Author Topic: Dual GPU laptops and LMCE  (Read 5347 times)

l3mce

  • NEEDS to work for LinuxMCE
  • ***
  • Posts: 1084
    • View Profile
Dual GPU laptops and LMCE
« on: April 10, 2012, 07:56:18 pm »
So I have written some code to handle disabling the discrete GPU and running the powerful GPU on laptops supplied with two for battery saving. Unfortunately I have no such hardware, so I have just sort of guessed at it. I would like someone with a laptop that has two GPU's to test this code. Chances are my first guess won't just work OOTB, so it would be helpful if someone would drop in IRC at their convenience to try and get this going. I am particularly interested in anyone with a weak and a powerful Intel based GPU pair. Typically there is a low power intel and a higher power nVidia or ATI, which is frankly a lot easier to deal with. Having the code figure out WHICH Intel to pick is a bit dicey, esp without any actual hardware to test on.

Thanks in advance to anyone who will test.
I never quit... I just ping out.

JaseP

  • Addicted
  • *
  • Posts: 526
    • View Profile
    • JaseP's LinuxMCE Wiki User page
Re: Dual GPU laptops and LMCE
« Reply #1 on: April 11, 2012, 08:23:40 pm »
Two questions, ... Does an Atom D510 count? ... and ... Have you considered Jupiter for power savings?

Link here describing Jupiter,...
http://www.webupd8.org/2010/07/jupiter-ubuntu-ppa-hardware-and-power.html

You might want to take a look as to maybe merging your code and Fewt's code...
See my User page on the LinuxMCE Wiki for a description of my system configuration (click the little globe under my profile pic).

l3mce

  • NEEDS to work for LinuxMCE
  • ***
  • Posts: 1084
    • View Profile
Re: Dual GPU laptops and LMCE
« Reply #2 on: April 12, 2012, 12:17:24 am »
Power saving is of no interest to me for LCME. I am not creating two setups and dealing with switching etc. I am just trying to prevent it from puking at two cards, and run the best card for video playback.

And all that counts/matters is when you type lspci | grep VGA it returns two results.
I never quit... I just ping out.

JaseP

  • Addicted
  • *
  • Posts: 526
    • View Profile
    • JaseP's LinuxMCE Wiki User page
Re: Dual GPU laptops and LMCE
« Reply #3 on: April 12, 2012, 04:25:27 am »
Sorry, misunderstood/misread what you were getting at...

You might want some people to check some small footprint machines too. Some of them were created for laptops originally, but were adapted for desktop use.

Also, would any machine with two active display adapters suit your needs?? I am going to get 2 machines for home lab use that will likely have both an onboard display and a low-to-medium end, low profile nVidia card. I can try to keep them both active in the BIOS.

Is there any hope for using this code you're working on as a dual head setup with the better being grabbed for video/TV playback and the lesser of the two driving a separate on-screen Orbiter?? The second as, say, a small touch panel ...
See my User page on the LinuxMCE Wiki for a description of my system configuration (click the little globe under my profile pic).

l3mce

  • NEEDS to work for LinuxMCE
  • ***
  • Posts: 1084
    • View Profile
Re: Dual GPU laptops and LMCE
« Reply #4 on: April 12, 2012, 06:48:38 pm »
I can do that... I am looking specifically for this hardware setup. It is fairly common now days, and I used to have scads of laptops floating through, but due to my hatred of laptop repair have really shied away from them.

I am attempting to disable a piece of hardware and blacklist it... which would sort of be the opposite of going for dual orbiter support. I have flirted with the idea before, but cannot come up with a scenario that makes that at all advantageous, when you can get any number of remotes/tablets that are much more convenient and less distracting from the media being played. It would be a WHOLE lot of work, wrought with problems and hardware/driver conflicts for a scenario I cannot see any advantage to.
I never quit... I just ping out.

JaseP

  • Addicted
  • *
  • Posts: 526
    • View Profile
    • JaseP's LinuxMCE Wiki User page
Re: Dual GPU laptops and LMCE
« Reply #5 on: April 18, 2012, 05:40:26 am »
Found this link...

https://launchpad.net/~hybrid-graphics-linux

Might be useful to you...
See my User page on the LinuxMCE Wiki for a description of my system configuration (click the little globe under my profile pic).

jaypetey

  • Guest
Re: Dual GPU laptops and LMCE
« Reply #6 on: April 22, 2012, 08:23:40 pm »
As an Ubuntu user of one of these laptops (nVidia Optimus: Intel + nVidia), I can tell you right now that the whole switchable graphics thing is a nightmare. It's getting better, slowly, but you run across all sorts of problems. I suggest looking into the Bumblebee Project (apparently I'm too new to post links, but Google "bumblebee nvidia" and you should be able to find something), at least for the nVidia side of things.

Basically, though, my understanding of Optimus is that, even though it has a dedicated GPU, everything still gets sent through the Intel GPU before being displayed. Therefore, you cannot turn off the Intel GPU, but you can still leverage the nVidia card. Nonetheless, it seems what you're trying to accomplish can already be done through the use of Bumblebee's optirun command. The basic state of things is that everything goes through the Intel GPU, but when you precede a command with "optirun" it will turn on your dedicated card and use it instead.

Frankly, though, it's all just a big pain. While using Ubuntu I very, very rarely touch my nVidia card. The Intel GPU built into the Core i3's seems to be plenty to handle any tasks I've thrown at it. Basic day to day use sort of things.

l3mce

  • NEEDS to work for LinuxMCE
  • ***
  • Posts: 1084
    • View Profile
Re: Dual GPU laptops and LMCE
« Reply #7 on: April 23, 2012, 03:31:19 am »
As an Ubuntu user of one of these laptops (nVidia Optimus: Intel + nVidia), I can tell you right now that the whole switchable graphics thing is a nightmare. It's getting better, slowly, but you run across all sorts of problems. I suggest looking into the Bumblebee Project (apparently I'm too new to post links, but Google "bumblebee nvidia" and you should be able to find something), at least for the nVidia side of things.

Basically, though, my understanding of Optimus is that, even though it has a dedicated GPU, everything still gets sent through the Intel GPU before being displayed. Therefore, you cannot turn off the Intel GPU, but you can still leverage the nVidia card. Nonetheless, it seems what you're trying to accomplish can already be done through the use of Bumblebee's optirun command. The basic state of things is that everything goes through the Intel GPU, but when you precede a command with "optirun" it will turn on your dedicated card and use it instead.

Frankly, though, it's all just a big pain. While using Ubuntu I very, very rarely touch my nVidia card. The Intel GPU built into the Core i3's seems to be plenty to handle any tasks I've thrown at it. Basic day to day use sort of things.

I am familiar with bumblebee, and vgaswitcheroo... for whatever reason bumblebees bash version has been wiped from the map and the devs wont give me any idea where I might find it.

What you say, while disheartening, makes sense about the optimus, which is why you can't disable the thing in the bios to begin with... however I have changed my methodology (even though I only had ONE user test for a few minutes for me :P) has changed.

I am no longer disabling the hardware, I am installing the appropriate driver and pointing to the bus id of the appropriate card in xorg. I have it all written out, I just need to figure out the right place to introduce it in our very chicken and egg unique xorg.conf creation.

I believe I have it worked out in the right places. Just need testers.
I never quit... I just ping out.