On Sun, 16 Jan 2011 10:46:27 -0800 VDR User user.vdr@gmail.com wrote:
On Sun, Jan 16, 2011 at 10:22 AM, Tony Houghton h@realh.co.uk wrote:
Maybe a better idea is to not assume anything at all, but rather actually look up real life data or just buy one and see for yourself (as I did). There's no reason to take guesses about any of this stuff, plenty of users have posts their results and specs at various forums. A good place to start would be nvnews.net and read the thread "VDPAU testing tool".
The results don't give the right information to determine how well a card can handle 1080i.
You apparently don't know the results come from analyzing actual playback of actual samples of actual content. Yes, the data tells you exactly what kind of performance you can expect since it's generated from actual use cases. Again, stop assuming everything and turning your nose up at first-hand experience. I've ran those tests myself, obviously know what deinterlacers I'm using, and have watched plenty of content seeing the result with my own eyes from the hardware we're talking about. Additionally I've done the same with various hardware configurations.. What you're telling people simply doesn't agree with reality.
I've attached the results of qvdpautest on my desktop PC. Some of the examples appeared to have no more than 2 or 3 frames. Does the test generate a 'realistic' stream using the same few source frames over and over again? Even if it does, it seems a rather narrow sample.
The MIXER results show unrealistically high fps. Evidently the deinterlacing is not being performed at the same time as decoding in these tests. I suppose it's easy enough to caclulate the frame rate of both operations combined for a worst case, but how do you know to what extent they can be performed in parallel?