Using NVIDIA's VDPAU On Mobile Platforms

[phoronix]

Phoronix: Using NVIDIA's VDPAU On Mobile Platforms

We know that NVIDIA's Video Decode and Presentation API for Unix (VDPAU) works very well for exposing PureVideo capabilities on Linux. We have benchmarked VDPAU and found it to perform very well in that under Linux it's possible to play HD videos with a $20 CPU and $30 GPU thanks to this video acceleration method. VDPAU is the best video acceleration / decoding API on Linux and is widely adopted by various multimedia applications, which is all in contrast to AMD's XvBA and their troubled implementation. But how does VDPAU work on mobile devices? With the ASUS Eee PC 1201N that is built on NVIDIA's ION platform we ran a new set of VDPAU video playback tests.

http://www.phoronix.com/vr.php?view=14483

[walkeer]

How can be CPU utlization while decoding 1080p h264 less than 100% while not using GPU acceleration (X-Video)? I am totaly sure that atom is not able to decode any "normal" real world 1080p h264 video, neither my Pemtium M dothan 1.5GHz does. Even 720p is very offen too hard to decode. I think there is some error in your suite. I had big problems getting 1080p work on intel 2GHz core duo CPU in apple mini, atom has simply no chance.

[Kano]

The used test file is just a joke, that's the problem. The bitrate must be very low, nothing compared to m2ts with 30-40 mbps average and 48 mbit peak.

[MiUNX]

I have this laptop, might check later today to see what the cpu usage is when playing 1080p vid.

[n0nsense]

Indeed I can't agree with this results.
Kano probably right, because my over 3GHz C2Q sometimes fails with some videos (they are stuttering or in slide show mode) and usually played nice on C2D with 9400 IGP.
Another way to look at it is that ION supposed to provide complete acceleration for MPEG-1, MPEG-2, VC-1/WMV9 and H.264.
H.264 was supported from the very beginning (of NV HD HW acceleration).
I'd like to see results for each one of them and of course to see if this thing can handle 50GB movie (real BD).
So, bottom line (or title) for this article should be "we are going to test 4x4 vehicle ability to deal with curb".
Disappointed

[Zhick]

Below the Power Usage Monitor-Graph

Using X-Video actually caused the system to consume a third of a Watt less than using X-Video.

I suppose that second X-Video is supposed to be VDPAU.

The decreased CPU usage though is clearly worth it if you wish to enjoy HD content on a NVIDIA-powered netbook or other portable device.

I don't agree with this. How is it "worth it" to use more power with VDPAU when the CPU alone is also capable of fluently decoding the video and actually consumes less power at it?

Anyway I agree with the others that it seems highly unlikely that an Atom is actually capable of decoding a 1080p h264 file fluently, even though your graphs do suggest so.
Maybe you could give us a download-link for the used video Michael?

[mgc8]

This has been a problem with Phoronix "1080p" testing for a long time... I believe they are using "Big Buck Bunny" or other such free media for testing, which has a very low bitrate and is quite easy to process. Alternatives have been suggested on the forum, however they appear to have been ignored.

Additionally, the reported CPU utilization is aggregated -- what this means on a 4 (logical) CPU machine like the Atom330 is that 25% utilization actually means 100% of one core. So when looking at the graph, it's quite obvious that with no acceleration, one core is almost pegged at 100%, while other system activity adds to get the 30-36% spikes. Even more telling, one can see a >25% spike with VDPAU as well. This would translate into lost frames or stuttering -- something that would be painfully obvious with high-bitrate movies, but it probably was unnoticeable with the tested content.

I have done my own Atom330+ION testing under Linux especially with 1080p/x264 high-bitrate scenes, and I can say it does not play them 100% perfectly (it loses frames), however that only happens rarely, it's quite useable for 99% of content. Of course, Flash is another question altogether (unless one rips and plays the FLVs with MPlayer).

[monraaf]

LOL, when I read the article my first reaction was wtf. An atom at 1.6GHz can easily decode 1080p H.264 without using GPU acceleration where my Athlon X2 @ 2.8Ghz struggles. I don't think so.

[deanjo]

This has been a problem with Phoronix "1080p" testing for a long time... I believe they are using "Big Buck Bunny" or other such free media for testing, which has a very low bitrate and is quite easy to process. Alternatives have been suggested on the forum, however they appear to have been ignored.

PTS uses Grey.ts for it's h264 test. It's bitrate chart looks like this.



One thing is for sure a better clip is needed. Re-encoding BBB during a scene with lots of movement at a high bitrate from the source PNG should suffice.

[L33F3R]

DEANJO! that looks like a windows screenshot to me!

You traitor!