Nvidia Developing Linux Driver for Optimus

By Wolfgang Gruener
published

Nvidia is apparently preparing to introduce Linux support for its Optimus graphics switching technology.

A developer thread dug up by PC World indicates that the GPU maker is still taking Linux seriously, despite harsh criticism expressed by Linux founder Linus Torvalds earlier this year.

In the thread, Nvidia engineer Aaaron Plattner writes:

"So I've been experimenting with support for Dave Airlie's new RandR 1.4 provider object interface, so that Optimus-based laptops can use our driver to drive the discrete GPU and display on the integrated GPU. The good news is that I've got a proof of concept working."

The bad news is that the Linux support is not finished and Plattner has several concerns, which he hopes he will be able to resolve with the help of open source developers. It is unclear how much participation of the open source community is possible and whether Torvalds' talk had any influence on the development on the Optimus driver. There was no information when the driver could become available.

Contact Us for News Tips, Corrections and Feedback

Wolfgang Gruener
7 Comments Comment from the forums
  • mitch074
    The title is misleading: Nvidia is doing NOTHING to port Optimus to Linux - it's the opposite: Xorg developers are working to make it possible for a GPU to offload tasks to another GPU on Linux (and any system that ports the Direct Rendering Manager 2) without going through a proprietary interface: instead, any driver that implements RandR1.4 can be used to offload rendering tasks, making combos such as an AMD card displaying on 6 screens using an Nvidia card to generate the 3D on them a possibility, or an Nvidia Tegra using any integrated chip to render, say, a video, or, as in the present case, an Intel IGP using an Nvidia chip to render the 3D content.
    And, as indicated, since the interface is open and shared across all GPU makers (RandR1.4), stuff like Nvidia disabling PhysX on systems using an AMD card cannot happen... Or at least, can be thwarted.
    Reply
  • shahrooz
    still cant run linux on my PC because of sli :(
    Reply
  • proxy711
    otacon72Optimus graphics switching is the worst "innovation" ever to come out of Nvidia. I bought an HP laptop about a year ago that had this great technology. Too bad it never switched to the discrete GPU ever no matter what you did. Ended up returning it.K so one person had problems, so the whole technology is trash? Right....
    Reply
  • joytech22
    otacon72Optimus graphics switching is the worst "innovation" ever to come out of Nvidia. I bought an HP laptop about a year ago that had this great technology. Too bad it never switched to the discrete GPU ever no matter what you did. Ended up returning it.
    I own two laptops with Optimus and it works fine. Except in one instance where it wouldn't switch to dedicated in that zombie game dead something.. Forcing it for that particular game worked fine though.

    Without Optimus, many hours of my study would have had to be completed in my personal time instead of in-class (as my battery would have been vaporized by that load).

    I would much rather a computer with both an IGP and dedicated GPU over one or the other. It just gives you the choice of having either battery life or high performance computing.
    Reply
  • captainblacko
    sounds like they need the creation and leadership matrix to get this working.
    Reply
  • actually you are wrong, the guy who they are specifically talking about, is a nvidia engineer, doing some things WITH the new implentations with randr by xorgs dave airlie so that optimus can use official nvidia driver to work.

    "So I've been experimenting with support for Dave Airlie's new RandR 1.4 provider
    object interface, so that Optimus-based laptops can use our driver to drive the
    discrete GPU and display on the integrated GPU. The good news is that I've got
    a proof of concept working."
    Reply
  • army_ant7
    I'm not sure what that RanR 1.4 is. But doesn't running a discrete GPU "behind" an integrated one sound like LucidLogix' VirtuMVP?
    Reply