Please add nvidia-418xx driver packages

Only Mobile ones are being dropped M series, the desktop ones should be fine.

Some mobile GPUs, but it still leaves those people with a broken system.

It is one link further on: https://www.nvidia.com/object/linux-amd64-display-archive.html

Nvidia advices the 418 long term driver so the driver is not dropped and will be available for a while.
So it looks like Nvidia has taken the correct step making the 418 long lived driver. now its up-to you lot linux in general to utilise the 418 long lived driver

2 Likes

There's a little discussion here:

https://devtalk.nvidia.com/default/topic/1055820/linux/device-support-quietly-reduced-in-430-release-series/

It may be that there's another legacy support series coming, but that's unsupported supposition.

2 Likes

those are individual releases, not a list of branches.

I perfectly understand and admire the enormous work of keeping up with building all those extramodules for all (new) kernels.

What can I say to convince you that providing a 418xx legacy driver and integrating it into MHWD/MSM is really worth the effort?

(A) Old machines get a strong 2nd wind with Linux. Side remark: Are machines built in say 2011-2013 really old? And those were more often built like a tank than today‘s machines.

(B) Manjaro is applauded for having one of the best - if not THE best - hardware awareness.

(C) Manjaro is applauded for providing a huge variety of Linux kernels.

(D) Manjaro is applauded for its effortless MSM/MHWD set of applications.

I do feel that continuing to fully live up to these features will give Manjaro many pluses with its existing and future user community. You might even advertise the provisioning of a good range of nvidia legacy drivers - i.e. 340xx, 390xx, and 418xx - on your website (didn‘t check if you do so already) such that it gets communicated to reviewers and users.

Hope that helps you to take the burden and go for a 418xx.

—————

Afterthoughts:

Of course, some things need to be confirmed first, a.o.:

Does the current 430 series really fail today on mobile Kepler GPUs? Mine is running fine as of now, cf. the first post in this thread.

I do not object to using nouveau, but I would really like to point out again what I experienced with this driver on Manjaro, cf. my first post here also for this fact. Maybe I can get help in making nouveau run smoothly from you folks here.

And off topic: Agreed, big corporations suck in many respects, from I$ to N$ to M$ ... But there is a place in this world for everyone ...

1 Like

That will now make 4 nvidia drivers required to be built and maintained going forward ... 1 mainline (430 currently) and 3 legacy (340, 390, maybe 418).

Where will it end? What a fricken mess.

Buy AMD people.

Therein lies the problem ... all drivers do not easily integrate and build with all kernels.

Eventually something will have to give, which is not ideal because nouveau is not really a like for like substitute, lacking both functionality and performance.

Oh ... and ... buy AMD people!

2 Likes

Maybe that could be a compromise between user friendliness and maintainer load, legacy drivers for LTS kernels only. In that case, one should weigh in what the latest kernels give to the everyday user (keep side streams like Btrfs in mind).

Dropping legacies altogether would be a HUGE deviation from the Manjaro concepts as we know and love them.

Oh yes.
It's a really bad situation with all those legacy drivers, and only Nvidia is to blame for that.
My next card will be AMD too - the first one since 2004.

I like the idea of LTS + Legacy though.

Depends on the hardware. In most cases the latest kernel doesn't offer all too much for normal users compared to current LTS, except for those with very new hardware and AMD GPUs (and to some extent, also Nouveau).
I've been using LTS kernels exclusively on my desktop since 4.9.

2 Likes

Another approach I can think of is to use dkms for Nvidia modules, so Manjaro team won't be maintaining a bunch of driver versions for every kernel it offers. There are quite a lot of LTS kernels available in repos, and having to build 4 Nvidia modules for each of them looks pretty resource-consuming task.
Of course this means full shift to dkms drivers with regard to Nvidia, otherwise the mess will be inevitable.

Or - I should note that I am saying this without any access to fresh statistics - the team could evaluate the need in 340-series driver and nuke it if it's user base is vanishingly small, so again we will have 3 series of drivers: 390, 418 and current one.

Older (but not too old) nvidias might work sufficiently well with nouveau (I can confirm that for a Fermi GTS 450). The biggest problem is power management and missing advanced features.

A look at the "feature matrix" gives more info:
https://nouveau.freedesktop.org/wiki/FeatureMatrix/

If you look at the list of supported linux drivers 418 got dropped. Also it is part of their deprecation schedule. As soon as X.Org Server 1.21 gets released, most of the legacy drivers are gone anyway. There was already a huge discussion about 340xx over at Arch. Yes, Manjaro is hardware friendly and supports more proprietary drivers, but we won't go thru extra hoops, only when needed.

1 Like

I read that deprecation schedule as indicating that the 390 legacy series remains important until end of 2022:

Support for Fermi-based GPUs
Fermi-based GPUs are no longer supported starting with release 396. These GPUs will continue to be supported through the release 390 legacy branch, which will receive critical bug fixes and updates for compatibility with new Linux Kernes and X server versions through the end of 2022. See http://www.nvidia.com/object/IO_32667.html for a list of Fermi-based GPUs.

Please note the mentioning of new kernels and new X servers. IDK whether such updates work right away, it is nvidia fixing the dead...

This should also be true for Kepler GPUs, since they are one generation younger than Fermi. OTH, that nvidia page doesn‘t mention mobile Keplers. They are specified in other nvidia places. Kind of messy regarding the 418 status...

Currently, I use 390xx on two laptops, one with Kepler (Quadro K1000M), one with Fermi (Quadro 1000M). In addition, I have the current 430.26 on the Kepler machine (multiboot). All doing well.

... nouveau is not like for like ...:

True. On my Kepler machine, I can run Tumbleweed only with nouveau (tried nvidia, but failed). At boot, Linux throws a

Type mismatch between the voltage table type and the GPIO table. Fallback to GPIO mode

error. Google shows this happens for Kepler cards and remains unfixed since more than 5 years. Some people report that their laptop monitor sort of dies over time with this. I don’t understand this matter at all. I haven‘t had anything bad yet, but who knows? [Huuhuu, nouveau ultimately frying a Kepler GPU???]

this makes a lot of sense for desktop users since a cheap AMD card can easily be swapped out but nvidia is focusing on dropping support mostly on the mobile gpu's which means a new laptop and not just a GPU.

people that really like manjaro and dont want to change will then be forced to use nouveau and since it would be the only driver that works with the older gpu's, users will expect support for it from manjaro when things go wrong. if the amount of nvidia related issues from using nvidia drivers seem like a lot, just imagine the amount of nouveau related issues there would be. :scream:

1 Like

Not talking about replacing working laptop systems, but anyone buying a new desktop / laptop for Linux usage make sure to go down the AMD route.

Older nvidia optimus laptop usage will simply become more and more restrictive in Linux, performance wise. These laptops should still function perfectly fine using the integrated intel gpu, but doing anything that requires dedicated gpu grunt probably will not be feasible.

1 Like

Actually, despite nvidia Optimus appearing first in an ASUS notebook in 2010, other manufacturer(s) chose to not implement Optimus in some business-oriented notebooks for a while. The integrated Intel graphics is not accessible then. My HP EliteBooks 8560w and 8570w from 2011 and 2013 are examples. Heavy duty workstations with perfect casing, keyboard etc; main weaknesses no Optimus and thus shorter battery lives. HP stayed behind Lenovo and Dell back then.

Browsing nvidia‘s news releases wrt legacy drivers shows the 390 series to be most important and have the longest future, as it will receive critical fixes as well as updates for new kernels and X servers until end of 2022. The 340 series will be abandoned completely by the end of 2019, EOL. The 418 series for mobile Keplers have been dropped from full support right now, critical fixes (no word about new kernels / X servers?) will end by April 2020. The other Keplers will follow suit...

Cf. these:
Support timeframes for Unix legacy GPU releases
and
Support Plan for Kepler-series GeForce GPUs for notebooks

Doesn‘t that sound like a plan?
(a) Forget about 340xx very soon, poll your users first.
(b) Keep 390xx alive in the current fashion (MHWD / MSM) for LTS and current kernels, which means ”for everything“. NVIDIA living up to their promising bug / new kernel / new X server fixes is the critical prerequisite for this.
(c) Decide on a 418xx as you like, maybe provide it for the latest LTS kernel only, maybe more, maybe ditch it right now. Its lifetime is less than one year from now anyway.

I don't want to be always that guy.. But nvidia situation is not really the same of fglrx, which can be wholeheartedly replaced in 98% of scenarios.

Kepler is the most nicely supported architecture of them all, in nouveau, and *even then* you have to perform reclocking manually and you loose vulkan. I don't even want to think what happens with other generations..

And if this makes the burden on your shoulders unberarable (which I can perfectly understand), maybe it'd be time to consider solidifying once and for all the "keep older closed drivers" system.
Linux 4.4 in particular, is going to be officially supported for so outrageously long that you could not care anymore then for a decade.

Forum kindly sponsored by