What's the difference between Copyback and Native?

[Snappy Phoenix]

DVXA Copy back or Native

What is the difference?

Some details please

My system specs:

Lenovo IdeaPad Y570-2349
• Intel Core i7 2670QM @ 2.20 / 3.10 GHz.
• 8 GB Hyundai DDR3 1333 MHz. RAM
• Kingston KC100 120GB SSD SATA-III + Toshiba THNSNB064GMCJ 64GB SSD SATA-II
• nVIDIA GeForce GT 555M 2GB DDR5 VRAM + Intel HD Graphics 3000
• 15.6' HD LED Screen (1366x768)
• Windows 7 Ultimate (x64)

[LoRd_MuldeR]

I would assume that "copy back" means that the decompressed frames are copied back to main memory after they have been decompressed by the graphic card's hardware decoder.

At the same time "native" probably means the default behavior of DXVA, i.e. present the frames right away without copying them back. That's what you would want for "playback only".

[heyer]

I would assume that "copy back" means that the decompressed frames are copied back to main memory after they have been decompressed by the graphic card's hardware decoder.

At the same time "native" probably means the default behavior of DXVA, i.e. present the frames right away without copying them back. That's what you would want for "playback only".

This is exactly what the difference is. For postproccesing like sharpening and avisynth, you need copyback mode. I recall, that once you couldn't use subs with normal DXVA. I do not know if this is still true.

[CruNcher]

Not correct for postprocessing you can use Shaders and that's very common since NT 6 to do mostly everywhere every GFX Vendors Driver has some PP implemented this way these days in their Driver stacks ;)
And it is a trend to add more 3D Engine things these ways also for Direct Manipulating Applications without needing to change them directly integrating AA and other interesting near Realtime stuff via Shaders :)
And some clever guys create crazy things with these possibilities making it even possible to almost paint into any Engine these days and be very creative, it's a very cool development. In Movie Production there is the term to "Paint with Light" many people today are already "Paint with Shaders" ;) :)
And seeing those things going on as you correctly realized is easy a normal Print Screen wont capture it as it isn't reaching the Ram @ all witch makes it also so efficient :)
Subs with DXVA is very well possible though most implementations aren't doing it well from a efficiency point it's the only correct way to do it under Windows for Playback, we are also in a situation where many things become mixed and most aren't aware of the sheer potential this is bringing combining DSP, Shader + multiple GPUs), CPU efficiently together into Workflows (especially as we get more and more DSP stuff for free), though there are also still a lot of issues in efficiency to handle better Driver models for example which are evolving more into these preemptive multitasking direction, latency and so on (which will be enhanced another time in Win 8) :)
The efficient combination of this allows very efficient Workflows @ Low Power people dreamed of years ago see for example http://www.mediafire.com/?0ksko7hdcdn29u6 where i combined some of these things "Realtime". Though it is taking place in every part of the System not only Video and 3D also Web Browsing HTML5 benefits these days i left this out in the Demo though due to limited Space ;)

And for many things these days in Video you wont need Pre Rendered Processing anymore because you can do it efficiently in "Realtime" and it evolves constantly so to get the same result on another system you wont need to render it out @ all just share your edit script (embed them as metadata) you fixed everything with and their you go no no need to waste hours on rendering anymore but instant results that are changeable at anytime as you wish (maximum flexibility) :) .

[Snappy Phoenix]

Thanks a lot for the info guys