First Came VGA Switcheroo, Now Comes ASUS Switcheroo

[phoronix]

Phoronix: First Came VGA Switcheroo, Now Comes ASUS Switcheroo

Last February there was the first bits of hybrid graphics support under Linux when Red Hat's David Airlie began working on what he called "vga_switcheroo." The vga_switcheroo support allowed switching between multiple graphics processors -- primarily for notebooks -- by issuing a few commands and restarting the X Server. This work was later merged into the mainline kernel, but as of late there hasn't been much more to say...

http://www.phoronix.com/vr.php?view=OTM3MQ

[89c51]

is the problem of not being able to switch between cards the same way as win/mac related to X.org or are there other pats of Linux that also need to change??

[FunkyM]

Hello NVIDIA. I would like to buy notebooks. From what I see the search always leads me to OPTIMUS based notebook solutions if long battery life is a factor.

Too bad I can't efficiently use them on the Linux Desktop!

[madjr]

There's also pending upstream patches for Linux 2.6.40 that would make ASUS-Switcheroo not needed

what are these patches, who is the upstream offering them and how are they better?

will this alleviate in some way the problem we have?

[markc]

Working ASUS Models

Right now this ASUS-Switcheroo support is limited to those with the basic WMI MXM interface switching such as the UL30VT notebook.

Is there a list of known models that the ASUS Switcheroo will work with?

The ASUS-Switcheroo module though doesn't support the ASUS Optimus laptops at this point.

So is there any hope that Optimus laptops may be supported sometime soon?

Note to ASUS: you can have my money but not if I can't use Full HD gfx on linux.

[dmbtech]

Now its time to have a switcheroo for apple computers! I would love to be able to easily switch my MBP from ATI(discrete) to Intel(IGP) even with the restart of X.

[elanthis]

is the problem of not being able to switch between cards the same way as win/mac related to X.org or are there other pats of Linux that also need to change??

I'm not sure I entirely understand your question, but there are basically two things that need to happen to do things the "nice way":

(1a) A driver needs to be written that transparently routes graphics commands to the correct card. It's okay if this is just a driver loader that selects either the low-power or high-power GPU when the app starts (this is what Optimus basically does; it requires the driver/OS to know which GPU each app is best suited for, but it works).

OR

(1b) A driver needs to be written that dynamically switches between GPUs for each app (which is significantly harder, but lets the OS make a judgement call as the app runs as to which GPU is needed).

(2) A driver needs to be written that knows how and when to blit the framebuffers on the "disconnected" GPU to the GPU that's connected to the monitor.


The 1b option is significantly more powerful and flexible, but ridiculously more complicated to write. The 1a option is much easier, but requires the rest of the OS stack to get updated to allow setting attributes on the binaries to denote which GPU to use and to allow users to edit the attributes (this may just be a file in /etc that maps binary names to GPU modes, or an extended attributes on the binary file, or even a flag in the .desktop file you use to launch the apps that sets an environment variable, etc.).

[Carmella]

I think this is a great feature! This is the first that I have heard of this technology, and I am really excited about it. This is a great solution to save battery life while you cannot plug in. I think it would be great if it could be set up so that it automatically runs one card while plugged in, and the other when not. Anyone know if this is possible?