AMD's Windows Catalyst Driver Remains Largely Faster Than Linux Drivers

[phoronix]

Phoronix: AMD's Windows Catalyst Driver Remains Largely Faster Than Linux Drivers

With last week having delivered our latest Linux vs. Windows NVIDIA benchmarks where we found that the NVIDIA Linux driver can outperform the Windows 8.1 driver with OpenGL workloads, the tables have turned to looking at the AMD Windows vs. Linux performance using the latest code. In this Ubuntu 14.10 vs. Windows 8.1 comparison, the open-source Radeon driver on Linux is also being tested against the Catalyst drivers.

http://www.phoronix.com/vr.php?view=21115

[profoundWHALE]

Results like this makes me wonder what AMD is doing on their Linux side for the Catalyst drivers to get that performance. That being said, 90-95% of their install base is most likely still going to be Windows, so it's not all that surprising either.

[M@GOid]

Unigine Valley

It work in my R9 290 with RadeonSI. (Kubuntu 14.10 and Kernel 3.17.2 + Oibaf)

The tests in Windows are using OpenGL? The AMD driver for OpenGL in Windows is very bad in the R9 290. I have slowdowns and stuttering in Rage and Euro Truck Simulator 2 have abysmal performance in OpenGL, compared to using DirectX in the same game.

A monday install of Oibaf PPA and Kernel 3.18rc3 reduced the stuttering I was having with ETS2 in Linux. L4D2 have more consistent performance too. Too bad the problem with 120Hz displays continue.

[dungeon]

Results like this makes me wonder what AMD is doing on their Linux side for the Catalyst drivers to get that performance. That being said, 90-95% of their install base is most likely still going to be Windows, so it's not all that surprising either.

I don't see a clear winner there - it is 6:3 (-1) kind od situation on these tests... largely faster does not mean completely faster

BTW i see large/micro stutter on both Windows/Linux with in Unigine Valley, with Aero/composite enabled . So somehow catalyst/flgrx in some cases ( if not all ) does not like composited desktops .

[curaga]

No cpu and gpu usage

It's this kind of articles that would have benefited.

[nadro]

No cpu and gpu usage

Yes, these statistics would be very useful.

[Michael]

No cpu and gpu usage

It's this kind of articles that would have benefited.

This testing had started before your recent requests but will be trying in the future to at least have them on OB for you. At least for CPU usage as I don't need to install anything extra (ala radeontop). There's also unfortunately no Windows CPU usage monitoring support in PTS code right now.

[lowflyer]

Please include other AMD cards in these tests too

Michael,
I'd like to see cards like the HD 6450, HD 6950, HD 6990, HD 7980, HD 7970 in these tests. Cards like the Radeon HD 6450 did hit the market in 2011, these are only 3 years old and by no means outdated. I suspect these cards to be in more widespread use than the newer cards you tested. I don't think the newer ones are adopted by the market. Seeing the problems that these newer cards have, this would not be a surprise.
And the test results you produce with half of the tests failed are really not saying anything.

[curaga]

This testing had started before your recent requests but will be trying in the future to at least have them on OB for you. At least for CPU usage as I don't need to install anything extra (ala radeontop). There's also unfortunately no Windows CPU usage monitoring support in PTS code right now.

Thanks!

[edward.81]

Michael,
[...]And the test results you produce with half of the tests failed are really not saying anything.

We need more benchmark with quake based engine.