• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Killzone: Shadow Fall Multiplayer Runs at 960x1080 vertically interlaced

VanWinkle

Member
I absolutely agree. Those Sony fans who didn't hesitate to call out Sony and GG on the fact that they lied have earned my respect. Others clearly tried to find ways to downplay or obfuscate the issue. Thuway's post in particular is troubling, since it's coming from an insider who had a starring role in Resolutiongate. And I still wonder how not one of the insiders got wind of this for so many months. As I said, troubling.

Fair point.
 

Oersted

Member
The question, which bothers me, is, how did journalists miss this for so long?

That a company is lying, well yeah, but missed by so many?
 

RoboPlato

I'd be in the dick
The question, which bothers me, is, how did journalists miss this for so long?

That a company is lying, well yeah, but missed by so many?
Technique hasn't been used before so we didn't know what to look for and blamed the blurriness/artifacting on a poor AA method since we were told it was native 1080p and that the AA in MP was different than SP. In certain situations this method can actually trick pixel counting as well.

It seems that GG probably shared the info with Leadbetter but he left it out of his tech article on the game for some reason and just happened to mention it now.
 
There isn't a thing wrong with 1080i on a proper display built to handle it, but when the content being sent isn't prepped for it and the display it's being sent to isn't designed for it at all I guess you can have issues.
So is it like 1080i...but just vertically interlaced instead of horizontal?
No, the thread title is wrong. Killzone's MP is not interlaced, vertically or otherwise.

Honestly, I think directly scaling one dimension of the frame (most similar to the GT5 process) with reasonable AA results in fewer artifacts and arguably a cleaner image than the process for MP Shadowfall. We've also seen Ryse's 900p with custom scaling and AA give some of the better IQ of the launch-window games, at least of those that aren't native resolution.
Did you see dark10x's comparisons of the real technique and upscale from 720p a bit back? The upscale looks far, far worse. Unfortunately, we don't know how a 900p upscale (like Ryse) would compare, since there's no way to force the PS4 to output at that resolution. However, given what the "Guerilla method" screenshot looks like, I bet upscale still looks worse.

Even if it looked the same or a little better, they're not really equivalent to use. A 900p render saves 30%, whereas the Guerrilla method saves 50% minus calculation overhead, which I can't imagine is 20%. So Killzone's approach frees more resources while (probably) looking about as good. Which makes sense, because why else would they invent it?
 
No, the thread title is wrong. Killzone's MP is not interlaced, vertically or otherwise.


Did you see dark10x's comparisons of the real technique and upscale from 720p a bit back? The upscale looks far, far worse. Unfortunately, we don't know how a 900p upscale (like Ryse) would compare, since there's no way to force the PS4 to output at that resolution. However, given what the "Guerilla method" screenshot looks like, I bet upscale still looks worse.

Even if it looked the same or a little better, they're not really equivalent to use. A 900p render saves 30%, whereas the Guerrilla method saves 50% minus calculation overhead, which I can't imagine is 20%. So Killzone's approach frees more resources while (probably) looking about as good. Which makes sense, because why else would they invent it?

Your conjecture sounds rather baseless. Come on man, a higheramount of pixels looking worse then a lower amount? That is not how sampling works no matter how much temporal reprojection you have...
 

RayMaker

Banned
Holy shit, talk about moving the goal posts.

The argument went from "lolz ps4 owners didn't notice the lower resolution. Biased exposed!" To "Boasting something something" when the previous argument was disproven.
If trying to argue against "fanboys being fanboys", it helps not to act like one.

Call it moving goalposts if you will, but it was two sepreate points what I said i still believe .

and going on about bein a fanboy is a bit rich with someone with this under there name

''breakup letter to Xbox''
 
Sony's 1080p pimping and trash talking about MS looks hilarious now.

Thankfully i don't give a **** about console resolution and will still enjoy my PS4 when it finally gets a game worth playing.
 

AgentP

Thinks mods influence posters politics. Promoted to QAnon Editor.
Your conjecture sounds rather baseless. Come on man, a higheramount of pixels looking worse then a lower amount? That is not how sampling works no matter how much temporal reprojection you have...

You should give a talk at GDC and explain all this to GG and the world.
 
Your conjecture sounds rather baseless. Come on man, a higheramount of pixels looking worse then a lower amount? That is not how sampling works no matter how much temporal reprojection you have...

It depends entirely on the accuracy of that reprojection and the effect the scaling has on the 900p image. He wants to see evidence, you're the one who wants to draw baseless conclusions.
 

emko

Member
i don't understand isnt it still 1080p with data filled in from previous frame so its 1080p frame but not all new pixels? so this makes the game have 2x more fps right? if we turned it off it will still be 1080p pure but at 30fps. What's the big deal? why are people saying its lower then 720p. I see it makes movement a bit blurry but i rather have that then lower fps even thought the MP has very low fps in some situations. Is the lower then 720p thing coming from people thinking its like interlaced video?
 
Your conjecture sounds rather baseless. Come on man, a higheramount of pixels looking worse then a lower amount? That is not how sampling works no matter how much temporal reprojection you have...
Well, I certainly could be wrong, given that we don't know what a hypothetical 900p Killzone MP would look like. But my conjecture isn't entirely baseless, because we do know what the current MP looks like. Here's a direct feed:

ARA.png


That doesn't look considerably worse than typical 900p to me. Without a real comparison my estimate of "maybe the same, maybe slightly degraded" is of course subjective. Nevertheless, the image is nothing like what I would normally expect hearing of a 960x1080 render. I know reprojection isn't magic, and 900p might indeed be too many more pixels for the technique to make the gap negligible. But the results here are undeniably closer than any previous solution, which makes me want to see some true comparisons.
 

RoboPlato

I'd be in the dick
Tried MP again to check out the new maps. When the game is actually running at 60fps, which was the first couple minutes of the game it really does look like 1080p with some bad FXAA. After that, when more players joined and things got chaotic, the effect kind of falls apart. This technique could be useful for games that are locked at 60fps but unless they can do that I think it hits an awkward point where it hurts image quality too much and doesn't provide enough of a performance benefit in return.
 

Violater

Member
The game is so horribly disappointing that none of this resolution revelations even matter to me. Tried the MP again just now, and its just a lensflare ridden blurry slow paced boring shadow of what KZ2 was.
GG needs a wakeup call.
 
The game is so horribly disappointing that none of this resolution revelations even matter to me. Tried the MP again just now, and its just a lensflare ridden blurry slow paced boring shadow of what KZ2 was.
GG needs a wakeup call.
Single player yes, was the best in the series but muliplayer no.

What exactly is it about kz2 multiplayer that people liked? It is just a massive nade spam fest with far worse spawn camping then kz:sf ever had and the lag was awful. KZ2 had too many cramped maps as well and the unlock times were atrocious for anything. Gunplay is meaningless when everything else was garbage.
 

Friction

Member
This discussion still going. Wow.

Checked and nothing mentioned in the OP. But did anyone from GG commented on this? I saw this news on IGN a few days ago. So I'm pretty sure they know its out there already.
 
What exactly is it about kz2 multiplayer that people liked?
I can only speak for myself, but: Warzone. Spawn grenades. Salamun Market. Pyrrhus Rise. The M82. Server lists.

Spawn camping definitely could happen, and the smaller maps were a bit of a mess, but lag was your problem, not the game's. I only rarely had issues.
 

Haines

Banned
The game is so horribly disappointing that none of this resolution revelations even matter to me. Tried the MP again just now, and its just a lensflare ridden blurry slow paced boring shadow of what KZ2 was.
GG needs a wakeup call.

It was always interesting to me watching everyone hype this game up when it looked so bloody average.

Hell kz2 was good but still had game breaking flaws in the mp gameplayl
 

HTupolev

Member
Your conjecture sounds rather baseless. Come on man, a higheramount of pixels looking worse then a lower amount? That is not how sampling works no matter how much temporal reprojection you have...
Consider the situation where the scene is completely static. Obviously in this case even a garbage implementation of KZSFMP-esque temporal sampling is going to be able to achieve a perfect native 1080p image, whereas raw spatial 900p render will remain upscaled 900p.

Now consider the situation where the scene completely changes to something entirely new from one frame to the next. Obviously in this case temporal sampling is worthless, and so unless you're a MASSIVE fan of horizontal-only upscale, the 900p will probably yield a better image on account of having access to significantly more useful samples when generating the output.

In the real world, the former situation happens occasionally, with the rest of cases mostly being covered by something in-between.
It's certainly possible to run into problems with temporal samples, since you will run into coverage issues, and because shading on particular objects may change between frames, and because reprojection may itself not be a perfect process. But there's nothing intrinsically invalid about temporal samples that always makes them completely worthless as a substitute for spatial samples. The reason that temporal sampling is getting a lot of attention these days is that games usually DO have a lot of shading redundancy from one frame to the next, and the pixels from the previous frame can be used as extremely close approximations for information that exists in the coverage of the current frame.

So yes, a higher number of spatial samples absolutely will, in some cases, give inferior results to a lower number of spatial samples, if those spatial samples are backed up by snazzy temporal sampling.
 
Well, I certainly could be wrong, given that we don't know what a hypothetical 900p Killzone MP would look like. But my conjecture isn't entirely baseless, because we do know what the current MP looks like. Here's a direct feed:

ARA.png


That doesn't look considerably worse than typical 900p to me. Without a real comparison my estimate of "maybe the same, maybe slightly degraded" is of course subjective. Nevertheless, the image is nothing like what I would normally expect hearing of a 960x1080 render. I know reprojection isn't magic, and 900p might indeed be too many more pixels for the technique to make the gap negligible. But the results here are undeniably closer than any previous solution, which makes me want to see some true comparisons.
TBF, the difference isn't going to be noticeable in static shots like the above since, the difference between two frames is minimal to non-existent. So for still shots, it is a fine technique for producing crisp frames. Problems arise when there is motion and a more significant difference between frames.
 
Call it moving goalposts if you will, but it was two sepreate points what I said i still believe .

and going on about bein a fanboy is a bit rich with someone with this under there name

''breakup letter to Xbox''
If you read the article it'll make sense as it was a joke piece from a comedy website. Or you just could, you know, throw stones from your glass house.
 
I wish GG just came out with this info instead of misleading people. Sure, resolution might not matter all that much in reality, but this is still a product, and if someone bought that product because they wanted a nice shiny 1080p game to play, then they got scammed.
Sony trash talked about how MS looks (sic)?

Can you point me to them tash talking about MS looks hilarious(sic)?
Reading comprehension :/
 
TBF, the difference isn't going to be noticeable in static shots like the above since, the difference between two frames is minimal to non-existent. So for still shots, it is a fine technique for producing crisp frames. Problems arise when there is motion and a more significant difference between frames.

Probably why he/she emphasizes 60fps.
Should be an interesting tech to cheat your way to 120fps :p
 

JLeack

Banned
The game is so horribly disappointing that none of this resolution revelations even matter to me. Tried the MP again just now, and its just a lensflare ridden blurry slow paced boring shadow of what KZ2 was.
GG needs a wakeup call.

It's just... So slow.

Some people like that but I don't.
 

dark10x

Digital Foundry pixel pusher
TBF, the difference isn't going to be noticeable in static shots like the above since, the difference between two frames is minimal to non-existent. So for still shots, it is a fine technique for producing crisp frames. Problems arise when there is motion and a more significant difference between frames.
What you missed is that I took that shot in motion. That is not a still shot. It was snapped while rotating the camera at 50% speed. The other shot I posted was at 100%.
 

Norml

Member
Here's a shot taken while rotating the camera at a medium steady rate (taken from a capture card rather than using the built-in sharing feature).

Considering the technique being used, I'm impressed with how artifact free the image appears in motion. From what I can see, this type of rendering has the most noticeable impact on thin objects (such as fences) but even then it simply looks as if it is part of the camera blur.

ARA.png


Here's the same area taken with maximum camera rotation speed. Motion blur is in full effect. If you look at the metal flooring you can see increased aliasing with larger steps but the image still looks good. Without the excellent motion blur it would certainly be more obvious.

BRA.png


I'd love to know just how many resources this technique frees up. It could be a real alternative to traditional rendering that would allow for a higher framerate without the massive loss in image quality associated with lowering overall resolution (which really only looks bad as a result of scaling).

Also, just for fun, here's a shot with the PS4 set to output at 720p. The system is downscaling the image so jaggies are minimized compared to what you'd get with a traditional 720p image. When blown up to 1080p it looks dramatically worse than the 960x1080 method they used.

Click on the images to see them at full resolution.

DRA.png

News is a bummer, but also cool since the technique looks to be far better than typical scaling. So now we have those two and dynamic resolution.I wonder which really is the best?

APNG of your shots.
kkkkrssr1.png
 
What you missed is that I took that shot in motion. That is not a still shot. It was snapped while rotating the camera at 50% speed. The other shot I posted was at 100%.

My bad. What's odd is that other captures people have provided are not nearly as clean.
 

Estocolmo

Member
Xbox is mentioned so many times in this thread, even though the OP has absolutely nothing to do with the Xbox what so ever.
 

mrklaw

MrArseFace
I wish GG just came out with this info instead of misleading people. Sure, resolution might not matter all that much in reality, but this is still a product, and if someone bought that product because they wanted a nice shiny 1080p game to play, then they got scammed.

Reading comprehension :/

Considering how people in this thread are having difficulty concisely explaining what is happening, how do you expect GG to have done that in short press previews? The closest they've come is in the DF interview I think. With hindsight you can read it and see some elements of what they are doing, but at the time it would have been easy to interpret that as some form of temporal AA.

I think brad genz nailed it though - we simply have no simple description to cover what they are doing, and calling it interlaced is unfair to the technique and invites ignorant arguments from those reading it as literally interlacing
 
Considering how people in this thread are having difficulty concisely explaining what is happening, how do you expect GG to have done that in short press previews? The closest they've come is in the DF interview I think. With hindsight you can read it and see some elements of what they are doing, but at the time it would have been easy to interpret that as some form of temporal AA.

I think brad genz nailed it though - we simply have no simple description to cover what they are doing, and calling it interlaced is unfair to the technique and invites ignorant arguments from those reading it as literally interlacing
Yeah, fair point.
Why use that technique ?
PS4 is not powerful enough to render full 1080p60 ?
Now THAT is some news
No matter how powerful a system is, you can always design a game that will push it so far it can't do 60fps.
 

RulkezX

Member
I haven't played a whole lot of the MP so can't say I noticed anything off.

The SP is still a proper 1080/unlocked , right ?
 
My bad. What's odd is that other captures people have provided are not nearly as clean.
I think there are two major things causing that. First, once you get to a certain speed the actual motion blur kicks in. Some of the blurred-out textures seen in other shots may be partially due to that effect, rather than the reprojection technique.

The second point is that the reprojection provides different results depending on what's onscreen, how fast it's moving, etc. There are definitely times when more artifacts are visible than in dark10x's shot; it's not always clean. But the artifacts are literally 1 pixel wide and only onscreen for about 20 milliseconds before disappearing (possibly to reappear nearby). In most viewing situations this will make them very hard to spot, especially if the observer is following the game action rather than trying to look for graphical errors.This is why no one really noticed them before (despite the fact that they're quite obvious in stills once you know what you're looking for).

This method isn't magic. It does not look as good as native 1080p. But the degradation is surprisingly graceful, meaning the final output is far better than a naive upscale from the same number of rendered pixels. Put another way, it looks as good as a traditionally upscaled render of more pixels. (How many more is both uncertain and debatable.)
 

le.phat

Member
The game is so horribly disappointing that none of this resolution revelations even matter to me. Tried the MP again just now, and its just a lensflare ridden blurry slow paced boring shadow of what KZ2 was.
GG needs a wakeup call.

Co-signed on that wake up call. Guerilla needs a new pet because it feels like they no longer know how to improve on the franchise amd onstead just change stuff up and hipe it feels fresh.
 
Co-signed on that wake up call. Guerilla needs a new pet because it feels like they no longer know how to improve on the franchise amd onstead just change stuff up and hipe it feels fresh.

They just need to hire a writer with actual talent to flesh out the world/storyline. They are very talented coders/artists/sound designers/etc...but their ability to create compelling storylines and rich characters leaves much to be desired.
 
Top Bottom