Image Quality and Framebuffer Speculations for WIP/alpha/beta/E3 games *Read the first post*

6lG8yAL.png


Lego Movie:

Decided to look for something a bit more solid (more steps/pixels). Again, it's just a guess/arbitrary count, but it looks like it's around 1280 for the vertical axis (64/54).

1200 would have given 60/54, mind you, let alone 20/18, which I kept finding to be 21/18, and so... another <unspecified time> investigating.

Kind of weird, but not an unfamiliar number, exactly 1.5:1 anamorphic. At least the FOV isn't affected due to it, which does say something about the engine.

---
 
AlNets said:
Lego Movie:

Decided to look for something a bit more solid (more steps/pixels). Again, it's just a guess/arbitrary count, but it looks like it's around 1280 for the vertical axis (64/54).

1200 would have given 60/54, mind you, let alone 20/18, which I kept finding to be 21/18, and so... another <unspecified time> investigating.

Kind of weird, but not an unfamiliar number, exactly 1.5:1 anamorphic. At least the FOV isn't affected due to it, which does say something about the engine.

---
Click to expand...
Can i ask you how you can count more steps than the maximum of the resolution of the image? I mean, if the native resolution of the screenshot is below than the res of the image i get it but how it works in the other way?
 
Jcustom said:
Can i ask you how you can count more steps than the maximum of the resolution of the image? I mean, if the native resolution of the screenshot is below than the res of the image i get it but how it works in the other way?
Click to expand...

It's tricky, but you're basically looking for a leading edge. Some steps have the erm, subsample, and others do not. It kind of just... works for non-integer scaling ( in which case all steps get extra sub-sample shades, and makes it harder although we know how rotated grid MSAA generally looks. It gets more complicated as you go up from there, and you have to look at other factors in image quality).

It's still guesswork, of course. :)
 
AlNets said:
It's tricky, but you're basically looking for a leading edge. Some steps have the erm, subsample, and others do not. It kind of just... works for non-integer scaling ( in which case all steps get extra sub-sample shades, and makes it harder although we know how rotated grid MSAA generally looks. It gets more complicated as you go up from there, and you have to look at other factors in image quality).

It's still guesswork, of course. :)
Click to expand...

Thanks, looks tricky indeed :smile:
I've tried to see some images of Lego Movie PS4 from DF but i'm not quite sure how to check the right edge when the post-AA kicks in.

DF has just updated the article ;)
 
I have misjudged the AA solution in MGS5. I previously thought it was a "blurry FXAA" when in fact it's definitely not. Sorry Mr. Kojima!

The AA is a lot better than I thought. We can see some artifacts in some areas (the wire transformed into a dotted line, the pattern on the fences, the strange stuff on top of the rocks top-left on this pic).

But the important geometry sub-details are preserved (surprisingly well, in fact notably on the tower) and the textures have still some sharpness in them (ground textures with sun reflection are definitely not wrecked by a strong FXAA):

13162705685_20c82e7e19_o.png


Maybe the pic I first saw had a strong motion blur in effect during this moment (it was during some stealth kill).

But it does look great and clean enough for a cross-gen game.
 
Seems likely the funky dithering is just their LOD system (fade in objects with view distance/object size).
 
I remember you said wait for some better screens because you saw some odd artifacts from the compressed ones a while back. So is it native Al?
 
Last edited by a moderator:
Looks that way. I haven't seen it in motion, but it was probably just some shoddy capture from a 1080i recording (it happens) for those anomalous shots.

--

Heh, right hand in third shot shows a very tiny amount of typical temporal ghosting issue. Also on that piece of paper to his left.

Actually, quite a few edges get missed by the edge detection here, probably too low contrast (hard to notice).

---
 
Call of Duty: Advanced Warfare (Xbox One)

booozp.png


kljozf.png


Not a good capture, but it sure doesn´t look like 1080p, right?
 
Last edited by a moderator:
I'm a very poor pixel counter, but I think I can safely say it's less than 1080P. Beyond that I wouldn't trust me much.

For a horiontal rez, I'd guess maybe 3/4 ratio, which should be around 1440, but that's kind of an odd resolution. And again, dont take me to know much, but it seems safe to say it's less than 1080P.

Could be 900P or even 720P I think.

Edit: Ughh, on a better measurement I think it's maybe ~720P. I'm pretty solid on 720P. Of course if it's some oddball resolution like 780P, I wont be able to suss that out, but it looks "around" 720p.

Edit 2: and that's on that particular shot. A lot of you have sharper eyes than me, and Globalisateur seems to think it might be a dynamic resolution, or that different parts of the trailer may sport different resolutions (different versions even?)
 
Call of Duty: Advanced Warfare (Xbox One)

I also find roughly a 4/3 ratio for both vertical and horizontal resolution which is exactly 810p.

A very long vertical edge can be found in the first image between the window and the book shelf. For horizontal resolution it's more tricky but we can thanks the FXAA naughty artefacting for giving us several long horizontal edges on the white truck here:

http://postimg.org/image/4tdrbk46r/full/

We can safely presume the native resolution (of gameplay-ish scenes) is approximately 800p.

But it's maybe exactly 810p!! After the Titanfall 793p resolution everything is possible!

EDIT: Horizontal resolution in a Kevin Spacey scene has a more precise 31/40 ratio with very detailed edges:

http://postimg.org/image/4j68s7pkj/full/
http://postimg.org/image/v2ytucq43/

Which is exactly 837p. Statix you may be right with your 840p number.
 
Last edited by a moderator:
On one of the screens posted on NeoGAF, I was getting ~840p pretty consistently. Could be dynamic resolution though, because some parts of the video are definitely blurrier than others.

ild9MH8ucmJgG.jpg


Edit: I think Globalisateur has the resolution pegged, at least for some scenes. 810p looks about right.
 
Last edited by a moderator:
Shortbread said:
So, we’re currently looking at 840-900p for cutscenes and mostly 840p gameplay? Maybe when everything is said and done within 5-6 months, 900p across the board.
Click to expand...

I would not take Starx's word, as I do not know him/her or their background.

The other 2 I would trust to get pretty close. So maybe 'low 800-ish p' is a ballpark until there is an official consensus.
 
KKRT said:
No, T1x TAA is much superior to T2x.
Click to expand...
I quickly looked & they I think they dont say that at all

I dont think they even mention T2x TAA, do they?

They mention smaa t2x though perhaps youre getting confused.

The general rule = with the same method the higher the samples the better the quality but at slower performance
 
You must log in or register to reply here.

Similar threads

X
Shimmering lines, common in menus, HUDs: Term for the artifact? And possible to "fix" with post-processing?
2
Replies
26
Views
3K
Davros
D
A
Machine Learning to enhance game image quality
2
Replies
27
Views
11K
Alucardx23
A
S
Modern textureless deferred rendering techniques
2
Replies
27
Views
62K
upnorthsox
U
S
Object/texture space lighting
Replies
2
Views
8K
sebbbi
S
P
Portable console for all PC games - PGS
2
Replies
31
Views
13K
Cyan
Cyan
Back
Top