Nvidia DLSS 1 and 2 antialiasing discussion *spawn*

DLSS quality in the Matrix Awakens City Sample looked pretty bad and here's the possible cause that I investigated.

1. Coarse jitter sample pattern.
This makes the scene look blurry. But not a major problem I think.

UE5 City Sample:
UUwKmPm.png


DLSS SDK Sample (good case)
fJkKqT6.png



2. Bad motion vectors. I mean, REALLY BAD especially in thin objects.
Cross-patterned 5 pixels in the air are all birds. You'll be noticed that birds are barely visible although they are flying around the city all time.
VDo1BZt.png


The motion vectors of the thin branches are very chunky and dirty. Compare with actual geometry.
xaE4Pzw.jpg

nOFYgLZ.jpg


This causes the branches to appear disconnected and blink while moving camera.
4YkWmn8.jpg


Also people walking far around have chunky motion vectors too. They always moving with the ghosting if you look closely.
MGAI6VG.png


This is what good motion vector should look like:
gEpn9n4.png



So, who should address this problem?

Nvm, I've found the option for disable motion vector dilation and tested it. There wasn't meaningful changes in reconstruction quality. :cry:
 
Thank you @Flappy Pannus for the link! Not quite there yet, is it?

This video is showing GPU load in native at 83%. DLSS obviously won't do anything (besides changing or lowering image quality) to a render which isn't GPU limited to begin with.
Make sure to read the YouTuber's comments; the comparison is TAA + render upscale in the game. As such, there is no "native" comparison in the video.
 
This video is showing GPU load in native at 83%. DLSS obviously won't do anything (besides changing or lowering image quality) to a render which isn't GPU limited to begin with.
'Load' has nothing to do with it, I'm not posting it for performance critique, rather the severe shimmering which blatantly stands out in the timestamp. That TAAU looks far superior (even at 50% scaling!) is indication something is broken with DLSS, at least on his system.
 
'Load' has nothing to do with it, I'm not posting it for performance critique, rather the severe shimmering which blatantly stands out in the timestamp. That TAAU looks far superior (even at 50% scaling!) is indication something is broken with DLSS, at least on his system.
Ah, well, it's a beta for a reason.
 
It doesn't matter. If you're not GPU limited you won't see any benefits from DLSS.
While your statement is correct, the video as-presented gives you no way to make a logical statement of GPU load at native resolution -- because no native resolution was recorded.
 
While your statement is correct, the video as-presented gives you no way to make a logical statement of GPU load at native resolution -- because no native resolution was recorded.
83% GPU load points to GPU not being fully utilized which means that you won't see performance benefits from using DLSS. Native or not.
 
We don't really know what part of the gpu is fully busy or not with a single load %age. For all we know, dlss can help even in a <100% load.
 
83% GPU load points to GPU not being fully utilized
I'm going to stop you right there. You're seeing 83% GPU load while an upscaler is being used -- not native resolution. I'm not sure how much more clearly I can point this out...

So, again, you have no native raster resolution data to compare to and thus you have no data to say the GPU wasn't being fully utilized when running at native resolution.
 
I'm going to stop you right there. You're seeing 83% GPU load while an upscaler is being used -- not native resolution. I'm not sure how much more clearly I can point this out...
So? I was talking about performance comparison. With GPU being at 83% you can't compare performance because the render isn't running GPU limited. That was my point.

So, again, you have no native raster resolution data to compare to and thus you have no data to say the GPU wasn't being fully utilized when running at native resolution.
Again: it doesn't really matter how it's running in native resolution for discussion of performance with DLSS.
 
Back
Top