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

From my playthrough of the beta, only sun was casting shadows and most shadows in environments were baked into textures, so it shouldnt be a big problem overall.

Still cant understand why resolution is so low on Xbone.
 
From my playthrough of the beta, only sun was casting shadows and most shadows in environments were baked into textures, so it shouldnt be a big problem overall.

Still cant understand why resolution is so low on Xbone.

I was thinking that it was because it was a forward renderer and that they rewrote the renderer to be able to draw many more things onto the screen - the additional load now must account for medium -> high amount of overdraw? If it were true, then perhaps 16 ROPS may not be enough.
 
I really wish we lived in a parallel Bioshock Universe where FXAA hadn't been invented.

In this Universe Timothy Lattes would have directly invented the already refined SMAA instead.

But in our FXAA Universe I guess we'll definitely have to endure on another whole generation of consoles. Confirmed.
 
First HQ (kinda) direct feed screenshot of Infamous SS I've found, courtesy of Neogaf:

ij0n5YfNkMfX6.jpg


The games seems fairly clean.
 
(About the 2 last Infamous pics) And most importantly, they've been smart and light with the post effect AA. In the first image I can even find some (rare) instances of low chrominance jaggies not taken care of by the AA which is a good sign. On the second image you can clearly see the detailed (non-blurred) sub-details geometry of the grass and the fences even with the compression.

I am completely reassured by those recent Infamous screenshots about the image quality. :cool:

This smart level of post AA reminds me Uncharted 3 and TLOU (in which I could notice some rare instances of non-blurred low chrominance jaggies). Maybe they use the same or similar AA solution.

On the other hand I have been completely disappointed by the strong FXAA (Thief vaseline level) they added in the retail version of MGS5 (video captured by a NeoGaffer, low PNG compression):

13069482435_5bd1fba52e_o.png


Just compare this blurry version with this previous build (same assets, no vaseline):

mgsv_1080p.jpg


Who still wants to argue with me that a strong FXAA is a good AA solution? ;)
 
This one made me say wow. (credit GAF)
edit: and that's with the share button.

1655456_10203498791089558_1214241031_o.jpg

lol could you imagine if this console generation followed all other console generations and released with a GPU equivalent to a R290?

The game is already so impressive running basically off a 5870, I can't imagine a game being designed specifically for an R290x.
 
Taken from MGS5 thread:

What kind of post AA creates this kind of artifacts? I thought it was FXAA (by the general level of vaseline) but now I have a doubt...

MGS5_cropped_cable.png

MGS5_cropped_fences.png
 
What kind of post AA creates this kind of artifacts? I thought it was FXAA (by the general level of vaseline) but now I have a doubt...

MGS5_cropped_cable.png

MGS5_cropped_fences.png
If you're referring to the funky patterns on the fence, that's likely just a result of insufficient sampling combined with the method of drawing those transparencies, and might not have much to do with post-AA.

Of course, you could always cast what I just said as "it's a problem with ALL post-AA, because supersampling would deal with this just fine!"
 
If you're referring to the funky patterns on the fence, that's likely just a result of insufficient sampling combined with the method of drawing those transparencies, and might not have much to do with post-AA.

Of course, you could always cast what I just said as "it's a problem with ALL post-AA, because supersampling would deal with this just fine!"

Ok for the fence it could come from bad sampling.

But about the wire? does FXAA normally modify a power line into a dotted line?

MGS5_cropped_cable.png



/wash my eyes with Infamous SMAA quality screenshots :mrgreen:
 
Any aliasing minded people out there : Is there some crippling drawback to SMAA Tx based solutions?

Ryse uses T1 and Infamous uses T2...and outside of MAYBE crysis trilogy(not sure about that), these are the only two console games to ever use SMAA. This is the case even though apparently according to graphs i found, SMAA T1 and 2 are the second lowest performance sappers of any aliasing solutions out there(with FXAA being the first obviously) while standing up very favorably to MSAA in terms of quality. It seems like SMAA should be the standard, yet FXAA is used much more....why is that?

If there isn't much drawback, i'd like to see SMAA T1 and T2 become standard this gen and not FXAA
 
Any aliasing minded people out there : Is there some crippling drawback to SMAA Tx based solutions?

Ryse uses T1 and Infamous uses T2...and outside of MAYBE crysis trilogy(not sure about that), these are the only two console games to ever use SMAA. This is the case even though apparently according to graphs i found, SMAA T1 and 2 are the second lowest performance sappers of any aliasing solutions out there(with FXAA being the first obviously) while standing up very favorably to MSAA in terms of quality. It seems like SMAA should be the standard, yet FXAA is used much more....why is that?

If there isn't much drawback, i'd like to see SMAA T1 and T2 become standard this gen and not FXAA

I would also like to know why. One reason could be ignorance, most don't know this effective and cheap AA even exists. SMAA is fairly recent and only used in a few games.

The only drawback is that it's a bit more gpu expensive than FXAA, but really it brings so much in the image quality category that it's worth removing a few shaders or details to gain the SMAA visual quality. Anyway FXAA will wreck any polygons (and the high resolution shaders/textures) you will add on the 3D scene...

With a 1tflops card on a 1080p image a decent FXAA costs around 0.7ms, SMAA 1x will cost 1ms so it's like 40% more expensive than FXAA. We're maybe talking on a PS4 on a 30fps game about 1 fps max more expensive than FXAA. On a 60fps game, we're maybe talking about 2fps more expensive than with FXAA.

Remember than even the blurry FXAA is not free! People easily tend to forget that.

SMAA_excerpt.png


With the superior temporal reprojection AA (which really makes the AA in another category than FXAA), SMAA 2tx costs 1.32ms, 88% more than a decent FXAA. But in reality it's just like 1 or 2 fps more expensive than SMAA 1x on a 60fps game.

All those "fps costs" are just rough approximations I have done from several diagrams I have found on the SMAA reference and PC AAs comparisons. You are free to correct me.

I think Ryse (SMAA 1tx) may use a less advanced temporal reprojection than SMAA 2tx, Crytek themselves did annouce a new version called "1tx" just used in Ryse but still with temporal AA but... 1 < 2, right?
 
Seems that way. I don't think 1 or 2 fps even really matters if that's all you save in comparison to a significantly lower quality solution(FXAA). Hopefully SMAA becomes looked at much more widely in the future.

Thank you for the information.
 
lol could you imagine if this console generation followed all other console generations and released with a GPU equivalent to a R290?

The game is already so impressive running basically off a 5870, I can't imagine a game being designed specifically for an R290x.

The GPU in the PS4 is far, far more powerful than a 5870...
 
Yeah, different architectures. The PS4's GPU is generations ahead.of that GPU. A 7850 would handily beat a 5870 in performance even though its flop count is much much higher, and the PS4's gpu is actually marginally more powerful than a 7850.

What's needed now is for developers to make sure have engines that can fully take advantage of multi core CPU's. The six jaguar cores are all necessary to take full advantage of the system
 
Back
Top