Intel SNA Ivy Bridge - September 2012

[phoronix]

Phoronix: Intel SNA Ivy Bridge - September 2012

Due to the prolific driver development work being done by Intel's Chris Wilson to advance the SNA 2D acceleration architecture for the xf86-video-intel driver, here's some new benchmarks of "Sandy Bridge New Acceleration" when being run from an HD2500-class Core i5 "Ivy Bridge" processor using the new driver release from Sunday.

http://www.phoronix.com/vr.php?view=17834

[bug77]

It's all nice and everything, but this driver seems to require a fairly recent kernel, as it's not available for Ubuntu 12.04 or Fedora 16

[abraxxa]

I've not have time to dig deeper so I ask here: can SNA be enabled on Ubuntu 12.04?
If yes can it be done without a complete xorg.conf and is it stable?

[ickle]

I've not have time to dig deeper so I ask here: can SNA be enabled on Ubuntu 12.04?
If yes can it be done without a complete xorg.conf and is it stable?

It requires an updated xf86-video-intel to a 2.20 series driver and an xorg.conf snippet to select SNA:

Section "Device"
  Identifier "intel"
  Driver "intel"
  Option "AccelMethod" "sna"
EndSection

It should be stable, but it is as yet unproven, hence why it UXA remains the default for the time being.

These set of benchmarks still have some bogosity behind them. There should be no difference in the copy throughput since that is GPU bound using identical operations, but most of the rest can be explained with the choice of unity wm and using ubuntu. Comparing results from a mobile GT2 device (a i7-3720qm) with this article a deskop GT1 chipset, http://openbenchmarking.org/result/1...SU-1209034SU66, we get the impression that the system under test underperformed significantly.

[Michael]

These set of benchmarks still have some bogosity behind them. There should be no difference in the copy throughput since that is GPU bound using identical operations, but most of the rest can be explained with the choice of unity wm and using ubuntu. Comparing results from a mobile GT2 device (a i7-3720qm) with this article a deskop GT1 chipset, http://openbenchmarking.org/result/1...SU-1209034SU66, we get the impression that the system under test underperformed significantly.

I have some more cross-desktop 2D tests coming soon.

[ickle]

I have some more cross-desktop 2D tests coming soon.

That will be fantastic. As always the challenge I have is to address why the out-of-the-box experience with Ubuntu or Fedora or Gentoo or Debian or... never quite matches up to my expectations, which is frustrating.

[kozaki]

I have some more cross-desktop 2D tests coming soon.

+1, this is great a news indeed! looking forward to read these

[ickle]

I have some more cross-desktop 2D tests coming soon.

Just an idea for a future pair (series? ;-) of articles: Pick a single benchmark (of reasonably short duration) and ask everybody to run it on their current system and upload to openbenchmarking.org, then have fun trying to make sense of the data.