VDPAU and Generation 1 PureVideo

[jmfernandez]

Then, although most 7xxx and 6xxx based hardware supports some H.264 and VC-1 acceleration (see
http://www.nvidia.com/docs/CP/11036/...Comparison.pdf ), they are unsupported?!?! What a shame!!!

[Thunderbird]

On old geforce6/7 and G80 models there was no dedicated hardware for decoding. The so called 'acceleration' were a bunch of shaders which accelerated a small part of the decoding. In case of h264 modern cards can accelerate all steps and in case for geforce9100/9200/9300 cards also VC-1. The vdpau API is made to accelerate all parts. Nvidia would have to offer a full software implementation of which parts can be accelerated in hardware. This is a lot of work and is very costly. I believe that on Windows all this preprocessing is done by DirectX and the driver only has to offer the parts it can accelerate.

[sl1pkn07]

Originally Posted by Thunderbird

On old geforce6/7 and G80 models there was no dedicated hardware for decoding. The so called 'acceleration' were a bunch of shaders which accelerated a small part of the decoding. In case of h264 modern cards can accelerate all steps and in case for geforce9100/9200/9300 cards also VC-1. The vdpau API is made to accelerate all parts. Nvidia would have to offer a full software implementation of which parts can be accelerated in hardware. This is a lot of work and is very costly. I believe that on Windows all this preprocessing is done by DirectX and the driver only has to offer the parts it can accelerate.

then change the specs of the cards. they give the option to confusion. offer things that do not have

[Thunderbird]

This is called marketing and ATI, S3 and others did the same. They offloaded some parts after the decoding to the gpu. The DirectX Video Acceleration architecture helped a lot for all. People are free to code their own acceleration using GLSL as that's basically what vendors did too.

[sl1pkn07]

DVXA meke inverse engineering? or nvidia (in this case) offer the especs for this feature...

OGL in this case has nothing to do. in this case exist XvMC/XvBA.

VIA and Intel has already made use of it, ATI/AMD is already starting to use it ... and NVIDIA?


inventing weird stuff (VUAPI) without being able to use the hardware ready for it (I remember that 6XXX/7XXX/G80 the series has implemented something: P). only make that users touch your balls (sorry for the expression)

[txf]

Originally Posted by sl1pkn07

DVXA meke inverse engineering? or nvidia (in this case) offer the especs for this feature...

OGL in this case has nothing to do. in this case exist XvMC/XvBA.

VIA and Intel has already made use of it, ATI/AMD is already starting to use it ... and NVIDIA?


inventing weird stuff (VUAPI) without being able to use the hardware ready for it (I remember that 6XXX/7XXX/G80 the series has implemented something: P). only make that users touch your balls (sorry for the expression)

6/7 series already use XvMC (for some time) 8 series and does not. XvMC only accelerates mpeg2 but it is an X standard so it is not nvidia's fault it doesn't do more. There are plans to extend XvMC to do more...

[sl1pkn07]

Originally Posted by txf

6/7 series already use XvMC (for some time) 8 series and does not. XvMC only accelerates mpeg2 but it is an X standard so it is not nvidia's fault it doesn't do more. There are plans to extend XvMC to do more...

Xvmc get support for H264/VC1, Nvidia no get support for XvMC

[txf]

Originally Posted by sl1pkn07

Xvmc get support for H264/VC1, Nvidia no get support for XvMC

I'm not sure what your complaint is. VP1 (i.e. purevideo 1) hardware can use XvMC, however XvMC is currently limited to mpg2, it's up to X devs to expand support of other codecs.

VP2 cannot use XvMC because it (xvmc) only does motion compensation (and inverse discrete cosine transform?). VP2 does decoding and postprocessing all in the same hardware (the gpu) hence why a specialized api is required to make use of it.

[sl1pkn07]

http://en.wikipedia.org/wiki/X-Video...n_Compensation

ok. no VC-1... but H264 (mpeg4 -avc) yes (for the moment. only uses for VIA unichrome

[txf]

Originally Posted by sl1pkn07

[url] (for the moment. only uses for VIA unichrome

I've yet to hear of anyone using this mythical via acceleration architecture for h264 (probably because so few people have one, and those that do probably don't care about these things).

The question then is, does nvidia want to expend resources providing acceleration to hardware architectures that already in the past and will never be resurrected in the future. By indication of the way they handle legacy drivers for unix (bugfixes only, no new features) it would appear that you're out of luck.

In the future though you might get it via Gallium3d and the Nouveau driver, but that is probably a long way off yet...

[Alejandro Nova]

As long as brand new computers (budget ones, but still brand new) keep selling with nForce 430/GeForce 6150 SE, or nForce 630i/GeForce 7050-GeFforce 7100, I wouldn't call the GeForce 6/7 family a thing of the past. Limited, yes, slow, yes, but definitely not a thing of the past.

[Alejandro Nova]

Originally Posted by txf

I've yet to hear of anyone using this mythical via acceleration architecture for h264 (probably because so few people have one, and those that do probably don't care about these things).

The question then is, does nvidia want to expend resources providing acceleration to hardware architectures that already in the past and will never be resurrected in the future. By indication of the way they handle legacy drivers for unix (bugfixes only, no new features) it would appear that you're out of luck.

In the future though you might get it via Gallium3d and the Nouveau driver, but that is probably a long way off yet...

I don't see that so far away... In fact, Nouveau has caught up with the binary drivers, at least in the framerate apartment, with some NV50 cards.