Poly/Tri count on combined model compared to another method

wat

Hi, I'm relatively new to modelling and have a question about which is the better method of making a model.
In the pictures below I have made the same model in two different ways, the first has a lower poly/tri count but is completely made of combined objects while the second, with a much higher poly/tri count was all created from a single object.




Is there any reason to not go the combined object route? It seems to require less poly's to get the same result. Or is that a bad practice?

Ghogiel

Depends.

There is zero context here to figure out what sort of thing would be better for the situation. Each have benefits, limitations and draw backs or advantages in different areas and situations. For example he top one you might be concerned about projection baking to it and hidden faces wasting UV space. While the second might be harder to quickly edit and move the box shapes around because it's all connected and has a load of loops going all the way around it.

wat

Thanks for the reply, sorry about the lack of context but I think you've answered what I asked pretty well, I just wanted to know some of the draw backs for relying too heavy on the attach tool when creating a model.
The context is that I'm trying to make some modular castle walls, trying to create the entire wall out of a single mesh required a lot of faces, so I was wondering what the negatives of separating the wall into different parts and then attaching them together would be.

poopipe

The first one wastes texture space compared to a watertight mesh (assuming unique bakes). 

The second could have significantly less geometry and still look the same
Sooo....
In this case the second option without the wasted geo would be your best balance  -. Provided we're just talking about a bit of static geometry with no weird shit going on. 

The balance changes as geometry gets more complex and the cost of wasted texture space is outweighed by the cost of the geometry required to make the mesh watertight (also when dealing with lods)

wat

Am I using too many loops on the second one? I thought I needed them to avoid creating Ngons, and yes Its just a static mesh. I'll work on improving my geometry with the second one then, thanks for the advice.

sprunghunt

wat said:

Am I using too many loops on the second one? I thought I needed them to avoid creating Ngons, and yes Its just a static mesh. I'll work on improving my geometry with the second one then, thanks for the advice.

For a lowpoly that would go in a game engine it's ok to have triangles. So you can get rid of the extra loops and have triangles instead. 

Ghogiel

^ everything is a triangle. Ngons and quads don't really exist. 

You are in Max there anyway. Which knows what a polygon is. You have complete control over all this stuff, especially the triangulation of polygons, be they ngon or quad

wat

So in this case ngons would be fine? I thought if a face with more than 4 sides isn't completely flat it can cause it to bend in strange ways because of how max tries to split the shape into triangles.
I guess that wouldn't be a problem on my mesh because its got nothing but flat faces. I'll have a go at making the mesh with ngons and then seeing how max changes them into triangles. Thanks for the feedback

poopipe

Ngons are a bad idea, triangles are good and quads are fine.theres a lot of very specific things about triangulation that only people who understand how renderers work know about but you can safely pretend they don't exist until it becomes an issue (which is almost never)

Before you bake you want to triangulate or you risk issues when the baker and target engine don't auto triangulate the same way. 

On something this simple I'd probably explicitly triangulate the whole mesh - by the time youve trimmed out the unnecessary geometry you'll be 90% there anyways

Ghogiel

wat said:

So in this case ngons would be fine? I thought if a face with more than 4 sides isn't completely flat it can cause it to bend in strange ways because of how max tries to split the shape into triangles.
I guess that wouldn't be a problem on my mesh because its got nothing but flat faces. I'll have a go at making the mesh with ngons and then seeing how max changes them into triangles. Thanks for the feedback

Ngons are tris. but obviously don't go out of your way to make them because you would probably be modelling a bit wierd and ineffeciently if that's happening. But things like caps of cylinders In max are ngons by default and there isn't a huge imperitive that you have to do anything about that for example. If you want, you can. You can see and change triangulation and not set them explict by hand unless you have a reason to. Things like turn to poly modifier exist in max too, which can tri up a mesh if you wanted as well.

wat

So use the poly modifier to tri up the mesh, and then optimize it for whatever I'm doing, unless the models tris already look ok and don't cause a problem, like when its on the cap of a cylinder.

kanga

wat said:

So use the poly modifier to tri up the mesh, and then optimize it for whatever I'm doing, unless the models tris already look ok and don't cause a problem, like when its on the cap of a cylinder.

This goes back to what you want as an end result. To optimize the help you receive explain what your end goal is. Making a castle isnt enough. Explain what you want the castle for. Just for modelling practice, for use in an animation, a still shot or a game asset? In some cases you can ignore ngons, in others you need a sealed hull. Depending on the structure you may have to divide the model up to avoid baking errors, or use a cage. Your question is like: should I walk my dog, if so how far?