For the Geometry Snobs

[blakjedi]

In PDZ and Kameo when you get right up along a wall... it had raised, not flat elements from every angle with fairly correct lighting... I can not see how geometry would have helped there... how do i do thumbnails so i do not post oversize pictures...?

 

[DudeMiester]

The only reason you need polys is so that you don't see faceted edges on objects when there should be smooth curves. Internal displacement is better done using raytracing techniques, imho.

 

[Pete]

how do i do thumbnails so i do not post oversize pictures...?

Probably { url = fullsize } { img } thumbnail { /img } { /url }.

 

[ShootMyMonkey]

not to mention that the parallax mapping in PDZ give a far more impressive result than COD2 and as i said is simply not achievable with geometry.

I really wouldn't word it that way. It's not that it "isn't achievable with geometry", it's that "the necessary geometry density isn't achievable" to match parallax maps. Parallax mapping is simply an illusion based on a texture, but geometry is geometry. If you could achieve geometry on the same level of resolution as the texture, it will inherently be more accurate and consistent than parallax or whatever because it won't be limited by the precision of texture data, and it won't simply be an illusion -- the shape of the object itself is changed (something parallax and bump mapping alone can never achieve).

 

[Jabjabs]

In PDZ and Kameo when you get right up along a wall... it had raised, not flat elements from every angle with fairly correct lighting... I can not see how geometry would have helped there... how do i do thumbnails so i do not post oversize pictures...?

Actually if you get a close enough angle they do become flat surfaces showing no bumps or details at all, you will see just how little geometry is used in some environments if you do this.

I think some of you here need to remember that one of the biggest cheats in computer graphics was the invention of the texture. Why use millions of polygons to define the details of a surface when you can just make two and slap a texture on top. Bump/Parallax mapping is just an extension of this idea.

 

[Grall]

A Wall with fine-crafted or even with just ok bm looks 3d even if you watch it from a side angle unless you are talking for real crappy bm.

I don't know what planet you're from; on planet earth, bumpmapping doesn't ACTUALLY create bumps, it just fakes it by modifying pixel colors to add illusional shading on an otherwise entirely flat surface.


Good bumpmapping looks reasonably realistic when viewed more or less perpendicular to the bumpmapped surface. The illusion quickly breaks the more parallel to the surface the viewing angle becomes.

Not to mention that what you can do in game enviropments (walls and grounds etc) with paralax mapping is not achievable with geometry unless you spend half of your poly-budget for this part of the enviropment and let the rest of your game to looks like crap.

Question:
Where do you think your bumpmaps come in the first place, magicked out of thin air perhaps?

Real answer:
Generated from geometry.

Btw, comparing PDZ and COD2 isn't going to be very fruitful when talking about two as aesthetically different games as these. COD2 is a straight PC port by the way, so naturally that's going to color its looks as well.

Apples and oranges, you know?

If we watch in the past when CryTek brought the PolyBumping technique on the table it was a revolution for the video gaming.

Nonsense. Not only was "polybump" not invented by them (they just slapped their own trademarked name on it), it was actually used by other PC devs before they did, namely id software and the Doom3 engine demo at that Mac expo back in 2001.


*G*

 

[ShootMyMonkey]

versus a supposedly high-poly lesser mapped game that has been touted as a better looking more aesthetically pleasing look

Higher poly maybe, but who says MGS4 is light on the "maps"? It's got normal maps and specular level maps and god knows what just as much as any other game. And they're also clearly pretty high res, too. I don't know what or who gave people the idea that it's "lesser mapped" than a large selection of other games out there. Maybe if you were comparing to something that just piles on the effects and includes things like ambient occlusion, subsurface scattering and other things, that would be true.

 

[zidane1strife]

Can anyone show me what the visual difference is between a game with high geometry and a game that is textured and mapped well? Its seems some prefer "loads" of geometry in their games to heavy mapping but I really can't see the difference to the point where someone is a snob about it...

Personally, I prefer high poly budgets on my characters. I just don't like rough-jagged heads, limbs, etc. I also prefer clothes that actually move for a change over static-sticker-like clothing. If we go into armor it'd also be nice if it slid around a bit(depending on the design and location of the piece of armor) and didn't just go through the other pieces of armor(Some aspects of GoW's character's armor where changed seemingly due to this, or maybe not, just know they were simply taken out. It also seems like they bumped the geometry some in the latest builds.). When it comes to skin folds/wrinkles it's best if they deform properly like on snake's face/neck than if they're just static-sticker-esque. faked/realtime-muscle-simulation is also another plus.

As for overall poly use vs mapping techniques. Some psone characters had organic features, individual fingers, and non-blocky facial features/limbs/etc, that is round and not poly-starved(as some so called next-gen characters have been.), with high levels of animation, seemingly surpassing some next-gen characters, iirc. I favor very low poly and even simple textured objects wherever they're possible so long as they don't break the illusion. Blockiness is just sad in this day and age, considering even psone managed to do away with it in some models so many years ago, just evidence of the wrong path both ati and nvidia have taken with overly focusing on mapping techniques at the expense of glorious geometry.(Thank god for the ps2's example, who knows how much lower our poly budgets would be today if it weren't for it.).

 

[AlNom]

Question:
Where do you think your bumpmaps come in the first place, magicked out of thin air perhaps?

Real answer:
Generated from geometry.

Not always true. You can generate normal maps by "painting" the details on instead of using geometry.

 

[Richard]

More/better grass.

 

[Danalys]

can someone explain to me why there isn't such a thing as poly mapping (that's if there isn't such a thing?) the bump texture could be a store for vertex data instead. the resultant polygons that are generated from the map are then lit following similar methods to how the pixel would be lit normally. the good thing is that the map could be scaled and rotated as if a texture meaning LOD would be perfect or perhaps it would go strangly wrong. someone would have to test an example to be sure.

actually if you just rotated and scaled the texture you'd lose the geometry. you could scale the texture then rotate the geometry tho.

 

[Acert93]

What you are suggesting sounds similar to Displacement Mapping. Vertex Texturing (or VR2B too?) can do this sort of task; on the consoles RSX should be able to do this sort of thing (seeing as G71 can through vertex texturing) and Xenos can as well (in addition the ATI dev PP slides have some notes about using the hardware tesselator in conjunction with such).

Of course there are some nice hacks/cheat that go beyond normal mapping, but are not quite real geometry like displacement mapping. For example Parallax Mapping (aka offset mapping or virtual displacement mapping). I am not totally sure how relief mapping fits into this equation (been a while since I looked at it). Some methods seem to overlap and even have more than one name (e.g. steep parallax mapping seems is a parallax, self-occlusion, and self-shadowing technique which seems to have a bit of an extension of parallax mapping, at least according to the site). Actually, now that I think about it, I always assumed Parallax Mapping and Parallax Occlusion Mapping were the same, but they may not be. Now I am confused

Btw, here is a good example of parallax mapping being used on a wall and how "art" is used to hide the silhouette edge. As the player cannot see the edge of the face of the wall the faked geometry effect cannot be easily seen. In situations where the effect cannot be ruined by getting flush with the polygon face effects like this stand up well IMO.

 

[Danalys]

thanks acert. i should probably have read up on them all before. sometimes i like to be free of knowlage of the conventional way to do things and work from base principals to think of ways things could be done. but that means i waste my time reinventing the wheel at times.

 

[Acert93]

thanks acert. i should probably have read up on them all before. sometimes i like to be free of knowlage of the conventional way to do things and work from base principals to think of ways things could be done. but that means i waste my time reinventing the wheel at times.

NP, I was actually pretty impressed that you came up with an idea exactly like a real world solution :smile: I am just a techie gamer, so I just read up on stuff and try to understand it. It is another deal altogether when you can imagine solutions to problems, that make sense, only to find our you were the 2nd (or 3rd, or 4th...) person to "invent" the idea. This is why patents can stink with software (or hardware even). Sometimes there are only so many logical ways of doing something.

It is pretty cool imo how pixel and vertex shaders are being used to provide a lot of detail from textures. Bump mapping was interesting, and normal mapping really brought a lot of detail to games. But POM and Displacement Mapping look to really up the level of detail in realtime games with decent fidelity. Nothing beats real geometry, but the processing needs for such at the same level of detail is still pretty high.

 

[Danalys]

okay another idea. pre tesselated displacement maps. the texture would contain pixels aranged in a brick type pattern. new scalers would be needed to keep the brick pattern. is that already done?


hmm i think i was talking rubbish then. the tesselation is in the creation of the polygons from the vertex texture right? so it can't be pre done.