Quote:
Originally Posted by madshi
I haven't really done benchmark tests for these things, but obviously running those algos multiples times costs performance. My future plan is to try to modify/tune the sharpening algos to only have to run once after upscaling, but still achieve a good effect, regardless of the upscaling factor. Currently if the upscaling factor is very large, most of the sharpeners (except SuperRes) lose their effectiveness.
|
But why would it cost so much more when run twice with NNEDI3 compared to super-xbr? AdaptiveSharpen is running the same number of times and at the same resolutions in both cases but the performance hit is 2600% greater when using NNEDI3 quadrupling. I am using NNEDI3 256 doubling in both cases.
I agree, AdaptiveSharpen is more effective when run twice in this situation. Looking forward to the tweaks for large scaling factors.
Quote:
Originally Posted by baii
What is the initial video size? I recently went up to 4k and any refinement got more expensive as it is done @4k. If you use every 2x, would it be possible that those refinement was done at high res then downscale to wqhd( was it you with wqhd and 980ti iirc?)
Sent from my 306SH
|
The source res was 720x480 (anamorphic 16:9). However notice the difference between super-xbr and NNEDI3. I get only one extra ms from running AdaptiveSharpen twice (once at 1706x960 and once at 3412x1920 instead of only once at 3412x1920) using super-xbr quadrupling but I get an extra 26ms when running AdaptiveSharpen twice using NNEDI3 quadrupling.
Running AdaptiveSharpen only at the quadrupled res of 3412x1920 costs ~1.4ms using either NNEDI3 or super-xbr quadrupling.
Yes, I am running WQHD with a 980Ti.