Oh I agree the game is great fun, but there was a noticeable IQ drop between MP and SP and its interesting to find out the cause of this, even if it's a bit disappointing that we have been mislead.
-
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
- Thread starter RoboPlato
- Start date
Oh I agree the game is great fun, but there was a noticeable IQ drop between MP and SP and its interesting to find out the cause of this, even if it's a bit disappointing that we have been mislead.
The fangs are out.
Someone here seems to have an agenda...
Sword Of Doom
Member
Dice has their own issues to deal with so I don't think that's a very good example. And please add players warping all over the place as another problem with the multiplayer. It's fun to just stand and watch everyone warp around the map.
Shouldn't have to explain anything to him. How a man can love a company so much I will never understand
That image looks better when its downscaled/downsampled, I wonder if this technique can be used to;
.render at more than 1920x1080 resolution
..utilize this temporal rendering method to render data for each frame at half the resolution
..put the final frame at their respective resolution
.downsample to 1920x1080 onscreen
boredandlazy
Member
18 short!
Sony-Wan: See it's 1080p for single player. So what I told you was true, from a certain point of view. Please Understand.
Liabe Brave
Member
A lot of the same assets are used in multiplayer as in the campaign--places, soldiers, weapons, etc. I think it's already really clear that this method looks noticeably worse than the same assets running in native 1080p.
A better comparison would be to see the multiplayer footage simply upscaled from 763p, without this new method. Because that's the alternative they didn't take. Going by simple logic and by the appearance of regular upscaled games, I bet it would look a lot worse.
If Sony fucking up is releasing a game as awesome as killzone, then they are okay in my books.
What about Thief? Last I checked, Thief runs and looks far superior on the PS4 than the Xbox One.
Also, "Sony GAF" will be fine so long as the sales numbers keep going the way they are.
Either way, one needs a before/after done where everything is the same except the technique discussed here. That way it's isolated and then only the impact of said technique can be analyzed.
SP and MP parts between games can be quite different despite using the same assets. Some games even load a different variation of their engine.
Liabe Brave
Member
Used that way, it would just be another method of AA...and one that generates artifacts scattered across the image, while being inaccurate in other areas. If you're able to render at a higher res than your output, that headroom would be better used on other methods, like MSAA or the regular types of temporal AA.
Eh. If the Xbox One drops to a 720p standard, it's possible that some PS4 games will do the same in order to achieve better visuals outside of the resolution.
FeiRR
Banned
You are close to devs so you should know that making a game look and run as good as possible with limited resources at hand is the most essential part of their work. If you can use tricks, you use them. It's not about a numbers race or console wars (PR takes care of that). It's about shipping an attractive product that people want to buy.
I smile (because I am a mean person) when I see people melt down over that famous Witcher gif or scraps of The Order footage. Because those people will go into meltdown again. Technology never lives up to the hype because hype has no limit.
velociraptor
Junior Member
KZSF's MP didn't have IQ as good as the SP but I always chalked it up to the FXAA. However, not once I thought to myself 'this isn't 1080p'. It looks like 1080p.
justsomeguy
Member
Someone with more skills than me needs to make a comedy Banderas gif maybe just showing the left half of the pic for 0.5s then the right half for 0.5s and so on.
Mobius and pet octopus
Member
WOW, and I thought I was done reading about ps4 a couple months ago.
Holy crap that's noticeable.I think GG made a bad move by trying to go for 60fps in the MP. The end result was pretty shoddy.
They should have kept it at 1920x1080, lowered some effects and try to get a solid 30fps.
instead of this:
Wish the plants had better AA.
Hmm I just noticed the odd low res rock texture on the right.
Liabe Brave
Member
All true, and it would be interesting to see the difference from a technical standpoint. I just think we know the results in general even without the exercise: this method looks noticeably worse than native 1080p, but noticeably better than any form of simple upscaling. After all, that's exactly what brought us here: people could tell MP was worse than native 1080p singleplayer, but it didn't really look like a true upscale. For example:
justsomeguy
Member
Damned good point. Am I right in thinking that most TVs will smear a bit anyway (because they're not properly set up -- see last night's tv set up thread) so you're not that likely to notice it? I guess the effect would be way more pronounced on a monitor?
I think it says wonders about the tech that GG used and how well implemented it is as myself who works in tech and design, loves games and I was one of the few to jump on the TR X1 lower Rez claim before it was public, missed this and NO_ONE in any part of the industry saw this and we all just thought a slightly blur AA solution was in fact an interlaced display (for want of a better description).
It is bad that DF never saw this (or never announced it until now all of a sudden) or that GG never came out and said it alittle after the game launched, but then looking at GAF with all the Resolution gate stuff hardly surprising, I bet they cannot believe they got away with it for so long.
I am not angry, just shocked to be fair
thunder_snail
Member
So do people think this method is better than 720p? only slightly higher pixel count compared to 720p but can be mapped per-pixel on 1080p displays.
I highly doubt we'll ever see a 720p game on the PS4. Reason being, the Xbox One will never go below 720p on any game (for PR reasons), and PS4 still outperforms the Xbox One handily at 900p (see: Battlefield 4).
Basically, PS4 has shown that it can render a full 56% more pixels and still run at a significant higher framerate than the Xbone.
Liabe Brave
Member
It's not regular aliasing you're seeing on the plants, it's artifacts from the reprojection method as the plants move. As for the "rock texture on the right", if you mean the one bound by the arc, that's actually a low-res reflection in water.Holy crap that's noticeable.
Hmm I just noticed the odd low res rock texture on the right.
RedAssedApe
Banned
let's see a 720p v. 1080pi v. 1080p shootout
CyclopsRock
Member
So this is basically just vertical interlaced? That worked fine on old TV's because the way they worked meant that the interlacing got ignored and the frames blended fine. I suppose this temporal blending is the modern equivalent of that!
But... it's a slightly weird existential question. At work whenever we render sequences that are particularly heavy, we tend to only actually render every other frame and then see well it blends using a frame blending algorithm. Some of them work literally flawlessly, where comparing them to an actually rendered sequence offers no difference. Some work terribly and it turns into a mess. The algorithms are complicated, but they work - mostly - on a purely pixel basis. There is not additional information and even for pure CG we don't provide it with any other information (like velocity data) that could potentially make its calculations more exact.
I assume, however, that KZ does. It has all the information - if it's rendering what is effectively half a frame, it knows where all the objects are on the screen, it knows where the shadows are falling (hohoho), it knows what objects are moving and what are static, how much the camera is moving etc. What it doesn't have to do is actually sample the pixels, AA them and all that gubbins. As such, I imagine that their ability to actually do this frame blending stuff well is very, very high (and the proof is in the output, obviously). It also leads to the question of "what counts as 1080p"? Afterall, the machine is outputting 1080p and whilst half of the frame is "made up", it's "made up" using a load of very useful data. It'd be easier to count the things it's not doing than the things that it is. Some games - like KZ Mercenaries - render certain effects at a lower resolution and overlay them onto the higher resolution main render. What resolution is that? It's all CG. It's all computer generated. I can understand why a 720p upscaled to 1080p can't be called "true" 1080p, but this? I dunno. I think it's more complicated than just "It's only rendering half the pixels".
Edit: This is all based on my understanding of the tech, which may be wrong.
But... it's a slightly weird existential question. At work whenever we render sequences that are particularly heavy, we tend to only actually render every other frame and then see well it blends using a frame blending algorithm. Some of them work literally flawlessly, where comparing them to an actually rendered sequence offers no difference. Some work terribly and it turns into a mess. The algorithms are complicated, but they work - mostly - on a purely pixel basis. There is not additional information and even for pure CG we don't provide it with any other information (like velocity data) that could potentially make its calculations more exact.
I assume, however, that KZ does. It has all the information - if it's rendering what is effectively half a frame, it knows where all the objects are on the screen, it knows where the shadows are falling (hohoho), it knows what objects are moving and what are static, how much the camera is moving etc. What it doesn't have to do is actually sample the pixels, AA them and all that gubbins. As such, I imagine that their ability to actually do this frame blending stuff well is very, very high (and the proof is in the output, obviously). It also leads to the question of "what counts as 1080p"? Afterall, the machine is outputting 1080p and whilst half of the frame is "made up", it's "made up" using a load of very useful data. It'd be easier to count the things it's not doing than the things that it is. Some games - like KZ Mercenaries - render certain effects at a lower resolution and overlay them onto the higher resolution main render. What resolution is that? It's all CG. It's all computer generated. I can understand why a 720p upscaled to 1080p can't be called "true" 1080p, but this? I dunno. I think it's more complicated than just "It's only rendering half the pixels".
Edit: This is all based on my understanding of the tech, which may be wrong.
Oh for sure. This is just to satisfy my own curiosity. GG has done a lot with resolution. Their 3D was done the same way WipEout HD was with the "stretched horizontal pixels" and KZ:M had dynamic resolution and now this title has the never before seen interpolation.
It would just be fun to see it go up against various methods and how much of an impact they have.
I'm open for exploration in those realms, but I'm also a person that doesn't care that much about resolution past a certain threshold. I think Thief at 900p on Xbone vs. 1080p on PS4 is visually not representative of what a 44% resolution difference would indicate.
I never played KZ:SF, but considering it took people months to figure out for certain that the game's MP mode runs at a non-native 1080p format, I would definitely have to say it's superior to upscaled 720p. It may even be superior to upscaled 900p (depending on one's preferences as to type/amount of scaling artifacts/blur).
Absolutely.
But when you downscale/downsample, some of the artifacts and inaccuracies would disappear with respect to the size of the footprint against the (higher)amount of downsampling.
If 2688x1512 resolution is to be targeted, pixels to render real time would be 1344x1512 or 2.03MP (Note: 1920x1080 is 2.07MP). Also hoping that the overhead for motion compensation/interpolation algorithm with downsampling would accommodate less resources. Then it might just be worth, just worth..
Framerate and resolution aside (features you can't sell) if the Xbox One has Quantum Break at 720p and the PS4 has The Order: 1886 at nearly 1080p the graphics of the Xbox One are competitive (because resolution won't mean much in that debate). You'll get games that end up looking as good as other disregarding resolution.
However, in the event that a Sony game decides to go 720p it will force a Xbox One game to go subHD or look worse. Upholding 1080p doesn't help PlayStation if Xbox sets the standard as lower.
Liabe Brave
Member
Yes, very much better. It requires no upscaling whatsoever in the vertical direction, and it limits horizontal upscaling-type effects to small selected areas on screen. (Plus it's still rendering more pixels than 720p to begin with, even if only half as many as native 1080p.)
The real question is whether this method looks better than upscaled 792p, or 900p. Those are rendering more pixels overall, but then using coarser methods to spread them over a 1080p screen. Ignoring personal preferences, it seems likely that Guerilla's method would give better IQ than 792p, and perhaps about equivalent to 900p.
In any case, I'd expect that Guerilla wouldn't have gone to all the trouble to invent an entirely new method if they didn't think it gave better results than a simple upscale that used equivalent rendering power.
foolishoptimist
Member
I wonder if Titanfall can be rendered in this res instead of 792p. Since the pixel count isnt that different.
Dictator93
Member
Lol even hardcore gamers can be fooled with resolution, how much do you guys think the average joe gives fucks? A little? Very little? Or not at all
tbh with you, if I had a PS4 and saw this in person, I would 100% notice. In fact, I did notice this in the pre-release footage but then just doubted myself when sony said it was 1080p.
A lot of people just took the devs and all the media content at its word (which is hilarious given so much of the resolution gate stuff).
Also, shame on DF for shitting the bed here.
Lol even hardcore gamers can be fooled with resolution, how much do you guys think the average joe gives fucks? A little? Very little? Or not at all
Well, if you mean 1080p whores can be bamboozled by an image that's damn near equivalent 1080p in detail/edge quality, then your point would be correct.
However, no one here would ever mistake a straight upscaled 900p or 720p image for a native 1080p image.
This seems to be one of those cases where we unfairly omit how it looks in motion: due to the temporal nature of the technique it is unfair to pick on screenshots. Picking on screenshots is part of the reason why we can't have nice things (e.g. 60fps) because devs prefer to make games that look good in screenshots but doesn't necessarily move fluidly (30 fps games that doesn't even run at 30fps.)
Edit: People seem to notice the blurriness also during play but the situation shouldn't be as bad as it looks on screenshots.
Edit: People seem to notice the blurriness also during play but the situation shouldn't be as bad as it looks on screenshots.
Lol even hardcore gamers can be fooled with resolution, how much do you guys think the average joe gives fucks? A little? Very little? Or not at all
1. Even the most ridiculous resolution arguments conceded the fact that not everyone cares. It's an issue to those wanting to spend money on the best possible version if the game.
2. Many people including myself noticd the disparity between the SP and MP versions. If anything this goes to show that resolution/fps is something you can notice.
SomedayTheFire
Member
Strange that the resolution was dropped like that and the performance was still not great.
AyaisMUsikWhore
Member
I doubt there will be any. The game has already been out for 4 months now. What damage control needs to be done when Infamous, The Order and other anticipated games with higher resolutions are coming out. The only people who are going to hold on to this are disgruntled Xbox users who can finally find a game in the Sony arsenal that isn't exactly what it is. AC4, BF4, NBA2k14 and others remain unbothered. At this point, it's a non issue but mostly moving forward (Sony exclusives only) we will have to keep a keen eye out.
The main issue I can see here is even if they are giving the illusion of 1080p, the interpolation algorithm does not bode well in terms of framerate. We do not know how costly it is in terms of resources or if the code is still in infancy stage (possibly unoptimized algorithm) that's causing performance issues. The tech indeed is amazing but it appears the trade-off they made was not enough to provide the 60 FPS experience people were looking for. Hope GG can give more insight on this issue.
Read the thread for christ's sake.
I'm assuming that this temporal interpolation method uses a nice, big chunk of video memory. 32MB of ESRAM would prohibit this method from ever being used in an Xbox One game.
Which is a shame (and ironic), because if there were ever a console that desperately needs a higher-quality scaling technique, it would be the Xbone.