Intel Sandy Bridge VA-API Performance

[phoronix]

Phoronix: Intel Sandy Bridge VA-API Performance

There have been a number of Intel Sandy Bridge articles on Phoronix since the January launch of this next-generation Intel CPU micro-architecture. It's ranged from the Linux support being challenging to dealing with motherboard problems and then the graphics performance being fast relative to previous generations of Intel graphics and for being based upon the classic Mesa driver architecture, but much slower than Windows. Last week then the Sandy Bridge Linux performance became much more interesting after a simple patch led to a huge performance win to the point that the open-source Linux driver performance is much closer to their full-featured Microsoft Windows driver. What is the next chapter in the Intel Sandy Bridge Linux story? A look at the VA-API video acceleration playback performance.

http://www.phoronix.com/vr.php?view=15756

[Kano]

I think i told you several times that the test file is a joke - bbb 1080p is compressed in a smart way to use a very low average bitrate, but that is not a good test to compare video accelleration. You need to use a high bitrate file, similar to what you get on bluray. Also vc1 could be tested, thats more interesting on nvidia ion 1 however. Very important is the cpu usage with vlc via vaapi as this is done much more inefficient than via mplayer.

[linuxjacques]

... You need to use a high bitrate file, similar to what you get on bluray. Also vc1 could be tested, ...

Totally agree. These tests are practically meaningless otherwise.

Show me the results from high-bitrate sequences from something like Avatar.

[Michael]

I think i told you several times that the test file is a joke - bbb 1080p is compressed in a smart way to use a very low average bitrate,

Where is any other high quality sample content that's widely available on HTTP/FTP mirrors and under a free license? I haven't seen such links.

[89c51]

http://www.sintel.org/ maybe?????

[liam]

Thanks Michael! Though the bitrate is low it should enter all the tricky bits of h264 decoding(cabac hopefully) so the test should scale similarly at higher nitrates.
These results are in accord with Anand where he found SNB to be capable of more simultaneous HD playback than any other common solution.
Great job Intel.

[curaga]

Ah, finally the va-api test. Thanks Michael!

Very important is the cpu usage with vlc via vaapi as this is done much more inefficient than via mplayer.

Why is testing bad code important? When better code is available?

[blackshard]

Ah, finally the va-api test. Thanks Michael!



Why is testing bad code important? When better code is available?

It's not a matter of better code, but it's a matter of different flow path.

[Michael]

I'm happy to test other media players too, it's simply a matter of someone submitting a test profile for such.

[deanjo]

Where is any other high quality sample content that's widely available on HTTP/FTP mirrors and under a free license? I haven't seen such links.

If you had a host that could hold the file I would recommend taking the raw pngs from sintel or elephant dreams and encode them using more demanding h264 settings. Elephant dreams has alot of motion going on in the background and would probably be the more demanding of the two.