Is there a path forward for better support of newer hardware on desktop Linux?
-
I've been in a fortunate position this past year of having some extra money to throw at shiny new hardware and I've experienced a side of Linux I haven't dealt with before...its poor support for shiny new hardware.
I grabbed a Ryzen 9000 CPU and an X870 motherboard...only to find that ethernet didn't work on kernel 6.11. I had to use a usb-c to ethernet dongle for several weeks until 6.13 released.
Just today and what prompted this post, I splurged on a 4k 240hz HDR monitor. HDR is obviously in-progress and I did not expect it to work out of the box. Critically, what I did expect was for the 240hz part to work, but I couldn't set it to anything beyond 120. Skip forward a couple hours, and I now know what EDID files are and how to use different ones. For more insight on my night, see this issue, this blog post, and this blog post. After all that, 240hz is smooth, goddamn.
For me, I'm not complaining. I love desktop Linux far more than shiny new hardware. I would return this monitor before considering not using Linux, and in the latter case it was a good chance to learn more about how Linux deals with display devices.
But I'm also one of many people here who wants to see desktop Linux become more popular, and if a regular person encountered either of those issues, they're going straight back to Windows. While that monitor issue has been fixed upstream, it's still broken in an up-to-date distro like Fedora and the monitor is over 6 months old at this point.
When it comes to stuff like HDR, that's obviously progressing quickly and is likely to become a non-factor in the future. But new ethernet controllers and new monitors with invalid DisplayIDs are likely always going to be coming out. Unless you're willing to tinker, your only option is to wait weeks or months before buying the new shiny thing if you want to use Linux.
That brings me to my question, is there a future where this isn't the case? And what would be required to get there?
Do motherboard/monitor/IC/etc manufactures need to submit their own kernel patches well in advance of product releases, like what AMD and Intel do for their CPUs and GPUs? Are we just waiting for them to give a shit?
Is there any possibility of hardware support-related patches getting backported to older kernel versions sooner rather than waiting for new major releases?
This is kind of an ungooglable question, and I figured it might make for an interesting discussion topic if anyone has more insight or thoughts on this.
-
L [email protected] shared this topic
-
[email protected]replied to [email protected] last edited by
I can't answer the main question, but anecdotally I find different distros work differently on different hardware. Like one laptop I have will not install, or if it does install it won't boot, any Debian based distro, but RPM based is fine. There is some bios bug or quirk that fails with deb, but rpm distros acknowledge and move on. So sometimes distro hoping can help.
-
[email protected]replied to [email protected] last edited by
the last time i was in your shoes, i bought from linux only shops. ie system76, tuxedo, kfocus, juno, novacustom, etc. and i had the exact opposite experience; it was as smooth as that of a mac owner, but on linux.
linux's forte is stability, that's why it's so dominant in the server space. adding in support for end user hardware is an accident and requires extra effort on hardware makers' part who don't always rise to the challenge when they don't believe it's profitable enough for the effort; in which case, volunteers have to step in to fill the gap.
-
[email protected]replied to [email protected] last edited by
"Distro hoping" Nice Freudian typo.
-
[email protected]replied to [email protected] last edited by
When manfuacturers are making their new hardware they do the work of making sure whatever OS they want on it will work flawlessly. Most machines are not made with Linux in mind but the community comes in once released and does the work to get it working. If manufacturers did that work prior to release it would be fine. Companies do exist that specialize in Linux machines that do work like this.
One of the main reasons it takes so long for new stuff to get support is the linux devs doing that work themselves need access to the hardware to make it work. They arent made of money tho, and cant go buy every new piece of hardware as soon as it releases.
Its a matter of manufacturer support vs community support. Typically community support takes longer to get rolling but also lasts much longer. A combination of the 2 would be ideal, but manufacturers would have to cooperate.
-
[email protected]replied to [email protected] last edited by
For sure, if I was in the market for a laptop, System76, Tuxedo, and (while not exclusively Linux) Framework would be at the top of my list
For general PC hardware though, I've always been late to the party. I upgraded to Ryzen 3000 right before 5000 was coming out, so hardware support was already perfect on Linux. That's basically been my upgrade strategy for the past 10 years, so I've personally never really encountered these teething problems before now.
adding in support for end user hardware is an accident and requires extra effort on hardware makers' part who don't always rise to the challenge when they don't believe it's profitable enough for the effort; in which case, volunteers have to step in to fill the gap.
That's really the crux of the problem. How can we make companies care and/or better support volunteers to get patches out sooner.
-
[email protected]replied to [email protected] last edited by
Bleeding edge is a bitch, ain't it. Thanks arch!
-
[email protected]replied to [email protected] last edited by
Believe me, it used to be so much worse than that.
Hardware vendors see the need to allocate their resources to support the majority of the users, so that means making drivers for all current flavors of Windows and Mac. Linux has a residual market margin, so no incentive there.
It usually is up to some talented person or persons somewhere out there to come up with support for dinner shiny new hardware, usually months or years after the shininess went away.
The path is clear: buy from vendors who support Linux, make yourself heard if they don't, or put up the work to make it work if you have the capability.
-
[email protected]replied to [email protected] last edited by
Honestly if you want the best chance of brand new hardware working, a rolling release distro running the newest release kernel as soon as possible is pretty much your best bet.
-
[email protected]replied to [email protected] last edited by
What you're asking is very difficult due to resources, creating new bugs etc. The reality is, go with true and trusted hardware. Linux can not possibly support every new shiny thing. That's Windows' job, because that's where the drivers are made for. And since it's difficult to install Linux drivers manually for most users, manufacturers don't bother with Linux at all. Especially, since Linus doesn't care about compatibility with older kernels as much as Windows does. Either support is in the kernel, or you'll experience problems.
Personally, all my laptops and PCs are more than 4 years old, for that reason. I often buy refurbished too. I write this in a refurbished Macbook Air from 2015, where I nuked MacOSX to run Linux Mint (with a binary broadcom wifi binary which is thankfully well supported by ubuntu). On my main PC, the only new thing I bought was an Intel gfx A- card, a 2 years old card, but I knew it had support before I bought it.
And even then, if it's some weird thing, e.g. some over-complicated sound capture device, gaming mouse with a thousand buttons, etc etc (in other words, non-standard hardware), don't expect great support for it, even if the years are passed. Stay with vanilla hardware to be compatible. If you had bought a 60 Hz 4k monitor, you wouldn't have had problems and you would have saved money. Personally, the only feature I look in monitors these days (apart from good color and enough ports), is that they are 32" instead of the usual 27", so that I don't have to use scaling (which creates yet another problem with Linux). I use 4k in its 100% resolution, fitting lots of windows in it, and not destroying my eyes because they're well visible.
-
[email protected]replied to [email protected] last edited by
is there a future where this isn’t the case? And what would be required to get there?
Hardware manufacturers of companies in a position of power who leverage it. Like Valve.
Are we just waiting for them to give a shit?
Speak with your money. They won't care though, the linux market is minimal.
-
[email protected]replied to [email protected] last edited by
the only feature I look in monitors these days
Not being smart.
-
[email protected]replied to [email protected] last edited by
When i see a monitor having network hardware built in, on the spec sheet, i get very suspicious.
-
[email protected]replied to [email protected] last edited by
I don't game, so it's smart. As long as I have the color I want (I used to shoot music videos for bands, so I'm into filmmaking), and enough ports, the only other thing that I need is to not scale. I don't need too much refresh rate, I don't need hdr, I don't need ethernet, I don't need webcams etc etc.
-
[email protected]replied to [email protected] last edited by
Speak with your money.
This is tough to do when there's often not a functional alternative. Buying old motherboard stock and a non-HDR monitor isn't going to teach manufactures people want Linux support, they'll just think you are saving money / don't care about new features.
Without some sort of clear 'Linux Certified' system they can compare sales against, no hardware manufacturer is going to be able to recognize the 0.1% sales increase stemming from the time random internet volunteers fixed the open source implementation of their display drivers.
-
[email protected]replied to [email protected] last edited by
Speak with your money.
This is tough to do when there's often not a functional alternative. Buying old motherboard stock and a non-HDR monitor isn't going to teach manufactures people want Linux support, they'll just think you are saving money / don't care about new features.
Without some sort of clear 'Linux Certified' system they can compare sales against, no hardware manufacturer is going to be able to recognize the 0.1% sales increase stemming from the time random internet volunteers fixed the open source implementation of their display drivers.
-
[email protected]replied to [email protected] last edited by
In the last decade, I've had that sort of issue affect me twice:
- I bought an AMD Vega 56 on launch day, and I had to run it with the proprietary driver for a while.
- I recently upgraded my three monitors, and was having trouble getting them all to do the 1440p/100Hz they were rated for. After a bunch of fiddling with xrandr etc. and trying to add modelines and whatnot, it turned out the real problem was that I needed to upgrade from HDMI cables to DisplayPort ones.
Anyway, I guess the gist is that I wouldn't have expected Windows to do any better in either case.
-
[email protected]replied to [email protected] last edited by
What do you do with 240hz on linux? I understand people who game on windows and need 240hz, is it even noticeable beyond 100hz when not gaming? Please elaborate
-
[email protected]replied to [email protected] last edited by
I understand people who game on windows and need 240hz
I don't. Isn't this just a numbers game? Should be unnoticable for sensitive people above 100hz.
-
[email protected]replied to [email protected] last edited by
That brings me to my question, is there a future where this isn't the case? And what would be required to get there?
Some kind of service where vendors can shove their propietary hacks of drivers in your face (as DKMS), like on Windows.
But nobody wants that.