My 2018 Linux Test

So it's all installed. Still working on a few things. Took a bit though because I had a very odd issue. Last night I had key errors while installing. Couldn't figure it out until this morning I just happened to see that my clock was saying 2119 instead of 2019. No idea how or why but corrected that (set-ntp true wasn't fixing it) and that was all I needed.

PRIME doesn't work right with the NV proprietary driver. So I said fuck it and am just straight running the NV card. So now it's just tweaking the config and scaling things as I use it to make it easier on my eyeballs. 4k-arch.png
 
So it's all installed. Still working on a few things. Took a bit though because I had a very odd issue. Last night I had key errors while installing. Couldn't figure it out until this morning I just happened to see that my clock was saying 2119 instead of 2019. No idea how or why but corrected that (set-ntp true wasn't fixing it) and that was all I needed.

PRIME doesn't work right with the NV proprietary driver. So I said fuck it and am just straight running the NV card. So now it's just tweaking the config and scaling things as I use it to make it easier on my eyeballs. View attachment 152370

PRIME does work, I just don't think the instructions on the Archwiki work. Not that I ever got it working either. But I do plan on giving it a try again here soon.
 
PRIME does work, I just don't think the instructions on the Archwiki work. Not that I ever got it working either. But I do plan on giving it a try again here soon.

I may try again as I'm going to rebuild tonight and go back to xfce. 4k on 15 inch screen is overkill and when you add that into my dual 1080p screens for work it just blows it all up.

Scaling in kde works great but it only scales the whole system so everything on the 1080p screens is fucking huge.

So 1080p it is.
 
Yeah, I can't remember why I went with 4k on the 9560. I knew in many ways I would regret it and I have, but it may have been the only way to get touchscreen at the time. I can't remember.

But, with the 1080p monitor connected it is fine.

When I disconnect the external monitor there is definitely some fine tuning I need to do. Just haven't got around to it yet. I've been fiddling around with Mozilla's HTML/CSS/JAVA web development stuff the past few days and having a good time learning that.

Who knows, I may just have to go back to school and finish that computer science degree I started 25 years ago. :p
 
Yeah, I can't remember why I went with 4k on the 9560. I knew in many ways I would regret it and I have, but it may have been the only way to get touchscreen at the time. I can't remember.

I can hint: Dell won't ship the more serviceable configs (i.e. 16GB RAM and 512GB SSD) without 4k. It's infuriating.
 
I had 200GB of space left on my hard drive so I installed Arch again on the laptop.

I'm going to preface what I'm about to say with this statement: The Arch wiki for setting up Optimus is absolutely freak'n useless.

So the Arch install goes as planned and I've logged in as root.

Rather than use the nVidia driver I decide to just leave the nouveau driver and go with that.

I download X, then attempt to configure it; crash and machine powers off.

Restart the machine, use lspci to verify what driver is being used; crash and machine powers off.

I forgot you can't use nouveau on this 9560 because everything including X, lspci, anything that does a call to the video hardware causes an immediate system crash.

So, I install the nvidia driver and xrandr per the Arch Optimus wiki, reboot, and lspci and X now work.

I run Xorg :0 -configure and generate the temporary xorg.conf.new file in /root so I copy it to the /etc/X11 folder as xorg.conf.

Of course now every time I attempt to run the X server test, it crashes with an error. Although when you review the X error log there are no errors.

So I'm going to do my next post in a complete step by step taken so I can try and work through in my mind what the heck is going on.
 
First, some more complaints about the Arch Wiki PRIME article.

One of the first thing the wiki tells you is there are multiple options for dealing with PRIME. Unfortunately, this one doesn't seem even remotely accurate: "using the official Optimus support included with the proprietary NVIDIA driver, which offers the best NVIDIA performance but does not allow GPU switching and can be more buggy than the open-source driver"

This is either a lie, or someone doesn't know what they're talking about. The nVidia driver supports this just fine in my Kubuntu install and I'm able to freely switch between the Intel and nVidia driver through nvidia-settings.

The first thing it tells you to do is install the nVidia driver and xorg-xrandr. Again, this is a HUGE mistake because it assumes you've installed X in the first place. X does not come pre-installed with the base package, even though in multiple places in the Arch Wiki it says it does. And even after install, X needs to be configured; which you can't do UNTIL you install the nVidia drivers because the system crashes if you try.

And don't even get me started with the Arch Wiki nVidia driver install page.

The configuration section opens with this statement: "The proprietary NVIDIA graphics card driver does not need any Xorg server configuration file." This too, is absolutely not true. Without the default xorg.conf file, nvidia-xconfig generates an incomplete configuration file that doesn't work.

Sorry, had to get that out of my system.
 
It looks like Ubuntu created a package called nvidia-prime which is what is allowing the switch between Intel and nVidia cards in nvidia-settings.

So, on that count, the Arch wiki isn't wrong, it just doesn't have that same capability.

#%$#$%^^
 
So got the new 5510 all up and running. Love it. The one thing I've learned with this though is that 4K on a laptop should be banned from existence. Period.

It's a gorgeous screen but having to effectively zoom in on everything with scaling kind of defeats the purpose of the hidpi. The fact that scaling isn't monitor independent sucks too. Basically I'm running it at 1080p because my two work monitors (so I have 3 running) are 1080p. If i had the 4K screen scaled up then the 1080p screens where GIGANTIC. ; ;

But I'm still tweaking the theme a bit deciding on what I finally want to settle on but overall it's done. XFCE with SDDM and Nvidia drivers. I chose SDDM over Lightdm because in my experience SDDM handles the xrandr stuff for the Nvidia card better than Lightdm does. The last major piece is I want to enable AppArmor for Snap confinement but the Arch wiki is kinda vague so I asked a few questions before making the config changes to enable AppArmor.

Screenshot_2019-04-03_22-38-46.png
 
So got the new 5510 all up and running. Love it. The one thing I've learned with this though is that 4K on a laptop should be banned from existence. Period.

It's a gorgeous screen but having to effectively zoom in on everything with scaling kind of defeats the purpose of the hidpi. The fact that scaling isn't monitor independent sucks too. Basically I'm running it at 1080p because my two work monitors (so I have 3 running) are 1080p. If i had the 4K screen scaled up then the 1080p screens where GIGANTIC. ; ;

But I'm still tweaking the theme a bit deciding on what I finally want to settle on but overall it's done. XFCE with SDDM and Nvidia drivers. I chose SDDM over Lightdm because in my experience SDDM handles the xrandr stuff for the Nvidia card better than Lightdm does. The last major piece is I want to enable AppArmor for Snap confinement but the Arch wiki is kinda vague so I asked a few questions before making the config changes to enable AppArmor.

View attachment 152644
OSX can handle this so it is doable in linux also. Someone just needs to implement it - I guess not so many people have this problem currently.
 
But nVidia in Linux even with a single dedicated card is a complete pain in the butt. If you're telling me you've never had trouble running nVidia under Linux you've either been extremely lucky, or delusional.

I've never had an issue running Nvidia hardware/drivers under Linux, and there's no delusion involved as we're talking about 30 different machines.

The thing is, I never install Nvidia drivers using the ancient .run method of installation as it's a useless method of installing drivers that results in more problems than it's worth. I make sure that I use a distro that supports a better package managed method of installation like the PPA method under Ubuntu. I even update kernels without issue provided I don't run bleeding edge, DKMS takes care of everything.
 
I've never had an issue running Nvidia hardware/drivers under Linux, and there's no delusion involved as we're talking about 30 different machines.

The thing is, I never install Nvidia drivers using the ancient .run method of installation as it's a useless method of installing drivers that results in more problems than it's worth. I make sure that I use a distro that supports a better package managed method of installation like the PPA method under Ubuntu. I even update kernels without issue provided I don't run bleeding edge, DKMS takes care of everything.

The ".run" method is really the only method. Your just running distros that pre build versions of the kernel with the closed driver already bolted on. And sure a few distros also use NVs scripts to build DKMS scripts. Still they are bolted on no matter who is doing the bolting the distro maintainer or the end user. (of course its much easier for everyone if its the maintainer)

I believe you that you have had no issues, I do. It is however also possible that you simply don't consider forcing full pipeline and issue.... or again the distro your using is doing some of that stuff behind the scenes. Which is cool too I guess... but no matter what most compositors need fixes or end user intervention to have a proper tear free NV experience. We can blame the open source compositor projects for following standards, or we can blame NV for not following them. :)
 
I've never had an issue running Nvidia hardware/drivers under Linux, and there's no delusion involved as we're talking about 30 different machines.

The thing is, I never install Nvidia drivers using the ancient .run method of installation as it's a useless method of installing drivers that results in more problems than it's worth. I make sure that I use a distro that supports a better package managed method of installation like the PPA method under Ubuntu. I even update kernels without issue provided I don't run bleeding edge, DKMS takes care of everything.

Sorry
I've never had an issue running Nvidia hardware/drivers under Linux, and there's no delusion involved as we're talking about 30 different machines.

The thing is, I never install Nvidia drivers using the ancient .run method of installation as it's a useless method of installing drivers that results in more problems than it's worth. I make sure that I use a distro that supports a better package managed method of installation like the PPA method under Ubuntu. I even update kernels without issue provided I don't run bleeding edge, DKMS takes care of everything.

Sorry, I didn't really mean delusional. I must have been in a bad mood. I apologize for saying that.

I just find it amazing you've had no issues, and I've had issues on every machine for 20 years.
 
  • Like
Reactions: ChadD
like this
I can't make any progress right now on the laptop because it is dead. The power port broke and I'm waiting for a part to come in tomorrow so I can fix it. $8 part takes out a $3000 laptop...filmed at 11.
 
I've never had an issue running Nvidia hardware/drivers under Linux, and there's no delusion involved as we're talking about 30 different machines.

The thing is, I never install Nvidia drivers using the ancient .run method of installation as it's a useless method of installing drivers that results in more problems than it's worth. I make sure that I use a distro that supports a better package managed method of installation like the PPA method under Ubuntu. I even update kernels without issue provided I don't run bleeding edge, DKMS takes care of everything.

I have used the built in package manager on every distro I've used that would at least allow me to boot.

pacman -S nvidia for Arch
Kubuntu Prime just worked out of the box so I didn't have to install anything.
Ubuntu I installed through the software center.
Fedora wouldn't boot on either machine.
 
The thing is, I never install Nvidia drivers using the ancient .run method of installation as it's a useless method of installing drivers that results in more problems than it's worth.

My fileserver has Plex on it, with CentOS 7, and it's the .run method or nothing. CentOS did work out of the box with the 1050Ti though, so hard to complain on that front. I just wanted Plex to have access to it for transcoding and whatever else (perhaps providing it to a VM or container).
 
I just find it amazing you've had no issues, and I've had issues on every machine for 20 years.

I can't say that I've used Linux on desktops with Nvidia GPUs much- usually just integrated- but I can also say that I haven't had much trouble at least getting a single window desktop up.
 
The ".run" method is really the only method. Your just running distros that pre build versions of the kernel with the closed driver already bolted on. And sure a few distros also use NVs scripts to build DKMS scripts. Still they are bolted on no matter who is doing the bolting the distro maintainer or the end user. (of course its much easier for everyone if its the maintainer)

I believe you that you have had no issues, I do. It is however also possible that you simply don't consider forcing full pipeline and issue.... or again the distro your using is doing some of that stuff behind the scenes. Which is cool too I guess... but no matter what most compositors need fixes or end user intervention to have a proper tear free NV experience. We can blame the open source compositor projects for following standards, or we can blame NV for not following them. :)

No doubt, that's one of the issues surrounding a tacked on binary blob. But the fact remains that developers have tackled and resolved the issue and now installing Nvidia drivers is quicker and easier with less issues than under Windows.

At the end of the day it's not like Nvidia could really have tackled the issue anyway, it was always a problem the devs of the particular distro's had to resolve (Nvidia could contribute more to the open source drivers, but that's never going to happen). All things considered, philosophies aside, I find Nvidia's support of Linux fantastic. The only time I'll consider force composition pipeline an issue is when it actually results in issues for me, as of yet I've had no issues as a result of enabling it.
 
So got the new 5510 all up and running. Love it. The one thing I've learned with this though is that 4K on a laptop should be banned from existence. Period.

It's a gorgeous screen but having to effectively zoom in on everything with scaling kind of defeats the purpose of the hidpi. The fact that scaling isn't monitor independent sucks too. Basically I'm running it at 1080p because my two work monitors (so I have 3 running) are 1080p. If i had the 4K screen scaled up then the 1080p screens where GIGANTIC. ; ;

But I'm still tweaking the theme a bit deciding on what I finally want to settle on but overall it's done. XFCE with SDDM and Nvidia drivers. I chose SDDM over Lightdm because in my experience SDDM handles the xrandr stuff for the Nvidia card better than Lightdm does. The last major piece is I want to enable AppArmor for Snap confinement but the Arch wiki is kinda vague so I asked a few questions before making the config changes to enable AppArmor.

View attachment 152644

I agree about the 4K on a 15" screen thing. You have to resize everything so much it looks terrible. Not to mention if you are streaming Netflix while out and about, and it defaults to 4K, there goes your mobile data allowance in like 10 minutes. lol

Its also very hard for a graphics card to power a game at 4K.
 
there goes your mobile data allowance in like 10 minutes

I know that Windows has an option to throttle mobile connections... but yeah, I'm not ready to deal with the issues concerning 4k (etc.) on laptops. Wound up with an ASUS instead.
 
I feel like at this point I'm just shooting spitballs at the ceiling trying to get Arch Optimus to work. I've done manually edited so many config files trying to follow the Arch Wiki that I don't even remember what all I've done and what I've probably broken.

Starting over in 3, 2, 1...
 
After a fresh install of Arch the first thing I have to do is install the nVidia driver on my laptop. If I don't, I get these funky error messages that pop up randomly and if I try and run any lspci commands, the system immediately crashes and shuts itself off.
 
After install of the nVidia driver, without running any type of configuration for X (Archwiki does say I shouldn't have to) I installed Xfce and of course, it wouldn't load. So I ran the Xorg configuration and ran the test X -configure /root/Xorg.conf.new and of course, even the test failed.

SHAZBOT!
 
Based on the Archwiki, I verified the following packages were either installed, or I installed them.
  • xorg-server (verified already installed)
  • xorg-apps (some elements installed)
  • xorg (some elements installed)
Next up is to verify my drivers are installed and working:
  • lspci | grep -e VGA -e 3D
I checked first for intel and see the Arch wiki recommends the following be installed:
  • mesa (verified already installed)
  • xf86-video-intel
  • vulkan-intel
I ran Xorg :0 -configure

Ran the X -configure /root/xorg.conf.new test

Black screen. -_-
 
At this point I'm just figuring the default X configuration file cannot handle Optimus, which in my mind makes sense. Two video cards trying to use the same output is probably not conducive to the way X seems to function. Which also explains why there are so many other options out there to try and fix this mess.

So the next option is to mess with PRIME: https://wiki.archlinux.org/index.php/PRIME

So I run the first command per the instructions: xrandr --listproviders

And guess what? It doesn't work. Error can't open display.

So the first instructions are to remove your proprietary video card driver, then run a command that won't work unless you have an xorg.conf file which I can't generate without the nvidia proprietary driver installed because when I do anything without that driver installed the system just reboots itself.

Honestly, this is almost like a joke at this point. Is it still April 1st?
 
At this point I'm just figuring the default X configuration file cannot handle Optimus, which in my mind makes sense. Two video cards trying to use the same output is probably not conducive to the way X seems to function. Which also explains why there are so many other options out there to try and fix this mess.

So the next option is to mess with PRIME: https://wiki.archlinux.org/index.php/PRIME

So I run the first command per the instructions: xrandr --listproviders

And guess what? It doesn't work. Error can't open display.

So the first instructions are to remove your proprietary video card driver, then run a command that won't work unless you have an xorg.conf file which I can't generate without the nvidia proprietary driver installed because when I do anything without that driver installed the system just reboots itself.

Honestly, this is almost like a joke at this point. Is it still April 1st?

You can't just install the system...install Xorg, XFCE, SDDM, enable SDDM, reboot and have it startup? The base system running on Intel should fire right up. then install Nvidia drivers, configure SDDM for the optimus xrandr command and reboot. poof Nvidia drivers and XFCE.
 
You can't just install the system...install Xorg, XFCE, SDDM, enable SDDM, reboot and have it startup? The base system running on Intel should fire right up. then install Nvidia drivers, configure SDDM for the optimus xrandr command and reboot. poof Nvidia drivers and XFCE.
I've tried that a couple of times, most recently yesterday. Well, as close to that as it will allow. After the fresh install and reboot I HAVE to install the nvidia driver immediately otherwise I get these funky error messages filling the screen and the system crashes almost every time I try and do something.

Base install.
Setup wireless network.
Add user and configure password.
Install Sudo and configure for user.
Install nVidia driver. Reboot.
Install Xfce.

And no, it didn't work.
 
You can't just install the system...install Xorg, XFCE, SDDM, enable SDDM, reboot and have it startup? The base system running on Intel should fire right up. then install Nvidia drivers, configure SDDM for the optimus xrandr command and reboot. poof Nvidia drivers and XFCE.
So I just tried installing SDDM, enabled it, started it, and it went to the same black screen that everything else is doing (X, Xfce, etc).
 
I've tried that a couple of times, most recently yesterday. Well, as close to that as it will allow. After the fresh install and reboot I HAVE to install the nvidia driver immediately otherwise I get these funky error messages filling the screen and the system crashes almost every time I try and do something.

Base install.
Setup wireless network.
Add user and configure password.
Install Sudo and configure for user.
Install nVidia driver. Reboot.
Install Xfce.

And no, it didn't work.

That's what I don't understand though...why is it trying to use the NV device at all? Did it actually load a driver? By default it shouldn't. Check to see if Nouveau or something has been installed and then remove it and leave the Intel driver in place. The system shouldn't give two shits about the NV card if the driver isn't there.
 
That's what I don't understand though...why is it trying to use the NV device at all? Did it actually load a driver? By default it shouldn't. Check to see if Nouveau or something has been installed and then remove it and leave the Intel driver in place. The system shouldn't give two shits about the NV card if the driver isn't there.
By default, yes, it is installing nouveau. Once this install is up and running I'll try removing it.
 
Although my guess is without the nVidia driver installed, even with nouveau blacklisted, I'll still get that PCe Bus error message over and over I keep getting if the nVidia driver isn't installed.
 
Although my guess is without the nVidia driver installed, even with nouveau blacklisted, I'll still get that PCe Bus error message over and over I keep getting if the nVidia driver isn't installed.

How are you seeing the error? Is it a popup on screen or just in a log?
 
Since I'm in the CLI doing the install it just puts it onscreen right in the middle of whatever you're doing. So like I can be in the middle of typing a command and it will output 4 or 5 lines of error message.

It looks like they just rolled out Arch 5.0.6 and while my old install USB stick was generating the error message, it doesn't seem to be doing it once the install is up and running.

I guess I'll need to redo my install stick.

Nouveau is definitely loading. I've tried blacklisting it 2 ways but it still loads.

How would you recommend I blacklist the nouveau driver from loading?

I tried using nouveau_blacklist.conf in /etc/modprobe.d and nvidia.conf in /usr/lib/modprobe.d and neither worked to block it.
 
Back
Top