I’ve read posts about it being unnecessary to instance materials and meshes. Do we still need to virtualize textures? Do I still need master materials? Is trying to reduce draw calls still a thing? Do I need to pack UVs still? Should I still merge meshes into static mesh chunks in the environment? I’m sure all these methods exist but are they worth the time with nanite and lumen now a thing?
I plan on creating a hyper detailed city multiplayer FPS map about the size of a CoD environment utilizing ultra high poly assets and materials with dynamic and destructible props, many of them. I just need to know how to go about creating and texturing my assets before I get too far into this. Thanks
I can say from experience that using virtual-textures tends to just-be a net benefit. Essentially you tile/chunk the various mips and stream into memory what you need as you need it. It’s like an object-pool but for parts of textures that you recently-viewed.
Virtual shadow-maps extend this concept to, well, shadow-maps, but also they auto-mip into the distance so they work to not render at resolutions where you won’t need them; they essentially auto-lod the shadow map.
Master materials, and all the material-tricks still work and apply with nanite. Especially with master-materials. Materials are binned, as in all the pixels for that material are drawn in one-pass. And I want to say I recall hearing on ‘some talk’ (I watch a lot) it also applies to things you can change in the instances, so it’s the same root-material that is counted for a bin, and variations.instances don’t count against you (unless there is a static-switch which essentially make s new material), Someone correct me if I am wrong here I cannot confirm either way from what I search for…pretty sure it’s ‘in a vid’ so…
WPO is expensive for nanite, and mask-materials have some caveats, but they work, you just pay for it.
All the things above, they can work with regular geometry but are made to best-work with nanite. Again, like just turning on lumen, nanite, etc you incur decent cost up-front to just turn stuff on, but the ideal is they help you scale to better quality-assets with ultimately less overall-cost. You’ll see a decent decrease in FPS just enabling stuff, but you’ll be able to top out on some costs as you fill you scene.
Yes, you might have to be flexible, and YMMV depending on your pipeline (particularly w/regards to foliage), otherwise it’s workable even if you have to adapt a bit.