I'm currently in school learning about game design, and game art production. My school teaches us how to model, texture, and light as well as getting our work into UDK. However since learning how to model within Maya I'm constantly having my meshes either being too low poly or too high poly for their purpose. My school doesn't really teach us how to optimize our meshes. I dug around the archives section to make sure I just didn't miss a awesome thread that might have been bookmarked (although there are a few that I'll have to revisit in there
![:) :)](https://polycount.com/plugins/emojiextender/emoji/twitter/smile.png)
)
I'm starting to feel a major crutch as I'm trying to work on a group Environment piece for our class and I'm having a hard time keeping up with the modeling when I don't quite understand when and how to optimize my meshes. I also feel that this slows me down when I go to UV, then eventually when I go to texture, so on and so forth. If you have a crap foundation, the house won't stand so to speak.
So I'm reaching out to the community for thoughts, and opinions on the matter. How did you guys learn to create good, clean game assets? Obviously on the job your forced to, but in school it feels like I'm unable to get that guidance. This whole topic started after a fellow classmate noticed my geometry had way more tri's then needed. Obviously its probably dependent on what your modeling, but I'm opening up to the community for advice on this.
Replies
Here's a decent place to start learning:
wiki: Polygon Count
specifically the Articles about Performance section.
You'll be wasting your time if you just take forum advice and never check PIX or any other performance tools.
Every engine is different, and the way you construct art assets will defer from the way you'd make an efficient asset in another.
Yes, we are taught to normal map, so I try to consider what I can leave in the normal map, and what needs to be in the mesh itself.
@Jon, Yes, that is very true. Engine, and hardware are critical. This is also the reason why I wanted to open up some discussion with the community to get a range of thoughts on the issue.
http://tech-artists.org/forum/showpost.php?p=9535&postcount=3
It's hard to be too specific here in terms of shaders and material costs. I can spit out shader percentage differences of our internal engines, but you'd really just need to verify with what engine/platform you're on.
Some general guidelines, however, these are not always 100% certain (YMMV)
-avoid small triangles of meshes, they render much slower,
-keep as much geometrical data stored in a normal map over putting it in the mesh, keep mesh details for silhouette
-learn how the platform you're working on handles draw-calls/state-changes, this HEAVILY impacts how you make your art, i cannot stress this point enough
-be sure to ask your local graphics engineer, especially if its your own engine, as they're the one(s) handling your art data, etc.
Last but not least, verify, verify, verify in some performance analyzer (this is more geared toward a commercial project).
The wiki is a great source that I have found about draw calls, texture map sizes, what to look out for in terms of resource hogs which are in fact not the poly's, but the various textures/materials.
Like I said I'm still in school, I'll be a senior next year so I'll still be using UDK. For me, its just trying to get a handle on the idea of creating good looking meshes, while not being inefficient on either end of the spectrum. Its something my school doesn't emphasis, so I'm trying to fill that gap in my skill set, and understand it so I can be more effective.
One of the nice things about UDK is that you can see some of the meshes made by Epic for the engine. So there's lots of great examples of how things should be done for that engine.