R&C has other problems though, like disappearing shadows.
Shadows in PS3 games looking really bad
- Thread starter gwaves
- Start date
But aren't those shadows projected shadows? You know like in Max Payne 2, Jedi Knight 2 or a lot of other games. Those shadows have no jaggies.
Hi, i've been having some problems with PS3 games that I own, Heavenly Sword and the GT5 Prologue demo seem to have these problems the most. They both seem to look awesome on every department, except with shadow. They really look bad, and playing Heavenly Sword and GT5 in cockpit mode are pretty much unbearable in my opinion. Here are 2 pics showing what I mean.
Most console games has that problem and it´s kind of anoying sometimes IMO, I´m a graphics enthusiast to the max
.I dont think it´s a problem with your PS3 though,
better shadows = higher resolution or more "smoothing steps" (can´t remember the real name) = reduced performance (fps = frames per second).
I think it´s more of a choice from the developers side,
will they choose better shadows to the cost of framerate or the other way around?
I can imagine that most developers are aiming at having a stable framerate the whole time or atleast most of the time, and with that there might come some sacrifices, It would not surprise me if one of those sacrifices are the shadows.
Max Payne 2 does show shadow aliasing... it all depends on your perspective. If you crouch and put the camera on top of Max, you should see aliasing (I do). Remedy just happened to pick a particularly good enough resolution that most of the time you won't see aliasing i.e. running around.
Jedi Knight 2 uses stencil shadows, which have different aliasing issues than shadow maps.
Hmm yeah it was some time ago I played Max Payne 2 neverthless they do look good for that time. Maybe I'll install the game to finish it if it works on Vista x64.
http://media.pc.ign.com/media/482/482121/img_1851259.html
Shadow maps have inherent problems, particularly with their current implementation. We in offline CGI had to manually finetune the bias parameter for every shot and every light to avoid this kind of artifact.
The solution is called midpoint shadow maps, I'm not sure how complicated it is to implement in real time though. I recall nAo or maybe Carmack talking about the issue some time ago...
The solution is called midpoint shadow maps, I'm not sure how complicated it is to implement in real time though. I recall nAo or maybe Carmack talking about the issue some time ago...
Are there any games implementing Summed Area Variance Shadow Maps? Is it even feasible? There was some neat discussion over awhile ago in the 3D Tech & Arch Forums. It seems like it would be a performance hog as implemented by Andy, but Mintmaster, nAo, ShootMyMonkey et al seemed enthusiastic about working with it.
Filtering on its own can't fix shadow bias issues. I'm not really into the actual technology but I'll try...
The general idea is that the shadow maps store the distance of the closest surface point to the light. This distance measurement is usually not entirely correct, for lack of precision, and that's when you start to have artifacts on surfaces, where they seem to move in and out of shadow. As the object moves around, this effect is emphasized by flickering.
Logically, the more distance a depth map has to cover, the less precise it will be. Obviously, large scale outdoor scenes are very good examples for this. You can however add near and far ranges to the light if you're sure that no object will have to cast a shadow that gets closer to, or further away from the light.
The bias value for a shadow is a certain small distance that's added to every value stored in the shadow map during the rendering. It's like you push the shadow itself a bit further out from the light. However if you increase the bias too much then objects that should be in shadow start to get lit (the light starts to "leak"), and shadows cast on the ground by objects start to move away, too.
There really is no fixed way to set this bias value, it requires tweaking in offline rendering to find the balance between artifacting and incorrect shadows (pushed out too far). This is obviously impossible in an interactive application where lights and objects are on the move all the time.
Midpoint shadows store the averaged depth of the two surfaces nearest to the light and this tends to solve shadow bias issues. But I think it's computationaly more intensive, and introduces other kinds of problems...
The general idea is that the shadow maps store the distance of the closest surface point to the light. This distance measurement is usually not entirely correct, for lack of precision, and that's when you start to have artifacts on surfaces, where they seem to move in and out of shadow. As the object moves around, this effect is emphasized by flickering.
Logically, the more distance a depth map has to cover, the less precise it will be. Obviously, large scale outdoor scenes are very good examples for this. You can however add near and far ranges to the light if you're sure that no object will have to cast a shadow that gets closer to, or further away from the light.
The bias value for a shadow is a certain small distance that's added to every value stored in the shadow map during the rendering. It's like you push the shadow itself a bit further out from the light. However if you increase the bias too much then objects that should be in shadow start to get lit (the light starts to "leak"), and shadows cast on the ground by objects start to move away, too.
There really is no fixed way to set this bias value, it requires tweaking in offline rendering to find the balance between artifacting and incorrect shadows (pushed out too far). This is obviously impossible in an interactive application where lights and objects are on the move all the time.
Midpoint shadows store the averaged depth of the two surfaces nearest to the light and this tends to solve shadow bias issues. But I think it's computationaly more intensive, and introduces other kinds of problems...
nAo has a very nice discussion on shadows on his blog. What I have noticed is most games on both platforms have huge shadow issues... and it doesn't both most. And some games with excellent shadows (e.g. Halo 3 which has nice soft shadows and dynamic shadows and lacks a some of the aliasing problems that immediately pop out in other games) it gets overlooked. The early Mass Effect media was driving me *nuts* totally destroying the visuals for me and yet a lot of people fawned over it. I think it is a question of what your eye is drawn to.
Uhm, as far as offline CG goes I can think of at least 3 different solutions that all address resolution and precision issues alike, and have had associated papers around for a couple of years.Laa Yosh said:
Personally I have no idea at all why CG toolmakers don't integrate at least one of them (if they haven't already). Or possibly the methods you guys use are tried and true enough that noone wants to switch, even though you're using hacks?
Or maybe I'm reading your post wrong and you've already been using something better now?
That said - application of those approaches in realtime is another matter alltogether. On that note, midpoint shadowmaps doesn't really address resolution issues (some of the other techniques do that as well).
It has to be - TombRaider on 360 was a lot worse then even MassEffect and it was a released game, to the point where I greatly preferred the look of PC and PS2 versions even though neither of them had global or selfshadowing.Joshua Luna said:
Yet many people still classed it as a good upgrade...
You might want to read nAo's blog for more about filtering improvements on shadows. Research has been going on on various analogous techniques to variance maps.[COLOR=black]AlStrong[/COLOR] said:
Last edited by a moderator:
Similar threads
- Locked
