Windows 7 vs. Linux w/ Sandy Bridge New Acceleration Architecture

[phoronix]

Phoronix: Windows 7 vs. Linux w/ Sandy Bridge New Acceleration Architecture

The new benchmarks going out today on Phoronix are looking at the performance of Intel's Sandy Bridge graphics with the latest Microsoft Windows 7 and Ubuntu Linux drivers. Not only are we using the very latest drivers, but there is also a separate Linux test run with SNA, the "Sandy Bridge New Acceleration" architecture enabled.

http://www.phoronix.com/vr.php?view=16216

[FireBurn]

Mentioning that the Linux drivers don't support OpenGL 3 yet has nothing to do with the benchmarks. None of the applications use OpenGL 3!

[MùPùF]

In a few days on Phoronix are more Intel Sandy Bridge benchmarks (from a Core i5 2500K) using the latest Linux Git code and comparing it to the Nouveau (NVIDIA) and Radeon (ATI/AMD) open-source drivers with a few different graphics cards.

Please try to upclock the card before doing so! I can give you detailed instructions if you want.

Safe reclocking is still being worked on, it works on most of my cards but not on all of them. I'm trying to get this done very soon but can't promise anything.

[deanjo]

Mentioning that the Linux drivers don't support OpenGL 3 yet has nothing to do with the benchmarks. None of the applications use OpenGL 3!

It is never implied that ogl 3 support is the reason for the performance difference. He was just pointing out differences in capabilities between the windows and Linux drivers.

[Shining Arcanine]

Please try to upclock the card before doing so! I can give you detailed instructions if you want.

Safe reclocking is still being worked on, it works on most of my cards but not on all of them. I'm trying to get this done very soon but can't promise anything.

I concur. The review will be worthless without this.

[smitty3268]

Obviously, the selection of multi-platform OpenGL games also continues to be limited since Mesa and Intel's Linux driver do not yet fully support OpenGL 3.0 and there are various other limitations of this open-source OpenGL library.

Is this true, or is Michael being "creative"* here? The Gallium drivers obviously have no problem with more complex games than Michael ever shows, and I suspect that the Intel drivers don't either but I'm not sure.


* Creative = technically true by default, but false for anyone who spends 5 minutes adding S3TC support, which is far easier and no less standard than, for example, the SBA feature he's also testing here

[LLStarks]

Try a real SNA test with hardware that actually benefits from it.

The 945GM is at least 10x faster now in benchmarks like the Fish IE test.

http://ie.microsoft.com/testdrive/pe...ce/fishietank/

Went from 2fps to 40fps.

[darkbasic]

SNA doesn't work for me (both Sandy Bridge and GM45) without the xorg patch. Where can I find it?

P.S.
With GM45 and Firefox4 it does 15 fps with 1000 fish (without SNA)

[smitty3268]

I know you don't want to do Firefox because it doesn't hook into your testing framework.

As an alternative, all it's really testing is XRender performance. So any XRender test should show the same results.

[MetalheadGautham]

This notebook has an Intel Core i5 2520M quad-core clocked at 2.50GHz,

Small (typo?) mistake here. Intel Core i5 2520M is a Dual Core.