AMD FSR antialiasing discussion

  • Thread starter Deleted member 90741
  • Start date
Cat Merc said:
Funny thing, I managed to set the slider to "100%", but it allowed me to enable FSR, so it's probably >99.5% at that setting.

I did some comparison shots and FSR has interesting behavior.

I have a twitter post going with a poll asking people to guess which image is which. I'm guessing the people here will be able to tell, as I noticed the more you know about rendering the more likely you are to figure out which one is using FSR:
https://twitter.com/x/status/1407941530673893376

I ask the people here to kindly not vote as I'm interested in what people who aren't from here think :)
Click to expand...
Which is why you should use CAS or some other sharpening on native res when comparing that to FSR.
 
Blooper said:
but looking at the neck areas on the side by side there it looks to me there's intentional blurring applied on the 1080p upscale and FSR images that's totally missing on the TAAU image.
Click to expand...
Post processing heavily depends on resolution. Higher resolutions require sampling larger windows of pixels to achieve the same filter radius as in lower resolutions hence the difference in DOF. it's still here with TAAU, it's just that DOF shader samples different windows of pixels hence the different amount of blurring.
 
DegustatoR said:
Which is why you should use CAS or some other sharpening on native res when comparing that to FSR.
Click to expand...
I find this also a reasonable thing as FSR is post-sharpening and creating local contrast in a heavy-handed way - equalising that with the native or other comparison shots when searching for "quality" seems reasonable.
 
Dictator said:
I find this also a reasonable thing as FSR is post-sharpening and creating local contrast in a heavy-handed way - equalising that with the native or other comparison shots when searching for "quality" seems reasonable.
Click to expand...
It's reasonable because FSR has (R)CAS as a part of it's algorithm. Thus when you do something like 99.9%->100% upscaling with it as in the example shown you're basically doing sharpening.
 
DegustatoR said:
Which is why you should use CAS or some other sharpening on native res when comparing that to FSR.
Click to expand...
Look at the character model. FSR successfully smoothed it to look far higher res. CAS wouldn't achieve this result.

Most obvious at the bottom of the character, the pants.
 
Cat Merc said:
Look at the character model. FSR successfully smoothed it to look far higher res. CAS wouldn't achieve this result.

Most obvious at the bottom of the character, the pants.
Click to expand...
Sure as it will still do its "reconstruction".
But the sharperning part you can equalize.
And I'd also expect the upscaling shader to make it into Reshade - so you'll be able to run it on native res in any game.
Results should be pretty equal at this point.
 
DegustatoR said:
Sure as it will still do its "reconstruction".
But the sharperning part you can equalize.
And I'd also expect the upscaling shader to make it into Reshade - so you'll be able to run it on native res in any game.
Results should be pretty equal at this point.
Click to expand...
There's a good chance post process effects mess with it too much and it will produce much inferior results as a reshade package.

As for the sharpening, FSR is surprisingly smart about it. HUB's video shows them trying to recreate it, and whenever they managed to match the sharpening on one part of the scene, another looked undersharpened or oversharpened compared to FSR.
 
As a side note about Riftbreaker, the blurring seen on the surface of textures with FSR on disappears when I disable TXAA from the ini. The surface of the texture remains crisp, though obviously without any TAA the image looks awful in motion.

Interesting interaction.
 
OlegSH said:
Post processing heavily depends on resolution. Higher resolutions require sampling larger windows of pixels to achieve the same filter radius as in lower resolutions hence the difference in DOF. it's still here with TAAU, it's just that DOF shader samples different windows of pixels hence the different amount of blurring.
Click to expand...

I'm not sure I follow. All three of those images are the same internal resolution of 1080p and all are upscaling to 4K, there should still be no differences in DoF.
 
Cat Merc said:
As a side note about Riftbreaker, the blurring seen on the surface of textures with FSR on disappears when I disable TXAA from the ini. The surface of the texture remains crisp, though obviously without any TAA the image looks awful in motion.

Interesting interaction.
Click to expand...
In what ways does FSR edge detection break without AA?

I would guess it might look like the Ps3 upscaler. (Works half decently on near diagonal edges, less so at pixel grid aligned areas.
 
jlippo said:
As they may be when comparing 1080p to 4k.
It's quite strange that sometimes camera fx like DoF are not tied to degrees in view, but to resolution of an image. (Or in some cases something is bugged with some features. ;))
Click to expand...

It can also be due to assumptions about screen size. At lower resolutions and smaller screen sizes you would want to blur less of the view.
 
You must log in or register to reply here.

Similar threads

Cyan
Microsoft's Auto SR: Universal AI upscaling discussion.
Replies
7
Views
2K
Janne Kylliö
J
B
DLSS versus FSR, again, neither being PSSR *spawn
2 3
Replies
41
Views
4K
homerdog
homerdog
Unknown Soldier
3DMark AMD FSR Feature Test
Replies
1
Views
1K
Putas
P
BRiT
Intel XeSS anti-aliasing discussion
9 10 11
Replies
200
Views
36K
Shifty Geezer
Shifty Geezer
M
Proposed upscaling technique - render vectors to pixels, convert to vectors, upscale and re-raster *spawn
2 3
Replies
43
Views
4K
JoeJ
J
Back
Top