AMD Radeon: Windows Catalyst vs. Linux Gallium3D vs. Linux Catalyst

[GreatEmerald]

What is wrong with triangle test on both intel and radeon drivers

http://openbenchmarking.org/embed.php?i=1403248-PL-RADEON28080&sha=7450c70&c=6cb8960a456f3799089e4f1e7d5c034b8a6c3eb3&p=1

I guess that needs to be decent performer if someone wants later tesselation to play at decent speed

Isn't the triangle test a simplified version of glxgears, thus not relevant to anything (but coil whine, I suppose) in the slightest?

[schmidtbag]

These are nice results. Last time I saw benchmarks like these, the radeon drivers weren't worth using and linux catalyst was behind windows in nearly every single test by a noteworthy margin. Now, the FOSS drivers are barely behind (at least for R600) and catalyst finally (overall) matches the speed in Windows.

Does anybody know if the catalyst game profiles for micro-optimizations are transferred to linux's catalyst? If not, that could explain why the linux catalyst falls behind in a few cases.

[dungeon]

Isn't the triangle test a simplified version of glxgears, thus not relevant to anything (but coil whine, I suppose) in the slightest?

Maybe, i just guess performance of triangles will be related to the tesselation's factor scalability and thus speed later... but of course we will know when we have OpenGL 4.0 with Mesa drivers .

[przemoli]

Maybe, i just guess performance of triangles will be related to the tesselation's factor scalability and thus speed later... but of course we will know when we have OpenGL 4.0 with Mesa drivers .

IMHO that lower score is more related to data transfers. Which is not the bottleneck for tessellation (number of output vertices per pass is).
So that bench should not matter. (And it really do not matter at all at anything...)

[bridgman]

AFAIK the triangle test measures how quickly a large number triangles can be fed from CPU to GPU through the driver, but tessellation generates its "large number of triangles" on chip using the hardware graphics pipeline, so there probably won't be a strong correlation between the two.

There's always a weak correlation, of course, but that usually just means "if nobody had time to optimize A they probably didn't have time to optimize B either"

[molecule-eye]

Interesting set of benchmarks, but I'm still interested in non-repeatable Source engine benchmarks, like L4D2 and Dota 2.

Since I'm not interested in Catalyst, these benchmarks make me want to do a cheap upgrade of my existing system by dropping in a used 6870. Pretty cheap on ebay!

[FourDMusic]

Phoronix: AMD Radeon: Windows Catalyst vs. Linux Gallium3D vs. Linux Catalyst

@Michael. Suggestion for the power wattage and total frame time graphs. At the moment you have two columns with the colors for each graphics card and the average/peak/low information is displayed right above the graph. I would recommend having one row of columns and putting the information next to teach column like so (but make the avg/peak/low horizontal):

[AJSB]

Here is *MY* Review...

MoBo:
ASRock FM2A88X-ITX+

RAM:
G.Skill 8GB @ 2133MHz

APU:
A6-5400K
Turbo OFF
APM OFF
Cool'n'Quiet OFF
iGPU OC to 800MHz

Game:
Enemy Territory: QUAKE Wars
Graphics settings same to all tests
Game always played in native mode and fullscreen in all OS....so , no WINE in Linux.


WindowsXP + latest Catalyst:
VERY High frame (peaks up to 200FPS ingame , NEVER goes bellow 120 FPS) , EVERYTHING very smooth , NO graphical glitchs

Windows7 + latest Catalyst:
Tops at 60FPS (older drivers peak up to 120FPS ingame and) goes down to 20 FPS , LOTs of stuttering in special when straffing, NO graphical glitchs

Windows8 + latest Catalyst:
Same as W7, but even lower frame rates

Windows8.1 + latest Catalyst:
Same as W8, but even lower frame rates

XUBUNTU 14.04 LTS + latest Catalyst:
Very decent frame rates (peaks more than 120 FPS easy, frame rate VERY consistent) VERY SMOOTH but LOTS of graphical glitches, so many that *i* don't want to play.

XUBUNTU 14.04 LTS stock using OSS RADEON:
I hear the music when game is launched....and..."black screen of death"...nothing more than the music...only way to get out is using the REISUB trick.


Didn't try WINE and even less WINE + RADEON + ET:QW


PS: Linux tests done with XUBUNTU, latest build of UBUNTU doesn't boot in my rig and i only get a black screen ....4 days ago build, booted.

[Michael]

@Michael. Suggestion for the power wattage and total frame time graphs. At the moment you have two columns with the colors for each graphics card and the average/peak/low information is displayed right above the graph. I would recommend having one row of columns and putting the information next to teach column like so (but make the avg/peak/low horizontal):

Patches welcome. It would be a nice improvement but I don't have time to do it currently. It would be a fairly easy change to make with refactoring just a few lines of PHP code in PTS via pts-core/objects/pts_Graph/

[F i L]

XUBUNTU 14.04 LTS + latest Catalyst:
Very decent frame rates (peaks more than 120 FPS easy, frame rate VERY consistent) VERY SMOOTH but LOTS of graphical glitches, so many that *i* don't want to play.

curious if you where using composting while testing on Xfce?