The Ironlake back-end regresses a great deal when enabling SNA for the JXRenderMark Put Composition test, but for the Gen3 back-end it is faster with SNA.
The JxRenderMark Simple Blit performance was a great deal faster for both back-ends of Sandy Bridge New Acceleration.
For the transformed blit linear test, the JXRenderMark performance for both Intel systems was faster when enabling SNA.
When going to bilinear, the Arrandale hardware was faster with SNA, but the Intel Gen3 performance was close to the same, if not regressed a little. This was similar with the transformed texture paint too.