It was earlier this month that version 2.1.0 of the xf86-video-intel driver was released, which among other things had introduced open-source Linux graphics support for the G33, Q33, and Q35 chipsets as well as fixing a horde of bugs and adding PCI IDs for the 945GME, 965GME, and 965GLE chips. As our last Intel graphics performance article was looking at the Q965 back in May, in this article we have enclosed some benchmarks from Intel's GMA 950 IGP using the new xf86-video-intel 2.1.0 driver.
The xf86-video-intel driver can possibly have some impact on 3D performance.
For example, up and coming versions of Mesa will support more advanced memory management schemes that will directly impact DRI driver performance. It will need support from the xf86-video-intel driver for that. Also driver releases tend to go along with new X.org releases that may have some overal performance increase.
But other then that... yep the xf86-video-intel driver is pretty irrelevent when it comes to 3d performance.
Mesa3D.org released their 7.0 release late June. This is the stable release supporting OpenGL 2.1 API. They also have a GLSL compiler and some shading language support among other things.
This does not necessarily translate directly to shading support in the DRI drivers, even if the hardware supports it.. but its definitely a step in the right direction.
As you probably know DRI drivers are derived from the Mesa OpenGL stack, which is obviously what Intel uses for 3D performance.
It would be interesting to see how this translates to GMA X3000 or GMA X3100 (Santa Rosa Laptops) performance since the hardware is capable of accelerating that sort of thing.
I know there is a 965-glsl branch for shading language support.