Next-gen graphical pet peeves

Apart from the light bleeding the only other 2 issues I have with it are:
1) prefiltering has a linear costs compared to PCF quadratic costs, but PCF runs on pixels you need
to shade, while prefiltering runs all over your shadow maps and often shadow maps pixels count is quite higher than frame buffer pixels coiunt
2) it consumes more memory (expecially at 32 bit!) compared to common shadow maps and memory is a precious resource, especially on consoles.
The second one doesn't quite bother me as much because I tend to see a lot of shadowmap usage cases that eat up massive amounts of memory as well. e.g. cascaded shadow maps, multiple perspective shadow maps, shadow cubemaps... and the thing is that all those are rather readily considered.

I would certainly hate to see 1024x1024 32-bit cubemap VSMs for point lights, though. Nothing says 'frugal' like sending out an email saying "Anybody have 48 MB of memory space to spare?"
 
I'll look forward to a generation where a stable 120 fps is a mandatory checklist for release.

Well, once 120 Hz becomes a mandatory checklist on new TVs...

My pet peeve is...I have no pet peeves. I'm easy to please with eye candy. I think developers do need to remember that reality is boring. I see water, grass, trees, snow, and buildings every day. However, I do not regularly see lasers, magic spell effects, huge explosions, weird monsters, mythical fortresses, and alien spacecraft. Or Hitler on a T-Rex stomping around the surface of the sun. Seriously, most games are a total joke.
 
zed said:
non unified lighting would be the biggest thing.
lack of solidity in the scene, 2d plants eg speedtree is a nono
repetitive textures
lack of AF
sub par framerates (more so with bouncy framerates, have no problem with a steady 30fps etc)
empty environments
Nice setup for my pet peeves --

People who measure everything in graphics... and no I'm not just talking about the graphics whores like babcat/MILR... I also mean using graphics as a measuring stick (e.g. posts saying things like "They can get UE3 working on the Wii! They just need to tone down the graphics!" or "If they're tapping only about half the power of the console, does that mean that it'll look twice as good when they really get the hang of it?" or some really fun ones like, "Can they make an engine that gives us both next-gen graphics AND allow players to jump longer distances?")

People who shout 'next-gen' from the hilltops as a sort of universal proof that everything should be solved. Oh yeah, we've got eDRAM on the 360, so there should never again be any problem with shadows. Load times won't ever be an issue on the PS3 because the Cell processor and its unparalleled "synergy" with the RSX is so mighty. BTW, I heard this rumor that the PS4 will be so next-gen that it will actually stop people from farting because the CPU is that powerful.
 
Nice setup for my pet peeves --

People who measure everything in graphics... and no I'm not just talking about the graphics whores like babcat/MILR... I also mean using graphics as a measuring stick (e.g. posts saying things like "They can get UE3 working on the Wii! They just need to tone down the graphics!" or "If they're tapping only about half the power of the console, does that mean that it'll look twice as good when they really get the hang of it?" or some really fun ones like, "Can they make an engine that gives us both next-gen graphics AND allow players to jump longer distances?")

People who shout 'next-gen' from the hilltops as a sort of universal proof that everything should be solved. Oh yeah, we've got eDRAM on the 360, so there should never again be any problem with shadows. Load times won't ever be an issue on the PS3 because the Cell processor and its unparalleled "synergy" with the RSX is so mighty. BTW, I heard this rumor that the PS4 will be so next-gen that it will actually stop people from farting because the CPU is that powerful.

Im confused on what you are saying. Are you using my post as a representation of what peeves you about next gen graphical pet peeves or consumer/fan ignorence?? (Cpi??)

Would it have been better if I said low res. textures or low poly count characters??

Sorry if I strike you as that way (ignorent). Although to be honest a good portion of my game enjoyment revolves around graphical presentation (I believe that is true for most of the gamers out there).
 
Last edited by a moderator:
Im confused on what you are saying. Are you using my post as a representation of what peeves you about next gen graphical pet peeves or consumer/fan ignorence?? (Cpi??).
I can't say I know for sure SMM's intentions, but I can certainly empathise with what I'm seeing in it. Basically this thread could be summed up as...

I dislike the way console graphics aren't totally realistic.

To whit the response is

'They're finite systems with limitations, so be happy with what you've got!'

It would perhaps be more productive if people offered some positive recommendations, though I'm not sure how. I just imagine devs looking at threads like these and seeing people whingeing about all the things their games don't do, as opposed to all the things they do do. Two years working on game xxx and the response is "why's it not perfect" isn't good for morale ;)
 
I would have to agree although I never meant to give the impression of belittling anyones personal efforts.

To a point we all are somewhat jaded by complaints from the masses (well at least I feel that way from time to time in my job :)).
 
Last edited by a moderator:
What annoys me a lot are normal mapped surfaces which look blurred or flat when the camera is to near their surface.

I also dislike low poly models with crazy normal maps. you see detail and the impression of complex textures and surface in the middle, while on the model's outlines you see rough edges.
 
Judging from my inbox recently, I suspect VSMs will be used in a bunch of upcoming games. Remember that they still are quite new.

Apart from the light bleeding the only other 2 issues I have with it are:
I'm actually really happy with the "lopping off part of the distribution" light bleeding reduction stuff. It works really well, is totally free and trivial to implement. Even if exact results are required, VSMs can still be used as a conservative oracle to quickly eliminate regions of all light or dark.

1) prefiltering has a linear costs compared to PCF quadratic costs, but PCF runs on pixels you need
to shade, while prefiltering runs all over your shadow maps and often shadow maps pixels count is quite higher than frame buffer pixels coiunt
Well for one I'd expect a good projection warping scheme to take care of this. Secondly, blurring is so cheap nowadays... I'm doing gigantic 50x50ish blurs on large shadow maps and still getting frame rates in the 300+ range in D3D10. Granted that's on an 8800GTX, but no one needs blurs that large anyways :)

Summed-area tables (and perhaps something like bicubic filtering) also solve the problem, although the former introduces a few new ones, at least until doubles are supported in hardware.

2) it consumes more memory (expecially at 32 bit!) compared to common shadow maps and memory is a precious resource, especially on consoles.
That's definitely a concern, although not as much on the 768MB+ monsters of tomorrow. I suspect consoles will catch up eventually.

I just hope people don't get the wrong ideas about VSMs. They're really a point on which to build, but people seem to like to dismiss them with trivial arguments that can be easily addressed. It's like dismissing standard shadow mapping because "it has biasing issues" (which incidentally VSMs don't ;)).
 
Last edited by a moderator:
my main gripe thus far is with substandard filtering/texture aliasing. some of the things other people have mentioned are issues for sure, but, for me, when a "next gen" game ships with crap filtering it shows. also, it's not like enabling higher quality filtering would break most games or require costly man-hours. this is all, of course, assuming the machine i question has the horsepower to do it at a desireable framerate.
 
Im confused on what you are saying. Are you using my post as a representation of what peeves you about next gen graphical pet peeves or consumer/fan ignorence??
Example, actually, as you echo a lot of the same things, and I used zed's for the same reason. I don't really care how ignorant or not you are. Your post was a springboard.

Although to be honest a good portion of my game enjoyment revolves around graphical presentation (I believe that is true for most of the gamers out there).
Yep. I also often find that the people who are more vocal about gameplay over graphics are also the ones who will readily join in on the "these graphics suck" bandwagon. And the simple fact of the matter is that people don't really have the capacity to perceive much more than that in general (unless of course their jobs depend on them perceiving more than that). I don't think that at any point within my lifetime will the buying public be able to comprehend the complexities of a lot of things that aren't at all visual or don't have any obvious feedback... especially when something which is internally very complex and difficult to manage accounts for something so minor on the surface. If anything, I think the ignorance will only increase and it will just *sound* as if the consumer is getting smarter because the technical terms of today will be buzzwords for tomorrow -- just like "FLOPS" and "HDR"...
 
Yep. I also often find that the people who are more vocal about gameplay over graphics are also the ones who will readily join in on the "these graphics suck" bandwagon. And the simple fact of the matter is that people don't really have the capacity to perceive much more than that in general (unless of course their jobs depend on them perceiving more than that). I don't think that at any point within my lifetime will the buying public be able to comprehend the complexities of a lot of things that aren't at all visual or don't have any obvious feedback... especially when something which is internally very complex and difficult to manage accounts for something so minor on the surface. If anything, I think the ignorance will only increase and it will just *sound* as if the consumer is getting smarter because the technical terms of today will be buzzwords for tomorrow -- just like "FLOPS" and "HDR"...

I never really thought about it in a sense of capacity more so time or availabilty. I doubt the average consumer would be able to dedicate a portion of his life to understanding what goes on during the developing of a typical game or have access to the knowledge, although you may be right about the general population not being capable of understanding. I would say that people who invest time and money into an entertainment medium should be able to criticize it without being in the know. As to say, if I watch a movie I should be able to criticize it whether or not Im a director and fully understand the ramifications of the job. Even if I teach a class then only those that teach or work in education should be able to criticize me or the job I have done because they can relate. I wouldnt necessarily agree with that.

I am honestly sorry if my statements or my digital persona represent what you dislike in the average consumer or fan. I didnt intend to give the impression that I dont appreciate the work that goes into game graphics or games overall, because I can assure you I do. I personally think that anyone that plays games or has some remote interest in games appreciates it to some level or another (despite how the internet may make people appear to be, or whether these people acknowledge it or not).
 
Last edited by a moderator:
I never really thought about it in a sense of capacity more so time or availabilty. I doubt the average consumer would be able to dedicate a portion of his life to understanding what goes on during the developing of a typical game or have access to the knowledge, although you may be right about the general population not being capable of understanding.
I wasn't referring so much to making sure everybody on the planet knows multivariable calculus before they're allowed to play a game. More to be aware that said knowledge goes into making one. People don't have to know in full detail the interdependencies inherent in something as simple (on the surface) as being able to switch weapons in an FPS, but be aware that it's there, and that there are things about it that they'll never understand.

Yeah, the industry itself has been rather closed off for a long time, and things like NDAs do mean a lot of knowledge stays within the industry -- but I'm at least saying that idiots shouldn't kid themselves into thinking they're anything more than idiots. Just like I don't kid myself into thinking that I'm not a sociopath.

I would say that people who invest time and money into an entertainment medium should be able to criticize it without being in the know. As to say, if I watch a movie I should be able to criticize it whether or not Im a director and fully understand the ramifications of the job.
A movie is actually a pretty good example because the fruits of everybody's labors is present in a purely audiovisual medium. Everything you get is straightforward feedback of what went into making that film. Sure, you may never know what a key grip, but at least you don't pretend to be better at his job than he is. That's rather different from how it is with the average gamer. That may be harder to see on a board like this one where intelligent conversation actually takes place, but this board or psinext or aceshardware are hardly representative of the masses.

Even if I teach a class then only those that teach or work in education should be able to criticize me or the job I have done because they can relate. I wouldnt necessarily agree with that.
Education is a bit of a bad example because the product is actually people, in a manner of speaking, and it's something many people have been through. Moreover, it's not something that can be judged so readily until after the fact. That too, I don't think the average child actually going to school has a clue what a teacher's job entails, but the parents of those children have a better chance of having a clue, and that too, they know what sort of a variable their own children are (hopefully) -- overall, it's not something that the "clientele" is totally divorced from.

I personally think that anyone that plays games or has some remote interest in games appreciates it to some level or another (despite how the internet may make people appear to be, or whether these people acknowledge it or not).
I can believe that of various individuals such as yourself, but I do think you give humanity in general too much credit.
 
I'm actually really happy with the "lopping off part of the distribution" light bleeding reduction stuff. It works really well, is totally free and trivial to implement.
It's perfectly fine for me as well, I would just like to have a way to reduce bleeding without overdarkening -> removing fine details from umbra edges.
Bleeding is also annoying when you have some shadow casted from a distant object moving and intersecting a shadow casted by a close object (close to the light..) and when those shadows 'composite' together it really looks really wrong, imho.
Nonetheless VSM are an amazing achievement, it's really rare and uncommon having completely new ideas being 'injected' in this field, I'm sure VSM just opened a new 'phase'..
 
It's perfectly fine for me as well, I would just like to have a way to reduce bleeding without overdarkening -> removing fine details from umbra edges.
Yeah I agree. At least one person has used VSMs as an oracle in the offline world, which works notably better than the envelope approach. I think eventually we'll be able to use an nice adaptive algorithm (SATs would be really nice here ;)) to get perfect shadows. Perhaps eventually we'll have deep shadow maps on the GPU or similar, but that's analogous to an A-buffer, which I was bashing in the "D3D11" thread ;)

Bleeding is also annoying when you have some shadow casted from a distant object moving and
intersecting a shadow casted by a close object (close to the light..) and when those shadows 'composite' together it really looks really wrong, imho.
Yup, there doesn't seem to be a good solution for multipenumbrae regions that isn't just brute force (PCF or better). Hence the need for an adaptive algorithm IMHO.

In any case I'd first be happy to see properly filtered shadows, even if they do have a bit of light bleeding or incorrect penumbrae. After games have achieved that, we can make them even better :).
 
Yup, there doesn't seem to be a good solution for multipenumbrae regions that isn't just brute force (PCF or better). Hence the need for an adaptive algorithm IMHO.
I'm sure solutions will be found.. :) Anyway..I've just realized Little Big Planet is using VSM, it's relatively easy to spot light bleeding when penumbras intersect, especially in the GDC presentation movie when they are showing off level editing capabilities..
Actually I think VSMs make a lot of sense for their game!
 
Back
Top