NVIDIA Driver Brings PureVideo Features To Linux

[phoronix]

Phoronix: NVIDIA Driver Brings PureVideo Features To Linux

Over the course of the past few months we have been saying that the NVIDIA 180 Linux driver to be released in the fourth quarter of 2008 would hold in store a few interesting features. Well, today that closed-source driver has been released in beta form. This driver adds a new VDPAU API, which provides PureVideo-like features on Linux, adds in CUDA 2.1 support, new workstation performance optimizations, X Render improvements, and other improvements.

http://www.phoronix.com/vr.php?view=13102

[evanjfraser]

*Falls off chair*

[greg]

Where's the driver for x86_64? The links are dead.

[bulletxt]

"Before installing the 180.06 driver, be forewarned that it is considered beta quality."


instead, you should use that sentence for every AMD Catlyst release.

[Dragoran]

There are patches to use the api with mplayer and ffmpeg using apps:
ftp://download.nvidia.com/XFree86/vdpau/mplayer-vdpau-3076399.tar.bz2
See http://www.nvnews.net/vbulletin/showthread.php?t=123091 but seems only PureVideoHD is supported (ie. only newer cards GF6/7 are not supported)

Once we do some further testing, bugfixing, and cleanup, we will
contribute the MPlayer patches to the MPlayer developers.

[fragro]

What's about GStreamer support? (even AMD's libXvBA)

[curaga]

Well. Well. Well. This certainly was unexpected.

While I do like this move, what I don't like in either Nvidia's or AMD's current moves is that they only support their latest generation's video processors, even though the last 2/3/4/5 have the same technology - meaning they could have Purevideo (non-HD) and UVD1 working quite fast using the more advanced code, but they don't.

[_txf_]

In nvidia's case they support the last two generations of purevideo as opposed to amd which only supports the latest gen umd2. Purevideo 1 quite a bit different (and much more limited) from subsequent pv hardware revisions, that it is probably justified.

[jeffro-tull]

wait a minute, not only do Nvidia's drivers (much) more often than not Just Work, but now they have the GPU doing some of the heavy lifting when playing videos?

man, I loves me some open source drivers, but this seems pretty big. Now I'll actually have some decisions to make when I build an HTPC.

[korpenkraxar]

This driver adds a new VDPAU API, which provides PureVideo-like features on Linux



Hang on, did they just get what has been bragged about and promised but never delivered to us poor ATI owners or potentially future poor ATI owners? Accelerated MPEG-1, MPEG-2, H.264, and VC-1? Right about the only thing except for the availability of open source drivers that would make it worth buying ATI cards? A driver which enables smooth non-teared Linux HD playback on a $35 GeForce 8400GS video card or a $70 GeForce 8200 motherboard? With a working implementation ready to drop into Mplayer? Instead of shame and trouble, spouse approval?

[deanjo]

Hang on, did they just get what has been bragged about and promised but never delivered to us poor ATI owners or potentially future poor ATI owners? Accelerated MPEG-1, MPEG-2, H.264, and VC-1? Right about the only thing except for the availability of open source drivers that would make it worth buying ATI cards? A driver which enables smooth non-teared Linux HD playback on a $35 GeForce 8400GS video card or a $70 GeForce 8200 motherboard? With a working implementation ready to drop into Mplayer? Instead of shame and trouble, spouse approval?


Pretty much, I hate to tell people I told them so but.......

[Kano]

Well VC1/WMV3 acc is only available for a few chips like G98 / NV 9300. So you an forget that. And currently h264 does only work with very few videos out there, but at least the NV dev are working on newer profiles (right now only up to 4.1). As long as it works the cpu usage is really low - even for hd material. Now it just has to be improved. Lets see if ATI provides mplayer patches too to demonstrate their new acc libs.

[korpenkraxar]

And currently h264 does only work with very few videos out there, but at least the NV dev are working on newer profiles (right now only up to 4.1).

Well it seems like H.264 is definitely becoming the de facto standard for HD "material". Are you also referring to DivX?

As long as it works the cpu usage is really low - even for hd material.

If it enables smooth 1080p playback perhaps with some optional post processing down the line on an el cheapo card, who needs a Phenom rig with Hyper Transport 3 or similar Quad Intel rig for his (ok, or her) HD HTPC?

Lets see if ATI provides mplayer patches too to demonstrate their new acc libs.

After they wake up wondering what hit them.

[Kjella]

And currently h264 does only work with very few videos out there, but at least the NV dev are working on newer profiles (right now only up to 4.1).

It's a bit confusing, there are levels (bitrate) and profiles (encoding methods). Blu-Ray uses L4.1 High Profile, so I assume you mean to say they're only supporting some of the lesser profiles at the moment.

[Kano]

Well you can not watch for example this:

http://mirror.bigbuckbunny.de/peach/bigbuckbunny_movies/big_buck_bunny_1080p_h264.mov

[sundown]

If I were AMD, I'd be really scared.

[Vadi]

Does anyone know if this fixes the weird Qt4 rendering + compiz bug?

[DeepDayze]

If I were AMD, I'd be really scared.

At least nvidia is moving in the right direction. Perhaps the stable build will have improved PV support by then. AMD/ATI need to get their act together if they want to stay alive in the Linux space

[bugmenot]

Hm. I don't know if I should think this is a good thing. :/

nVidia tries to create an own proprietary video acceleration although there are open source approaches. Can't they thy to use something like XvMC or try to improve it or even use VA API? But AMD is not better: They made an own proprietary video driver that isn't even ready to use. Especially they should have used VAAPI or something like this.

Now nVidia and AMD have made duplicated efforts...

[myxal]

Hm. I don't know if I should think this is a good thing. :/

nVidia tries to create an own proprietary video acceleration although there are open source approaches. Can't they thy to use something like XvMC or try to improve it or even use VA API? But AMD is not better: They made an own proprietary video driver that isn't even ready to use. Especially they should have used VAAPI or something like this.

Now nVidia and AMD have made duplicated efforts...
First - am I reading something wrong? Where is it mentioned that VDPAU API is proprietary? In fact, quoting the first news article on VDPAU: The VDPAU documentation is currently housed within a header file in /usr/include/vdpau/ that contains hundreds of lines of documentation and covers threading, extending the API, and more. VDPAU is also supported in the 180 releases of the FreeBSD and Solaris display drivers.Doesn't read like proprietary to me :-\

XvMC is dated and plain simply doesn't offer many features essential to future video-watching experience (no frame-by-frame stepping? subtitles? no PIP? etc... WTF?).
AMD's "solution" involves essentially porting DXVA to Linux, to minimize the work needed - they have already developed the code for their windows driver, so... you can't really talk about much effort on AMD's part.

As far as API wars go, I'd like to hear someone competent look at XvBA and VDPAU and compare them. I guess if VDPAU is signifacantly better than the DXVA port, it makes sense to support it (MOAR FEATURES!!! :p ) Otherwise the point of maintaining two APIs and codebases (DXVA vs. VDPAU) simply escapes me.