Some Good & Bad News For The Nouveau Driver

[phoronix]

Phoronix: Some Good & Bad News For The Nouveau Driver

With our big AMD Linux GPU / driver comparison we found its open-source Gallium3D driver to be noticeably faster than the classic Mesa DRI driver across an array of Radeon hardware from multiple generations. However, the official Catalyst driver was multiple times faster (roughly 5.18x faster) than the Gallium3D driver, not to mention its lack of proper support for OpenGL 3/4, VA-API/VDPAU/XvBA video playback, and many other features only found within the proprietary Catalyst driver. Now though it is time to see how the Gallium3D Nouveau performance compares to that of NVIDIA's proprietary Linux driver across different GeForce graphics cards.

http://www.phoronix.com/vr.php?view=15607

[marek]

I think VDrift just uses a different feature set for the GL3 and GL2 APIs, with the GL2 code path being simpler and therefore faster than the GL3 one.

[Artemis3]

The 8800GT and the 9800GT are virtually identical. Cards using the g92a chip can happily crossflash into either without issues. Some 9800GT come with the g92b chip which is 55nm instead of 65nm, and allows tri-sli. The 9800 firmware supports phys, and thats about it.

[Thunderbird]

Really make sure that both drivers are running the GPU at the same clock speeds else especially on the 8600GT / 8800GT you may be comparing apples to oranges.

[Shining Arcanine]

How is being within 50% of the proprietary driver's performance good? That is not acceptable for gamers.

[not.sure]

How is being within 50% of the proprietary driver's performance good? That is not acceptable for gamers.

You must be new here.

[r1348]

How is being within 50% of the proprietary driver's performance good? That is not acceptable for gamers.

Are you even aware on what and how nouveau is?

[0e8h]

It goes to show that the programming is so poor for the non-proprietary drivers as they should be lighting speeds ahead whilst they aren't doing all the features. Other words if a open source driver is only doing vectors and the proprietary-driver is doing vectors and buttering bread, then the former should be faster.

The non-proprietary coders must be using bad kernel techniques, or not using proper hardware abilities.

Get the vector speeds first, then start adding the buttered bread later else your project is pointless if it's going to be always slow!!!

[0e8h]

How is being within 50% of the proprietary driver's performance good? That is not acceptable for gamers.

I agree. Something is seriously wrong. People maybe willing to use a modular driver model if is was say 80% of another non-modular.

The programmers need to step back ten steps and looks at their project and see what they are overlooking or what needs to happen to fix the direction.

Should the driver stick to 2d, not 3d?

[saski]

given the fact that the nouveau driver is being developed without any technical documentation support or help from NVIDIA and with a much smaller headcount of developers this performance result is absolutely awesome!

great work folks! keep it up!