Intel SNA Acceleration Architecture Continues To Mature
A few days back when testing the Linux 3.1 kernel with Intel's Sandy Bridge hardware and then the Intel RC6 power-savings support, I also ran some updated benchmarks of SNA, the new Intel acceleration architecture available from their graphics driver.
The Sandy Bridge New Acceleration Architecture (SNA) was originally introduced in June, was quick to advance in order to work out very visible bugs, and was extremely fast. This acceleration architecture may be targeted for current-generation Sandy Bridge graphics, but it even speeds up 2D/3D for older generations of Intel hardware.
The last testing I did on Intel SNA was in early September, but Chris Wilson has been quick to work on improving this Intel feature. It is still not the default acceleration path in the xf86-video-intel DDX, but Chris works it on near daily. In fact, in the few days since these benchmarks were done, a number of SNA-related commits have already been pushed to xf86-video-intel Git master.
This testing was done using the Git graphics components on the 27th of October, including Mesa 7.12-devel git-6437a71 and the xf86-video-intel head was at 990043f. The Linux 3.1 kernel was also employed. Testing was from the same HP EliteBook used in other recent Sandy Bridge articles on Phoronix with the Intel Core i5 2520M CPU.
Ubuntu 11.10 was the base operating system and the results in this article compare the SNA performance when the xf86-video-intel DDX was built with and without the --enable-sna flag for enabling the Sandy Bridge acceleration architecture back-end.