22-Way AMD+NVIDIA Graphics Card Tests With Metro Redux On Steam For Linux

[Vulkan]

I know the concept about DX11 and its different feature levels.
But does it always require at least a HD 5000/Geforce 400 GPU to use OpenGL 4? The Windows version supports the 8800GT.

[Jabberwocky]

The other thing is you dont need 4 different Northern islands. 3 different Sothern Islands.... The highest end card from each generation would do just fine. (I understand then you wouldn't be able to tout a "22 card comparison", but the information gathered would still be exactly the same.)

It's very useful for driver developers that read Phoronix to sport regression(s)/optimisation-boost(s), also interesting for me as normal user. It's seldom that cards perform as expected.

[Passso]

It was just trying to be as large of a comparison as possible given it's the first time doing Metro Linux tests.

Michael, you are so right putting several models and generations.
I possess a 660GTX so I can compare my card to newer Nvidia and AMD cards.

The point is not only to have 780 VS 980 VS x290, it is very good to have a big picture of card performance and know if you have a nice one or need a change.

Great job I am very enthousiast about this "real" game complete benchmarks

[Swiftpaw]

Should do a comparison between Windows and Linux with Metro, specifically without SSAA. You'd be surprised of the results.

https://www.youtube.com/watch?v=Eab2UI1Iq2w

That is pretty amazing yeah, the OGL implementation of SSAA seems to be really poor because without it Linux screams ahead. Or it's a driver issue.

Is the scene quality identical between the two with SSAA disabled though? I'd guess the issue is probably poor driver optimizations for the more advanced OGL4 features due to those features seeing reduced use (until now) on the Linux side.

[peppercats]

Hi Michael, could you possibly include some Windows results too?(i.e, just a 970 and a 290x or something)

I remember reading that Metro games run awful on Linux compared to Windows, so it would be nice to compare.

[Michael]

Hi Michael, could you possibly include some Windows results too?(i.e, just a 970 and a 290x or something)

I remember reading that Metro games run awful on Linux compared to Windows, so it would be nice to compare.

Unfortunately cross-platform comparisons wouldn't work out as AFAIK the Windows version of the Metro benchmark program offer a variety of options to run at where as the Linux version doesn't offer any options really at all for benchmarking to control the image quality settings from the CLI let alone the resolution, so the feature levels in the comparison would likely not correlate. So only if the Metro Linux benchmark was improved upstream would it become more feasible.

[mum1989]

PhysX

If PhysX works on GNU/linux side
Is PhysX off on our test ?

[oleid]

Wow, Catalyst is really good for those non SI cards! Even better than comparable NVidia cards. I wonder if they can do the same magic for the recent cards!

[Tiger_Coder]

Should do a comparison between Windows and Linux with Metro, specifically without SSAA. You'd be surprised of the results.

https://www.youtube.com/watch?v=Eab2UI1Iq2w

I have noticed something like that long ago with Unreal 2004. Ubuntu trends to perform better on when cpu is bounding. Windows when GPU is bounding. Without SSAA, GPU limitations are not that severe.

[dimko]

Should do a comparison between Windows and Linux with Metro, specifically without SSAA. You'd be surprised of the results.

https://www.youtube.com/watch?v=Eab2UI1Iq2w

I have opened another thread.

http://www.phoronix.com/forums/showthread.php?112376-kernel-process-responsible-for-GUI

Look at your video min 4.30
Even when frame rate is high, you can see a slide show.
I have explained in my thread how to get rid of it.
So you my Linux brethren can have better experience.