May 2011: Gallium3D vs. Classic Mesa vs. Catalyst

[phoronix]

Phoronix: May 2011: Gallium3D vs. Classic Mesa vs. Catalyst

The open-source graphics driver landscape is ever changing with new work going into Mesa / Gallium3D near daily. While many improvements have been made in recent time, the open-source drivers have a ways to go in competing with the proprietary competition. Even the open-source AMD driver, which is developed using documentation from AMD as well as code and engineering resources within the company, it has a tough time competing with the well-optimized Catalyst driver. Fortunately, the AMD driver is now largely centered on the two Gallium3D drivers: R300g and R600g, and have pushed away their classic Mesa DRI drivers into maintenance mode. The R300g supports the R300 through R500 ASICs (up through the Radeon X1000 series) while the R600g driver supports all ATI/AMD hardware past that point up through the latest Radeon HD 6000 series and Fusion. In this article, we are seeing where the performance is currently at for the classic Mesa, Gallium3D, and Catalyst drivers under Linux.

http://www.phoronix.com/vr.php?view=15970

[jakubo]

i wonder how the performance would scale down if catalyst was used on a single core.

[mangobrain]

Still no Cayman support last time I tried

I have a 6950, one of the last remaining unsupported cards, it would seem. Well, unsupported in terms of 3D acceleration, anyway - KMS works, but no direct rendering. I really want to like GNOME Shell, but can't give it a fair try, because if I use the open-source drivers or llvmpipe it goes into fallback mode, and if I use the proprietary drivers I get creeping texture corruption until the session eventually becomes unusable.

Wake me up when Gallium3D support arrives for Cayman....

[NomadDemon]

this time you used good colors, its very "lookable"

nice test, I wonder, when i would be able to use OSS drivers with my 4850 with same performance

[Jimbo]

this time you used good colors, its very "lookable"

Agree, beside the ordering problems, the charts look great this time, nice job.

[pingufunkybeat]

What exactly was the problem you had with Warsow?

I tried it a few days ago with r600g, and gave it a platinum rating on RadeonProgram, because it worked perfectly.

Anyway, it's pretty clear that the r600 drivers are CPU-bound. Nice results on Urban Terror, though, that gives hope.

[Kano]

There are too many values at 30 fps. 60 = 2*2*3*5, so any combination like 10, 15, 20, 30 would sync with SwapBuffersWait. Basically the results should be higher when this would be turned off (like fglrx which has got no such limit). Also libtxc_dxtn was not mentioned, why not? The results are usually not that much higher, but at least it would be more like fglrx. Also nouveau has got no such limits... When you add those results then ati will even look slower

[chithanh]

I think in Urban Terror and possibly other tests fglrx is CPU limited. With a higher resolution (dual-screen 3840x1080 or so) maybe the difference between fglrx and the open drivers would be larger.

[b0uncyfr0]

You know what, the catalyst driver aint bad when it comes to performance. Its just the integration with the desktops that pissses me off. For example when im playing a video on my KDE desktop and switch activites, the video stutters then goes back to normal. When im browsing, a video playing in the background CONTINUOUSLY stutters. None of these happen with the open drivers.

Its these little things that really put me off their drivers. A modern day card has enough power to handle these ATI. You need to tune your drivers not performance wise but for general activities as well.

Nice comparison btw, the open-drivers are getting there but still needs more time.

[monraaf]

These benchmarks look fishy. Last time I tried openarena anholt demo my fps was much higher on HD 5750.

Are you running with swapbufferswait on again? It's likely to have a negative effect on fps. You can argue all you want about defaults being used, but with this setting on it's unlikely you'll see much performance progress being made in some benchmarks, making the benchmarks kind of pointless.

Insanity: doing the same thing over and over again and expecting different results -- Albert Einstein