jmite Dec 21, 2012 @ 9:14am
How to install AMD Proprietary Drivers with Hybrid laptop graphics?
I'm wondering, does anybody know how to install the proprietary AMD drivers on a laptop with the PowerXpress hybrid graphics? I'd managed to get this working in 12.04, and I haven't been able to get it in 12.10. Installing the drivers through jockey gives me a dead Xserver when I boot, as does directly installing the driver from AMD.

Getting these drivers installed would definitely make games run better.
Showing 1-15 of 15 comments
< >
mab879 Dec 21, 2012 @ 9:17am 
I wound't recommend doing this. This has proven to make the Steam Client not load.
jmite Dec 21, 2012 @ 9:19am 
Oh, good to know! Guess I'm stuck with intel graphics, then?
Last edited by jmite; Dec 21, 2012 @ 9:20am
RussianNeuroMancer Dec 24, 2012 @ 3:56pm 
Originally posted by jmite:
Oh, good to know! Guess I'm stuck with intel graphics, then?
What about "use 12.04" option?
jmite Dec 25, 2012 @ 10:33pm 
Had the same troubles on 12.04, the only difference was I had some luck with the hacks I'd found online, but still not what I'd call a reliable or working solution.
Last edited by jmite; Dec 25, 2012 @ 10:34pm
RussianNeuroMancer Dec 26, 2012 @ 12:44am 
Originally posted by jmite:
Had the same troubles on 12.04
Try to install driver follow this installation guide: http://wiki.cchtml.com/index.php/Ubuntu_Precise_Installation_Guide
Last edited by RussianNeuroMancer; Dec 28, 2012 @ 7:24am
eliaspirante Jan 27, 2013 @ 2:05pm 
Originally posted by jmite:
Had the same troubles on 12.04, the only difference was I had some luck with the hacks I'd found online, but still not what I'd call a reliable or working solution.

Well, i have a dell vostro 3550 with radeon hybrid hd6630m and intel 3000.... i use this link to make my videocard work! ^^. I only change de version of my ubuntu for make the .deb package.


edit -----

Sorry, i forgot the link.... and sorry for my english....


http://ubuntuforums.org/archive/index.php/t-1930450.html
Last edited by eliaspirante; Jan 27, 2013 @ 6:37pm
jonasan [#!] Mar 13, 2013 @ 5:45am 
running hybrid graphics on ubuntu has been a longstanding problem for me which ended up causing me to reinstall the os on a number of occasions. I really wanted it for the improved battery life when switching to intel instead of my amd card. Never found a workable drivers solution...... BUT...

by checking all the options in my system BIOS i found a way to shut off the intel chip and force the computer to use only the AMD card for graphics. Installing ubuntu like this and then the linux AMD drivers (following any guide) works fine. You dont have the switchable options but you have a system with all the 3d capability that your system should have. As a nice suprise ubuntu manages to get much more life out of my battery when using the AMD card than windows 7 ever did!

Turned out to be my prefered solution - now very stable and suprisingly capable 3d laptop. As far as i know there really is no SIMPLE supported way to get switchable graphics working so i highly recommend you just disable switching/intel in the bios! And steam runs perfect now too!! Good luck
Letalis Sonus Mar 13, 2013 @ 8:00am 
Don't forget that there are 2 different hybrid configurations: The old one with a multiplexer and the new Optimus-like one without a multiplexer. fglrx only supports the first one and there's no Bumblebee-like solution for the second one, yet - so it renders the discrete GPU rather useless. MUX-less configurations also lack any possibility to disable the integrated GPU in BIOS for static switching to the discrete one.

However, there's PRIME which can be used by the free drivers to fuel a MUX-less discrete GPU. It's still pretty cutting-edge, though, and comes with some RandR options which are AFAIK not even specified by any new RandR standard, yet.
mannerov Mar 13, 2013 @ 12:34pm 
"fglrx only supports the first one". Sorry to contradict you. I use an intel i5 and an amd hd 7730m (mux-less configuration) and it works with 13.2 beta and Kubuntu 12.10. I have to switch with the amd tools and reset the X server to use the dedicated graphic card or the integrated.
Letalis Sonus Mar 14, 2013 @ 5:45am 
And how do you know that it indeed is MUX-less? Just recently I once again stumbled over this line in someone's X log, printed by fglrx:
(EE) this is a Muxless PX A+I platform, we doesn't supported it

Supporting a completely MUX-less platform means getting the drivers to communicate with each other, sharing some of their buffers. If the Intel driver really already had such a non-GPL interface, we would already have true Optimus support as well - because it is really just exactly the same stuff, but instead Nvidia still works on some upstream workaround to get rid of the GPL-only restriction of the already available PRIME/DMA-Buf interfaces.

You have 2 GPUs and only 1 screen, you can't just stick those together. That's what a multiplexer is needed for. In the old times they muxed a lot more than just the screen so you really kind of exchanged the GPUs, nowadays the GPUs can just be used in parallel - don't confuse this. Optimus and entirely MUX-less AMD systems don't have any possibility to access the screen on their own and need to communicate with other drivers to do so, as mentioned above.
mannerov Mar 14, 2013 @ 6:53am 
amd hd 7700m-7800m-7900m series are mux-less.

Probably the X log you saw are with old catalyst versions.
The only thing that seems not to be implemented is vsync with the dedicated card. I have a line on it in my X log saying that TearFreeDesktop is not yet implemented.
Letalis Sonus Mar 14, 2013 @ 7:45am 
Originally posted by mannerov:
amd hd 7700m-7800m-7900m series are mux-less.
Concerning its PCIe interface, but not its screen. That's the point. There is still a multiplexer.
Last edited by Letalis Sonus; Mar 14, 2013 @ 7:45am
mannerov Mar 14, 2013 @ 2:41pm 
I advise you to read this article which points out the whole evolution of amd switching graphics: http://www.anandtech.com/show/6243/amds-enduro-switchable-graphics-levels-up/2
There is no multiplexer.
On windows, my dedicated card will only run heavy programs (I can choose which ones) whereas the integrated one will always run. I can have a window managed by the amd card while an other one is managed by the intel card.
On linux, I can change the card I use with amdconfig or amdcccle (after resetting the X server).
Probably there is a share between the two card because there is only one screen. Why not say the intel card to draw a buffer created by the amd card?. There is no issue do to that.

To have full optimus / enduro support, the cards have to use the same buffer, and due to license, they can't do that.

To understand better how the current implementation works, I'll quote an amd post on this bug report: https://bugzilla.novell.com/show_bug.cgi?id=780672
"The new issue is caused by Intel driver change to map the GTT aperture (bottom half of BAR 1) as WC, while the top half of BAR1 is still Un-Cacheable. Then the conflict happens when fglrx driver maps the whole BAR1 of Intel IGP device. We have already prepared a fix for this, which should be available in Catalyst 13.3. As a workaround before that, I think you can do some change in Intel graphics driver to map GTT aperture as Un-Cacheable."
Letalis Sonus Mar 15, 2013 @ 3:47pm 
Originally posted by mannerov:
I advise you to read this article which points out the whole evolution of amd switching graphics: http://www.anandtech.com/show/6243/amds-enduro-switchable-graphics-levels-up/2
It does not cover any technical details, just what can be seen fom the surface or in less interesting press releases. Anything specific you wanted to point at?

Originally posted by mannerov:
Probably there is a share between the two card because there is only one screen. Why not say the intel card to draw a buffer created by the amd card?. There is no issue do to that.
It is a HUGE issue. On the one hand, the cards (better: the drivers) really need to communicate to some point - imagine the Intel GPU writing to its buffers while the AMD GPU tries to fill them up. On the other hand it requires direct memory access across the PCIe bus - you can't just do this behind the back of the OS. Driver and kernel support of these is a must have. These issues are exactly what Nvidia is trying to solve for Optimus, and they still don't have any solution, as the only available kernel infrastructure provides only GPL interfaces and are therefore unaccessable for fglrx as well. If you simply put a multiplexer between the screen and the GPUs, it's a nobrainer - basically initialize the 2nd card and trigger the multiplexer, done.

The reason for the required X restart is simply that X was never designed to switch GPUs while running, once it has chosen a GPU to start with there is no going back - all sorts of dynamic configuration changes are limited to using a single GPU.
Bumblebee simply starts a 2nd root-less X Server using some OpenGL wrapper, which then connects to the primary X Server - so instead of buffer sharing it is just copying over all the stuff in user space, with a heavy impact on performance.

Originally posted by mannerov:
To have full optimus / enduro support, the cards have to use the same buffer, and due to license, they can't do that.
That's what I said. However, if they found a way to do this anyway, a restart of X would not be something required to achieve this - Intel has got KMS drivers, therefore the main part sits in the kernel and therefore does not care what X is doing. A Bumblebee-like solution with huge performance benefits would be easy to create in this situation.

Originally posted by mannerov:
To understand better how the current implementation works, I'll quote an amd post on this bug report: https://bugzilla.novell.com/show_bug.cgi?id=780672
That does not need to have anything to do with any memory interaction between both GPUs. WC is a processor-side feature of Intel to boost its IGP performance by using a burst-like write mode that would normally be counter-productive for GPU typical memory transactions, using some small buffers for transaction optimizations. WC is weakly ordered, which means that GPUs can only make use of it when they support this as they are usually only expecting strongly ordered memory. Being such a central feature in the system, incorrect usage can mess up a lot more than just another graphics driver.
As the Intel GPU is always the active one upon bootup, there's a lot that may go wrong when fglrx gets initialized, without even being related to actively sharing any memory regions (e.g. reusing old memory regions). Even worse: Since Sandy Bridge there's even per-process GTT - that's a lot more potentionally problematic code, and the support within the Intel driver seems to be rather dirty so far.
Last edited by Letalis Sonus; Mar 15, 2013 @ 3:47pm
The Hermit Mar 28, 2013 @ 4:01pm 
I've found that I can manually switch to one by using VGA_Switcheroo

https://help.ubuntu.com/community/HybridGraphics

However, as for what graphics driver to install, beats me
Showing 1-15 of 15 comments
< >
Per page: 15 30 50