http://demos.hacks.mozilla.org/openweb/HWACCEL/
60+ fps
(one only gets a few fps with Intel graphics)
I personally don't run Wine or games, since I use my coputer for computing. This laptop with Radeon Mobility graphics was the same price as one with Intel graphics, but the AMD/ATI graphics hardware is of course far superior.
The Radeon driver is not slow at all.
If Wine has a problem with the open source Radeon driver, then it is high time Wine fixed this. If they can't figure out how, perhaps the Wine project should simply adopt the Gallium3D state tracker for Direct3D.
http://www.phoronix.com/scan.php?pag...3d_d3d11&num=1
Last edited by hal2k1; 02-21-2012 at 12:48 AM.
You are correct. In Firefox 11, the trouble shooting information says this:
All the more impressive then that the Radeon open source driver can still achieve 60+ fps for this test (where Intel drivers get only a few fps) when Mozilla is not utilising 3d acceleration features directly. Firefox is getting this rendering performance boost via the Linux OS graphics layer drivers rendering the desktop.Code:Graphics Adapter Description: X.Org -- Gallium 0.4 on AMD CEDAR Vendor ID: X.Org Device ID: Gallium 0.4 on AMD CEDAR Driver Version: 2.1 Mesa 8.0-rc2 WebGL Renderer: X.Org -- Gallium 0.4 on AMD CEDAR -- 2.1 Mesa 8.0-rc2 GPU Accelerated Windows: 0
If it is still slow, then it is almost certainly a problem with Wine. AFAIK nothing else has a problem with the ever-imporving performance of the open source Radeon driver for Linux.I think it's the driver code, not Wine. Wine is coupling to the open GL routines, so i'd guess it's not entwining itself with any driver projects.
AFAIK full OpenGL 3 compliance for the open source Radeon driver just missed out by a matter of days for the release of Mesa 8.
http://www.phoronix.com/scan.php?pag...tem&px=MTA1MDM
AFAIK it is available in Mesa 8.0.1. So why can't Wine use it properly when everything else can?
Last edited by hal2k1; 02-21-2012 at 01:17 AM.
You seem to be under the mistaken impression that KDE devs have the ability to contact someone at AMD. They don't. Their options are:
1. Open a ticket on the public bugtracker. I think we all know how much good that would do.
2. Write a blog informing all interested parties and hope word trickles in to AMD. That's the current solution.
3. Maybe try and find a forum that AMD sometimes visits, and see if you can find someone there. That's basically Phoronix, where we can hope Bridgman notices this and passes on word to the fglrx team.
There really isn't a 4th option - AMD has no open source evangelist taking calls from open source teams. At least not on the KDE side, and clearly not on the Gnome Shell side either.
Perhaps one of the distros, like Fedora, could facilitate a phone call. I guess that's option 4.
In the case of fglrx, the driver is just clearly broken.
I see no reason why Window Managers have to bloat themselves with dirty workarounds and legacy code in order to make the broken drivers work with them. I really hope this move will finally convince AMD devs to fix their awful binary blob because it's really disappointing to see that fglrx still cannot handle OpenGL 2.x/ES correctly...
Anyways, don't forget it's just about the Composited mode of KWin. Users of legacy/broken/unsupported drivers/hardware can still run KDE just fine without Compositing.
I guess it will really bother only when Wayland will be around, since it requires Compositing to be enabled.
Last edited by Scias; 02-21-2012 at 01:22 AM.
It seems like I end up explaining this an awful lot.
Firefox accelerates that page through XRender, which uses the 2D X driver. It gets that automatically by drawing everything through Cairo.
3D rendering is supported through WebGL, using the OSS and fglrx drivers.
3D acceleration, through layers, is not implemented yet, but is frankly rather limited in what it will do anyway. The main work is being done by Cairo/XRender (at the moment - there are plans to change this eventually).
60+ fps - with intel graphics (HD3000)!
But there will be other problems, when people get forced from fglrx to radeon: missing OpenCL and video acceleration. Especially those low-power-devices (e450 etc) need it for decent performance. And ATM. it is only supported through fglrx (though radeon-guys say they are working on it).
And Xrender, in turn, uses the GPU hardware acceleration fratures exposed by the graphics driver.
http://en.wikipedia.org/wiki/X_Rendering_Extension
Performance
It is designed to target the 3D graphics capabilities of newer video cards.
And Xrender uses the GPU hardware acceleration features of the Radeon driver.
Mozilla are eventually going to switch to their own graphics canvas called Azure.
http://www.phoronix.com/scan.php?pag...item&px=OTcxNQ
Azure will use the GPU hardware acceleration via OpenGL.
All being well, Azure will be even faster in using the GPU & driver via OpenGL that Cairo was in using the GPU & driver via Xrender.The back-ends for the Azure library are Direct2D, Quartz, OpenGL, Direct3D, and Cairo. On Windows, Direct2D is the primary target for Azure while Apple Mac OS X users will have Quartz, and Linux users have OpenGL. If hardware acceleration fails, Azure can fall-back to having Cairo sit underneath. Cairo is also used for Gecko's printing support. Mozilla's copy of Cairo has also been modified to take advantage of Azure's internal stateless surface API. Direct3D 9 and Direct3D 10 back-ends for Azure are also being developed.
I prefer the radeon driver but on my laptop I have had to switch to the blob to keep save power. And for gaming the open driver is to slow, be it wine or not, the open driver can even be a bit slow running my desktop in certen situations.