[SOLVED] Optimus and Nvidia

[Arthanis]

I found this post (https://forums.gentoo.org/viewtopic-t-959568.html) and posted my doubt, but it seems dead, so forgive me if it is a double post:

Well, I just bought a laptop with optimus technology, and I tried bumblebee and didn`t like it. I read around and I understand that it is now possible to use the nvidia gpu without bumblebee even though my laptop lacks a hardware muxer to disable the intel.
So and followed all the instructions here but couldn' t get it to work. I uninstalled xdm and bumblebee to make things simplier.

These are my adapters:

[Yamakuzure]

It is possible to use nvidia *only* with intel disabled. Yes. If you want to switch, you need something that actually can do the switch. Bumblebee + Primusrun is the best I could find so far. (From "Bumblebee" Overlay)

Otherwise I do not understand what you want to achieve.

The xorg.conf you posted is one used by switchers like bumblebee. It does not let you use your nvidia card directly but configures your system to run on the Intel HD. With the configs you posted it will not be possible to use your nvidia card at all.
_________________
Important German:

1. "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
2. "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.

[Arthanis]

Well, I see. I guess I misunterstood this post (https://forums.gentoo.org/viewtopic-t-959568.html) because of this:

[krinn]

[Arthanis]

krinn, I linked a full dmesg in pastebin on the post, would you please check it out and see if it seems indeed an oops?

EDIT: Here is what I think is the relevant part of dmesg regarding nvidia:

[Arthanis]

Yamakuzure, doesnt this arcticle here (http://wiki.gentoo.org/wiki/NVIDIA_Driver_with_Optimus_Laptops) talk about bumblebee-less setup ?

[Yamakuzure]

[cyberjun]

Hi,

What worked for me was this: https://wiki.archlinux.org/index.php/NVIDIA_Optimus#LightDM. With lightdm, I get to see the login screen as the xrandr commands are run before login. I have a Lenovo G500s laptop with optimus. I had almost given up and then I tried this arch wiki.

Works very nicely. HDMI out works too, along with suspend/hibernate etc.

--cyberjun

[Arthanis]

[cyberjun]

Yes. it is bumblebee less. You have to remove bumblebee from your runlevels and also remove/blacklist bbswitch module.

--cyberjun

[Arthanis]

Cyberjun, would you please enlight me? It's all so confusing. How can I use optimus without bbumblebee if I have a muxless laptop? And to do that, do I nees Do I need CONFIG_VGASWITCHERROO ir not? No BBSWITCH?

And also, it all goes back to the beginning of my post. Everytime I try to startx with taht kind of xorg.conf, it gives me the dreaded "no screens found" error. And if am able to do it, does powermizer (nvidia gpu frequency scale) work?

[cyberjun]

Hi,
I don't know what muxless means. On my laptop, I can switch exclusively to Integrated Graphics or Hybrid (in BIOS) but not exclusively to discreete (NVIDIA). Does that mean muxless?
You don't need CONFIG_VGA_SWITCHEROO. In fact disable it if already enabled. Also I don't have bbswitch loaded. bumblebee service is also disabled. If you get this setup right, nvidia-settings shows powermizer working.

My Nvidia card shows up like this in lspci output:

[Yamakuzure]

From the wikipedia page on nvidia optimus: (MUX == "[MU]ltiple[X]er")

"Optimus avoids usage of a hardware multiplexer and prevents glitches associated with changing the display driver from IGP to GPU by transferring the display surface from the GPU frame buffer over the PCI Express bus to the main memory-based framebuffer used by the IGP. The Optimus Copy Engine is a new alternative to traditional DMA transfers between the GPU framebuffer memory and main memory used by the IGP"

In other words: You can not use your nvidia card directly, it is always indirect. Without a multiplexer there is no direct connection between any video out and the card.

And further: "The binary Nvidia driver added partial Optimus support May 3, 2013 in the 319.17.[5] As of May 2013, power management for discrete card is not supported, which means it cannot save battery by turning off Nvidia graphic card completely."

So while it is a nice thing that the nvidia linux drivers natively support optimus in some way, they are out of the window for people like me with an often traveling laptop. I need to be able to have the discrete card switched off completely, and only bumblebe can do this at the moment.

However, what is the problem anyway? Is it too much asked to write "primusrun foo" instead of just "foo"? Actually I find this more convenient, as I have absolute control over when the nvidia card is switched on and when not.
_________________
Important German:

1. "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
2. "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.

[Dr.Willy]

[Yamakuzure]

[Arthanis]

[Yamakuzure]

[Arthanis]

Well, I was finally able to use nvidia-only with proprietary drivers. But before I mark this post as SOLVED, I will try to give it back to the community by summarizing all the experience and knowledge I got about this subject, because its not clear what all the options for optimus on linux are. Please if anyone see any error feel free to correct any misconception I may have had.

As far as I'm concerned, these are the options when it comes to optimus on Linux:

Option A: If you have a muxed hardware, you can disable the integrated GPU, use only nvidia either with nvidia-drivers or noveau, and be happy

If you have muxless (or willing to use the integrated chip), the options are:

Option B: Use nvidia proprietary drivers support for optimus, using nvidia chip all the time

If you choose to use bumblebee (with iether noveau or nvidia), there are two backend that transport the data from the framebuffer:
Option C: Bumblebee + Virtual GL ("network" backend, slower)
Option D: Bumblebee + primus (local backend, faster)

Options A and B:
-PROS: Less hassle, better performance
-CONS: More Heat, more power consumption

Options C and D
-PROS: Flexibility, better power management
-CONS: More hassle, more daemons, more bugs, less stability, no HDMI sound, compulsory use of DM (i like using auto startx after shell login, without any DM)

I finally got it working with option B, with the following steps:

- Enable intel modesetting, intel i915 in kernel, rcu_igp_delay, agpgart, drm_i915_kms, agp_intel
- Enable the following options in /etc/default/grub:

[Moriah]

I recently upgraded from a Lenovo w500 to a w530, mainly to get 8 cores and 32 GB of ram. I always used the w500 iin intel mode and used the vga output to an external monitor when at my desk. When on battery, I had the same 1920x1200 resolution on the laptop screen as I had on the external vga. The 2 monitors had exactly the same picture.

The w530 is a downgrade as far as the built-in screen is concerned. The laptop screen is 1920x1080. But the graphics controller is either intel, or nvidia K100M, which has very high resolution capabilities via the mini-displayport. The vga connector on the w530 gives me problems. I tried to connect it to the same external monitor I used with my w500, and it failed to read the EDID info, so the resolution was early 1990's style 800x600.

So I decided to bite the bullet and got a Dell P2815Q monitor that does 3840 x 2160.

It works great under w7, which I only installed to prove that the hardware would work.

It must use the displayport input to the monitor, which means it must use the mini-displayport output from the w530. But the mini-displayport output only works with the nvidia controller. So I must get the nvidia controller working under gentoo linux.

I boot my laptop to a command prompt login, not an X-server, so all that flap about blank screen before login will not affect me. I do, however, switch between the tty consoles with alt-ctl-F1 thru F6, and between multiple X-servers on alt-ctl-F7 thru F11. F12 displays my syslog.

I run a minimal window manager: ctwm, which is basicly a very simple twm with multiple virtual desktops.

Since it will be impossible to keep the same picture on the laptop screen and the Dell monitor (given I run the Dell in it 3840x2160 mode), I am willing to give up the laptop display entirely when using the external monitor. I usually close the lid when using the external monitor on the w500 anyway.

I am currently running:

[Moriah]

I have installed the nvidia driver and have at least that much working, as when I boot the new kernel, I can see:

[Arthanis]

Moriah, O don't know how to solve your problem, but I do know that the module nvidiafb you have loaded, conflicts with nvidia-drivers, so for starters, I would remove or blacklist it.

Anyway, Im marking this thread as solved. The results for power management are:

intel only: about 4:00 hours of battery power
nvidia : about 3:40 hours of battery power

Obviously, I did not use any gpu intense software, just normal browsing and stuff. Thank you all for the help.

[Yamakuzure]

Well it surely depends on usage. I can only say that on my laptop 'powertop' reports a power consumption between 15 and 18 Watts on intel, and 48 to 52 Watts on nvidia. But then, I am using kwin with opengl backend, so I'd have more usage of the gpu anyway.
_________________
Important German:

1. "Aha" - German reaction to pretend that you are really interested while giving no f*ck.
2. "Tja" - German reaction to the apocalypse, nuclear war, an alien invasion or no bread in the house.

[Arthanis]

I couldn't agree more. Im using xfce with very little compositing. No compiz nor nothing. Would you please tell me which command line you used for powertop? Mine has the following output:

[Yamakuzure]

No special command line options. just 'powertop'.

powertop only displays the important values when you are on battery. Otherwise it is quite limited. Here is a comparison:

powertop on AC:

[Maffblaster]