View Single Post
Old 4th December 2012, 05:25   #15943  |  Link
6233638
Registered User
 
Join Date: Apr 2009
Posts: 1,019
Quote:
Originally Posted by ajp_anton View Post
HD 3000 at 1700MHz, is that close enough to HD 4000?
1920x1080 -> 2560x1440
Default (lanc3 AR / bilinear) renders at around 41ms.
Thanks. That seems like Lanczos 3 AR Luma, Bilinear Chroma is pushing it, and you're likely to experience frame drops. (or just avoid them)

I'd really need to have one myself to use the same material and settings to do a proper comparison/test though.

I would be interested in seeing what your numbers are for Bicubic 75 Chroma & Lanczos 3 without AR for Luma with the same material though. I really think that using Bilinear for Chroma is too much of a compromise, even if it would let you use Lanczos 3 AR.

In the test I posted above, those settings almost halved rendering times - as good as it can look, I think the anti-ringing filter is probably too demanding to have enabled as a default.

Quote:
Originally Posted by iamhacked View Post
Nice point. I have a Dimension E521 (AMD Athlon™ 64 X2, 1GB RAM, Nvidia GeForce 6150LE) and using Bicubic 50 for chroma upscaling does look better than Bilinear. However, any number above 50 will cause 720p video to stutter (using LAV filters). Using Bicubic for luma/image upscaling will cause video to stutter as well. Bilinear seems to be the only option for luma upscaling to me.
That's interesting, I was under the impression that all the Bicubic variations (Mitchell-Netravali, Catmull-Rom, Bicubic, and SoftCubic) were all the same algorithm with adjusted values, that wouldn't change the load they put on the GPU. I certainly don't see a meaningful change between them. (less than 0.5ms)



EDIT: And from doing some extra testing, at least with some scale factors (I don't know if it will change dynamically) Nvidia basically using Bilinear scaling with the DXVA2 option. (results are slightly different) So it's even worse than I thought, considering that with madVR's Bilinear Luma scaling I was getting render times of about 3ms compared to the 50ms+ of DXVA2.

EDIT2: Actually, it's worse than that - if you use DXVA2 for Luma upscaling, Chroma upscaling is basically ignored. Unless DXVA2 is handled a lot better on AMD/Intel, I wonder if it should actually be removed.

It's maybe not identical to the bilinear option in madVR, but the results are still equally bad.The DXVA2 results are a little smoother when compared to Bilinear in madVR (is chroma being filtered?) but you can see that there's almost no difference between Bilinear Chroma and Lanczos 8 AR when you use DXVA2 scaling.

Render times are considerably higher with DXVA2 scaling as well.

Last edited by 6233638; 4th December 2012 at 06:14.
6233638 is offline   Reply With Quote