Floaters or extra triangles - what's best performance wise?

sltrOlsson

Title pretty much describes my question.

Lets say you are doing a rail for a gun, is it best to leave them as floaters or connect them together with the barrel? And that is best performance wise.

Mark Dygert

For the high poly and baking? Whatever works best for you, performance shouldn't matter. Personally I like floaters, they're easy to edit and you don't have to add and manage a bunch of extra edges.

If its for the final mesh, there are some advantages and draw backs to clipping one mesh into another, but it starts to venture into "it depends" land. If you have a specific example like the gun is a First Person View weapon for XYZ game running on ABC hardware in 123 engine using DEF lighting, then it's a little bit easier to say, "yea floaters will work best" or "no you'll probably run into these issues if you're not careful".

cryrid

There's some info covered at http://www.polycount.com/forum/showthread.php?t=77428

[HP]

You should try and explain yourself better, do you mean floaters for high poly stuff? Or floating geometry within your LP?

If you mean your LP, basically, keep your valuable geometry for silhouette stuff. For HP, surface detail floaters will almost always get the job done, in some situations even work better.

sltrOlsson

Ah, I'm sorry for being unspecific. I mean for LP. But is it that much of a difference from engine to engine and hardware to hardware? I guess you just keep to one of them through the model then.

Sean VanGorder

Is this the sort of thing you're asking about?

http://racer445.com/pages/tutorials/rail-for-fps-guns.php

EarthQuake

Racer's little minitutorial, while helpful only really covers one side of the issue, there are certain pros and cons to each.

pros of using more mesh chunks:
1. Can use less geometry
2. Can use "cleaner" geometry, ie less long thin triangles, which are generally bad

cons of using more:
1. Much harder to get a "seamless" look to bakes, ie: if you break the tops of every rail off, you'll have to explode bake those and end up with pretty obvious uv/geometry seams there. Generally every time you split off a mesh chunk, you're going to have aliasing that you otherwise wouldn't with a proper "solid" mesh.
2. May waste more texture space, may use about the same but will rarely(if ever) save texture space.

So, there really is not "clear" choice here, they both are good and bad in different ways, I generally try to keep meshes seamless aside from parts that need to animate. However, if I'm making a gun that has 200 rail notches, sometimes its just impracticable to model that solid. If you've got a bit of freedom on the design side, it helps to have those bits physically *Be* different, so you're not creating negative artifacts.

IE: The highpoly and the low are the same in what is a "separate" piece.

sltrOlsson

Sean_EG wrote: »

Is this the sort of thing you're asking about?

http://racer445.com/pages/tutorials/rail-for-fps-guns.php

Yeah, i was thinking of something like this.

But i guess i get it, Haven't really thought about the smotth group and UV splits create vertices to. I always tend to split up edges and stuff that will have normalmap applied, so that i won't have any big gradients in my normalmap..

EarthQuake

You can still get a seamless bake when you have split smoothing groups, provided your projection cage is averaged(by default if you're using a cage in max, and not the "offset" method).

But if you have physically split the geometry, you cant.