Polygon Budget for Next-Gen Games
So the new age of gaming is upon us, and I wonder what budgets we are looking at.
Suggestions?
-Polygon budget for 3rd person action game with lots of enemies. (A bit like Warframe)
-Polygon budget for 1st person multiplayer game (Budget for 1st person weapons, enemies, and other player models)
-Polygon budget for 3rd person action game with little enemies (A bit like Deep Down)
-Total polygon budget for environments.
-Polygon budget for "hero" environment pieces.
Suggestions?
-Polygon budget for 3rd person action game with lots of enemies. (A bit like Warframe)
-Polygon budget for 1st person multiplayer game (Budget for 1st person weapons, enemies, and other player models)
-Polygon budget for 3rd person action game with little enemies (A bit like Deep Down)
-Total polygon budget for environments.
-Polygon budget for "hero" environment pieces.
Replies
There are simply way to many factors invovled to give single definite answer for that. ie. physically simulated meshes should generally have lower vertices count, because of simulations complexity.
In any caes, things like lighting, complex materials and amount of individual objects (draw calls), will bottleneck your game long before, some triangles will.
And yes, I agree that there are far more process intensive things than triangles - I think Warframe is a good example. They spend a lot on PhysX stuff, real time reflections, dynamic lighting and so on, and the polygon count of their stuff seems really low.
If you are making simple chair with 4000 polys you are clearly not game dev artist (;.
You should just be using few simple rules:
1. Add as many as geometry you need to keep silhouette.
2. On top of that, add small details that enachne close look and/or improve normal map baking. (ie. you can chamfer edges if that will help with normal map). Or add floating details like holes or big scrubs.
3. Add geometry that will help split mesh on per material basis, so you can add physically correct, tileable materials where you need them on polygon basis, instead of using complex masks.
Blending and masking materials based on textures is usually more expensive that using multi-materials.
Oh. And use LODs. Most engines support some form of auto-magic LOD generation. I think most people forget that such feature as LoD exist (;.
Star Citizen:
Fighter: 300k.
Carrier: 2mln.
Character: 150k.
I think my rule of thumb here is if you actually notice it in game, then it's a go. But I've seen models with such a tiny amount of detail that the actual result in-game is barely even noticeable.
Yeah, that would be awesome. It would be great to know the specs of recent games, like Destiny. But I'm thinking that kind of info is hard to find.
Thanks for the detailed reply. To add to your #2, I've done that a few times and then I actually get rid of the bake-helper edge loop, so you get the nice bake and you can actually then make it low again. It works pretty well.
And WOW, that's a lot of polys for Star Citizen! Man!
http://www.guerrilla-games.com/presentations/Valient_Killzone_Shadow_Fall_Demo_Postmortem.pdf
"Around 40k polygons for the highest LOD
‣ Enough to capture all detail for closeups
‣ We provided detail guide for LOD setups
‣ Up to 8 bone influences per vertex
‣ Most vertices use 4-5, drops with LOD#
‣ 6 x 2k x 2k textures for character body
‣ Plus detail maps and head textures
‣ 10ppi, everything authored as 4k
‣ KZ3 used 10k polygons, 3 LODs and 1k
textures"
Although as others would say the polycount is only one of many things that determine performance, many other things weigh in more into how well the game will perform.
iniside listed some good rules for modeling.
This PDF is awesome. Really good info here!
That seems to be the consensus here. Good judgement calls + Freedom = Awesome.