| 1 |
On Sun, Apr 15, 2012 at 4:45 PM, Volker Armin Hemmann |
| 2 |
<volkerarmin@××××××××××.com> wrote: |
| 3 |
> Am Sonntag, 15. April 2012, 16:44:43 schrieb Florian Philipp: |
| 4 |
>> Am 15.04.2012 16:22, schrieb Michael Mol: |
| 5 |
>> > On Sun, Apr 15, 2012 at 10:09 AM, Florian Philipp <lists@×××××××××××.net> |
| 6 |
> wrote: |
| 7 |
>> >> Am 15.04.2012 15:18, schrieb Walter Dnes: |
| 8 |
>> >>> On Wed, Apr 11, 2012 at 06:30:02PM +0200, Volker Armin Hemmann wrote |
| 9 |
>> >>> |
| 10 |
>> >>>> Am Mittwoch, 11. April 2012, 02:11:35 schrieb Walter Dnes: |
| 11 |
>> >>>>> If it's PCIe, so be it. Actually, a post that prevents me wasting |
| 12 |
>> >>>>> |
| 13 |
>> >>>>> money is helpful <G>. Would PCIe be significantly better on the same |
| 14 |
>> >>>>> CPU+GPU, or is it hype? |
| 15 |
>> >>>> |
| 16 |
>> >>>> a lot, lot lot lot better. No hype. |
| 17 |
>> >>>> |
| 18 |
>> >>> I've done some looking, and I'm back with more questions. I've also |
| 19 |
>> >>> |
| 20 |
>> >>> read the Nouveau-versus-NVIDIA thread. Questions... |
| 21 |
>> >>> |
| 22 |
>> >>> 1) Will PCIe 2.0 cards work in a PCIe 1.0 slot? I'm not expecting 2.0 |
| 23 |
>> >>> performance, I just want full backwards compatability. PCIe 1.0 cards |
| 24 |
>> >>> seem to be rare, and have to be ordered online, while I can pick up a |
| 25 |
>> >>> 2.0 card locally at a store. |
| 26 |
>> >> |
| 27 |
>> >> PCIe-2.0 is fully downward compatible to 1.1 and 1.0. |
| 28 |
>> >> |
| 29 |
>> >>> 2) My main "torture test" will be HD fullscreen video. Will there be |
| 30 |
>> >>> major improvement in that? That's 2D. Forget 3D. |
| 31 |
>> >> |
| 32 |
>> >> 2D video is still rendered using OpenGL if your video player supports it. |
| 33 |
>> > |
| 34 |
>> > I'm not aware of any video decoders using CUDA, OpenCL, or pixel |
| 35 |
>> > shaders for video decoding; AFAIK, unless you're using VDPAU you're |
| 36 |
>> > still using the CPU to render the video to a frame buffer. The most a |
| 37 |
>> > video player is going to use OpenGL for is stretching that frame |
| 38 |
>> > buffer to fit a window or screen, and possibly as a compositor to |
| 39 |
>> > place overlays like subtitles or playback control elements.. |
| 40 |
>> |
| 41 |
>> Agreed. Decoding is still usually done in software but offloading |
| 42 |
>> scaling and YUV to RGB conversion helps none the less. Mplayer, for |
| 43 |
>> example, allows a lot of customization depending on the amount of |
| 44 |
>> texture units. With high resolution displays and slow CPUs, this can |
| 45 |
>> have surprisingly large effects. |
| 46 |
>> |
| 47 |
> |
| 48 |
> and with vlc you can use vaapi which can make use of the video decoding engine |
| 49 |
> of the graphic chip. |
| 50 |
> |
| 51 |
> If the movie is using the right codec, of course. |
| 52 |
|
| 53 |
Also depends on whether or not the graphics driver and vaapi like each |
| 54 |
other. I'm not aware of NVidia cards supporting VA API yet. |
| 55 |
|
| 56 |
VA API is pretty new; it'll be interesting to see where it goes, and I |
| 57 |
hope it takes hold. Right now, the most tested and working solutions, |
| 58 |
AFAIK, are nVidia cards and VDPAU. At least, that combination has been |
| 59 |
working well for me for 3-4 years. |
| 60 |
|
| 61 |
-- |
| 62 |
:wq |
Archives