5. Wifi didn't work out of the box with stable releases of any of the distros I wanted to use, I had to use the testing release, and even then the wifi is unusably bad unless I'm sat right in front of the AP (all other wifi devices in my home work perfectly fine from any room).
Putting the OS or even just the display to sleep causes the whole thing to completely freeze, forcing me to hold the power button until it shuts off.
Other than that, usable, but some really bad quirks that would make me switch back to Windows if I didn't have workarounds (use an ethernet cable, never let the display sleep, never close the lid while the laptop is running).
My Vostro 1400 (Core 2 Duo) has kernel bugs on sleep-wake related to the Ricoh xD media reader (remember those? I don't either), and my Inspiron 15R SE (Ivy Bridge) randomly disconnects from all external USB devices until I use the internal keyboard to remove and reload the xhci_hcd kernel module. And my Ideapad Flex 4 models have a bug where you can press Page Up, release Fn, and release Up, which on Windows stops sending Page Up events but on Linux results in a stuck Page Up key (technically a laptop bug but affects Linux far worse). So older laptops are by no means trouble-free either.
How do we fix this? It seems like most people in the community have the mentality that these issues work themselves out eventually, so it's no problem. And there's nothing wrong with that. But a lot of people really really want better hardware support for new hardware, and fewer regressions in drivers for older hardware. Perhaps we need a special-interest group that keeps track of ongoing hardware issues, and generates kernel patches.
As a user of the E485 (basically a budget version of the A485) I can confirm, and agree with, everything but one thing here.
Regarding your Bluetooth issue, what chipset do you have?
I picked a Qualcomm one on mine because I had bad experiences with Realtek before. Never had an issue like yours with it.
For me its quite a usable machine now. But I'm currently giving a M1 Macbook a shot and it certainly is convenient not to have hiccups like this (yet).
Well, the reality is that Linux consists of less than 5% of the market (desktops and laptops), where Windows OSes make up around 75% and macOS around 15%. So, that is going to dictate the priorities of companies supplying drivers.
These numbers are always so skewed. I wish just one time we could get market share numbers for new sales WITHOUT corporate purchases, but Microsoft has made this a "thing" for 30 years now. On the one hand manufacturers would still want to target numbers, even if that includes corporate "fleets." On the other, they'd see that Linux makes up a larger portion of the market where people are spending their own money.
I've had multiple Thinkpad T-generations from T410 to the latest. Sometimes it does works flawlessly out of the factory at purchase.
This time, it did not. The 12th gen Intel CPUs have a heterogeneous design with traditional "P-cores" and low-power "E-cores". I'm suspecting the reason I see terrible performance is that the CPU scheduler does not handle this efficiently and assigns the wrong task to the E-cores.
Also the Intel WiFi does not even get detected. Have not dug deeper into that yet.
Anything > ~6(Intel) ~12(AMD/Realtek) months old tend to work smoothly out of the box, IME.
To be fair, this is also an issue with servers. I bought a server from a Linux server vendor and the chip was too new that it wasn’t supported on their custom Linux OS (same company, but the hardware and software sides didn’t communicate). Thankfully, it was supported on CentOS at the time, so I was able to switch pretty easily.
I just mention this to say, this can be an issue with any recent hardware. With Linux (the the most part) drivers are built-in and vendors do often ship drivers, so we have to wait sometimes for compatibility.
I have an E15 from work and have a myriad of just _strange_ issues. When I first got Fedora installed, the integrated mouse/keyboard didn't work without some kernel parameter tweaks. This was eventually fixed. Right now I'm dealing with some random crashes(?; there's no stacktraces, it's kind of annoying) of XFCE that I'm blaming on the 12th gen intel GPU firmware, but I have no evidence to back that up yet.
The annoying thing is that it's quite unpredictable. You can sometimes find information on the internet on which laptops Ubuntu is going to work on out of the box, but usually it's a gamble. Sometimes you notice something's not working a month after buying the laptop just because you never happened to try the feature before.
Old laptops & desktops too have been left in the dust for some time now.
Decades ago one of the most important benchmarks of Linux distributions was they were all higher performance using less resources than the original Windows that came with the PC.
If you have a PC with only 1GB of memory which still works fine with XP or W7, most distros are now unusable.
Thanks for honesty. During my studies I really tried using Linux on my laptop for a few years. It was amazing to tinker around. But when I finally switched to Mac OS, I felt I became instantly more productive.
That was 15 years ago in 2007. I never went back. Now macOS has its struggles, but I can work and focus on a clean UI.
Did Lenovo classify the device as Linux compatible? A lot of laptops from Lenovo, HP and Dell have Ubuntu or RedHat as an optional pre-installed operating system. Those devices usually work with Linux.
As a regular customer you can't order it with Linux though, it is only sold to enterprise customers.
> sleep and hibernation have been fixed and broken again intermittently over several kernel versions
How do people typically learn to debug kernel issues on their hardware? It seems like actively promoting widespread knowledge of the practical methods would benefit the community.
That and he says “most software has migrated to the browser.” Maybe most of what he uses, but if that’s the case then you’re basically doing a DIY Chromebook.
Unless you have a thinkpad or some other popular hardware, you'll find Linux barely works at all out of the box, and even with hours of fiddling around, you'll still have to live without some features.
For example, power saving features, sleep and hibernate, screen brightness controls, fingerprint readers, keyboard hotkeys and backlights, etc. rarely work. Prepare for broken external hdmi ports or USB stuck at USB 2.0 speeds. Have fun with the fan stuck on either max or zero, or the CPU stuck at the lowest clock speed.
There are still lots of things you have to go hunting for the right old firmware version for.
I think Linux is only great if you have whatever hardware distro developers have, because that will be all that works out of the box.
> Unless you have a thinkpad or some other popular hardware, you'll find Linux barely works at all out of the box, and even with hours of fiddling around, you'll still have to live without some features.
I've had it work first time, perfectly on:
- Tongfangs, 3 different models
- Lenovo, many different models
- Clevos, 2 different models
- Asus Zenbooks, 2 different models
- Too many Dells to count
- Asus Zen2 desktop
I have yet to find a device it doesn't work on. I've never had to mess about with the kernel params or do anything clever with fans except install the sensors package and run it.
The only shortcoming I've noticed is it the fingerprint readers were hit and miss, but this is mostly because the device manufacturers didn't bother with drivers.
In my world, running perfectly means, runs at least as good as windows.
Same battery life and performance. No glitches with suspend, hibernation, etc.
And I doubt that.
Not because windows is so awesome, but because hardware manufacture write and optimize their drivers for windows.
And linux is a way smaller market and one with intentional no stable driver ABI (to force the vendors to open source their drivers).
That is the situation. And it sucks, because I do not like to use windows, but I need my mobile devices to be reliable.
There are often things that do not work under Linux on different laptops and it depends on what you need and expect. However, consistently over many laptops (I am a computer hoarder and have 100s of laptops), after running my post install script, battery life always easily beats windows. For old and new laptops.
Recent example; someone gave me a cf19 which had a dodgy battery; it had the official Panasonic windows install with the Panasonic drivers and optimiser and the battery went from full to 0% in about 15 minutes; I did a windows 10 install, again with the official drivers and got about 15 minutes. Did a windows 10 install without drivers and got about 15 minutes. Did Ubuntu install with my post install script and get consistently 3.5 hours. Yes, I know this is a weird case; I bought a new battery by now and still get the same behaviour; no clue what’s up. I thought it was funny though because ‘windows better battery life’ people.
And this always happens; my x220 runs 8+ hours under Linux while barely getting to 4 under windows for the same work. GPD pocket 1 runs over 10 while under windows getting not even 4. And these are the ones I use all the time for work. I consistently see this and have no clue what people are doing who claim windows has better battery life; must be much different workloads. I use i3wm (which I believe makes all the difference; with Unity, battery life gets slashed by up to 80%), code in vscode and vim and browse in Firefox; but I do that under windows as well.
You are free to have a look at my devices, but I did manual optimizing, tld, thermalD, optimized grub settings, allmost bricked devices with it and still even stock windows is better.
But a proper installed windows, without bloatware and updated drivers, always won by a large margin.
And the bigger showstopper are standby/hibernation. Touchscreen glitches etc.
So I have 2 devices, one lightweight chromebook. A bigger windows laptop and manjaro on the desktop.
Yeah, the things like standby and hibernation and touchscreen are major issues especially in new laptops. But I never use these features myself. Guess I got used to it. Which devices (specifically) do you have? I might have one or both or know someone who has.
Out of the box linux experience is actually quite good. Performance loss I can handle, but standby-resume reliably introduced touch screen bugs and other annoyances. Or did not wake up at all. I actually just recently gave up with linux on it and installed windows.
(which was surprisingly a big huzzle, I had to resort to third party software to get the right windows drivers. But now it runs way better - and more importantly, more stable.)
I get that; a computer is a tool; it might be a love but mostly it’s a tool and it just has to work. That is really why I installed Linux in the first place back in the day but maybe things did reverse. I have a script which just makes any apt machine into mine; it is something I cannot do on Mac and win anyway and that is enough reason; I can get a laptop and run that script giving me a full working station in about 10 minutes. That is already nice. But then I don’t have much wishes; if it’s very heavy, I will offload to servers and I don’t need a lot of graphical stuff outside a modern browser.
" I can get a laptop and run that script giving me a full working station in about 10 minutes"
I even made my own manjaro (arch linux) version. So I have a USB stick with all my setup and programms, I can plugin and go hacking on any computer.
Or directly install it. Takes only 3 minutes ;)
> Same battery life and performance. No glitches with suspend, hibernation, etc.
Other hardware can be sketchier, but I've found power handling much more reliable on Linux than Windows. It just works on my Linux boxes, whereas my Windows machines I would regularly notice that my previously suspended laptop was 130F+ roasting its bag and emptying its battery, or I'd find it completely dead after the same thing had happened and I didn't catch it in time.
I had a Thinkpad with Ubuntu and still had many of the problems you mention and more:
- Barely ever waking from sleep, especially with external monitor connected
- Screen brightness keyboard controls didn't work (needed to use a CLI tool to control gamma as a hacky workaround)
- Had to power cycle repeatedly to get to a desktop when booting
- Not working reliably in clamshell mode
- Randomly forgetting external monitor scaling
- Accessibility features like screen zooming are very poorly done compared to Mac's Ctrl-MouseWheel (which zooms entire screen without crashing)
Things actually got worse as I upgraded to newer kernels. The wake from sleep problem is the #1 productivity killer I had. I had to leave the machine running all the time just to do my job.
Sleep has become less of an issue recently, at least in my experience. Modern laptop CPUs idle in such a low power state. I just set up my built-in display to disable when the lid is closed. Seems sufficient.
If you leave it anywhere in a bag not plugged in, then you have to be constantly thinking 'I only have 30 hours left till the battery dies and I lose everything I had open'.
I'm not 30 hours away from electricity very often, to the point where it wouldn't be a big deal to just save and shutdown, since it is a rare occasion. But I live a very predictable lifestyle, I guess if you are exploring London you probably end up in unexpected situations often.
Yes, running Linux on Windows hardware is often a recipe for misery, or at least dealing with obscure kernel parameters.
Which is why I've said and will say again: slapping Linux on Windows hardware is a mug's game. Buy it preinstalled, from a company that supports it. We actually have that option these days, and it's amazing.
Some days, I swear the smartest thing Apple ever did was prevent users from slapping OSX on commodity Windows hardware.
Or just look up your "windows hardware" before you buy and check compatibility. Companies that sell laptops with preinstalled Linux are far more often than not just selling rebranded "windows hardware." The benefit is that you get a support number, and that they have paid attention to the Linux compatibility of the models in their range.
Sure, I can look up Dell Linux products. However, the assertion was that they're just branded white box computers, so Im asking what the white box equivalents are, and to what extent they're really equivalent.
I didn't understand it that way at all. The assertion, as far as I understood, was that the Linux machines were the same ones that were sold with Windows on them otherwise. Which is true, Dell sell the same hardware with Windows or Linux.
Is the firmware confirmed to be the same? Even with the same chips, the firmware would likely diverge between Windows and Linux (ACPI code, BIOS, EC, etc)
From everything I've read, ThinkPads (IBM/Red Hat devs seem to use them), Acers (have pretty standard parts, nothing funky), Dells and HPs (both have Linux dev laptops) all seem to run pretty well.
The worst seem to be gaming laptops, non-Lenovo Chinese brands, Asus, etc...
Corroborating this on the asus, I got one back in 2020 when my laptop gave up the ghost. Tried trudging through but I couldn't get comfortable with opensuse, things like on boot the mouse not responding, spontaneously rebooting when I tried to change volume, wifi card being throttled or just useless. All of it failing just enough I never quite trusted it.
Less than the brand, I think it is the line and components manufacturers that counts more.
Like pro lines are what most linux devs receive from their employer and better supported than familial and gaming lines. Also intel integrated everywhere is better supported than a mix'n'match of chipset foo, network bar, gfx baz.
That has been my experience too. Even then if you get a next generation thinkpad that is slightly newer than what has been "blessed" by the community, there is a good chance that a lot of essential hardware won't work. Fortunately, in the case of Lenovo they do actively track issues with hardware and issue new bios versions that fix compatibility but even having to install new firmware when you are using Linux can cause major headaches and worries.
ya that service has been very helpful and tracking issues is great, but there can still be issues when installing. For example, I am dealing with a bug found in this list of issues: https://github.com/fwupd/firmware-lenovo/issues on one of my laptops.
But isn’t that basically “good enough” if you know you want Linux and can either afford the latest thinkpads or are okay with a slightly bulkier older Thinkpad?
Seems not worse than different from needing Apple hardware to use Apple software… (though in practice there is a significantly wider array of hardware that has very good support for the software)
I slapped Arch Linux on a new HP 2 in 1 and everything except for the fingerprint reader worked out of the box, including the stylus. Not even a single controller of my weird Chinese 10-port USB-C dongle refused to work.
I've been running manjaro on a dell xps 15 2-in-1 without issue for about 3 or 4 years.
The only oddity is that it has the intel kbl-g gpu, so sometimes you have to manually choose which gpu to use if the app is badly behaving and you don't want it to suck your battery dry in an hour.
I have an Acer Aspire and a self-built PC and both of those work just fine with Linux. All the hardware drivers work on both of them no problem. I didn't even build the desktop FOR linux, it just works when I run it.
Good catch. The traditional problem (from the era before T420) is waiting for the kernel to catch up with the new hardware, for any kinks to be shaken out.
At one point, there was a joke, if you wanted some new hardware to work with Linux, the easiest way was to buy two of them, and give one to Alan Cox or similar.
Then Linux became mainstream, and you had dynamics like Lenovo wanting Linux to work well at launch of a new ThinkPad.
I don't know how that's holding up, now that we're back to a large percentage of developers who are using Windows for development, and all that brings in. Which relieves some of the commercial motivation to honestly support open source, as well as eroding technical savvy about what's secure/sustainable/etc.
(I'm guessing most developers don't understand why there was commercial embrace of open systems, and then of open source. It's partly cost, but also outright abuse and counterproductive dynamics. In some sense, we're coasting, reaping many of the benefits of past battles that got out of abusive situations, while setting up the next generation for abuse. Only, the next generation might have it worse: tech will be vastly more ubiquitous, complex, and mandated -- and perhaps impossible to dig themselves out of.)
Had it not died I would still have used my 2013 MacBook Pro. For many use cases computers stopped being slow a decade ago.
There are certainly things I can do on my new laptop that was a major hassle on the old one, but web browsing, Python development and day to day sys admin stuff was perfectly fine on the old machine.
For me it's all about the screen, an 11 year old ThinkPad most certainly have a terrible screen (it might not, but most do). Getting a clear hi-dpi monitor is more important than having the latest CPU, GPU or 32GB of RAM, at least for my needs.
I’ve been buying MacBook Airs of the 2011-2013 vintage for 10 years. Love them.
I have to disable third party JavaScript, and I have to be careful what software I install, but I love this machine.
I will probably upgrade to an M1/M2 for my next machine, but it’s because of software not hardware. The software, after 10 years, is finally starting to be bloated enough that I feel like I might need more soon.
Installed latest Mint MATE (based on Ubuntu 22.04) on a MacBook Air 11 2015. Linux has a lot of rough edges on the MacBook Air, definitely not boring, it but works well enough for my needs:
* No fan control out of the box, so CPU overheats after a new minutes. Fixed by installing a 3rd party fan control package.
* Broken sleep. Always wakes up 2-3 seconds after putting to sleep. Fixed by a series of hacks to disable the keyboard and lid while sleeping. Only the Power button is able to wake it up now.
* Display brightness setting lost after sleep. Always wakes up at 100%.
* Webcam does not work. There is no compatible driver from what I understand.
* Two-finger scroll is awful on Linux, compared to the buttery smooth scroll of MacOS.
* Poor battery life compared to MacOS, I estimate about 25% less.
* It can be tricky to figure out how the Mac keys are mapped to normal Linux keys: Alt, Option, Command. Also tricky to figure out how to remap them so that they are more usable on Linux.
Weird, I installed Asahi on my M1 Pro and while the two finger scroll isn't butty smooth as MacOS, it works well enough and already better than in any Thinkpad touchpad.
I bought an M1 Pro a month ago. While I kinda miss linux, I do not miss x86 at all.
Seriously, going back from a performant arm64 to x86 feels like going back to the POWER architecture all over again. Big bloated chips where every little computation generates a lot of heat and you need big fans to dissipate all of it.
I wish I could do the same, although in the lower end, with my Raspberry Pi 4. Sadly, those laptop cases for the RPi are too expensive, if you include shipping, to justify them.
All platforms have issues, especially with uncommon hardware combinations. But if you buy any mainstream device odds of it working in linux are probably similar to the odds of it working in windows.
For older hardware the odds are much better that it will work out of the box in linux.
That forum thread you linked does not say microphone drivers are broken on Windows. The first reply sounds on point.
I'll add that all laptops produce noise on the headphone jack as the audio amplifier is preemptively switched on and off. Only difference is that it's normally just barely audible. I'm gonna take a wild guess and say they just choose a crappy IC.
Granted, I've always had these kinds of issues with new laptops, especially when it came to proprietary nvidia or AMD graphics (before AMDGPU) and I agree it's improved a lot, but I still need to tell people that there's caveats with some (especially newer) laptops.