been digging a bit 5 features. and when i see that nanite stuff, i’m really impressed. but in the same time it seems so great that there’s a little voice saying into my head can we really trust that thing. I mean according to the courses i’ve been watching, this is non less than a brand new level in game developping. i mean no more lods, and drawcalls concerns… just drop the textured mesh, and you’re good. what are your thoughts. Is it that easy now with 5? just drop any mesh, and it’ll work. What do you think, will you blindly trust nanite for your project
You are allowed to test it, you know… 
of course, but my question was more a global one. like if there were some drawbacks to such promising system. It looks so good on paper, and it probably is, but well… I come from the modding. I’m a bit old school 
Personally, I think it won’t be any use for your average gamer for a while.
Most people out there have Nvidia 1000 series card or less ( or similar ), the vast majority in fact 980 and below.
In my experience, you have to drive things so far before you see any benefit, it’s not worth it with these kinds of cards.
1 Like
the problem with nanite it encourages the devs to create ultra high poly assets. and even if nanite does its job pretty well you still have ridiculously high poly assets even after being reworked. here’s an example i downloaded an asset pack. i took a road sign that is natively 150k poly. if i try to reduce it with nanite, it does the job pretty well but it’s not the magic solution neither. as you reach a point where or you still have a ridiculously high poly mesh, or you have an ugly mesh. i tried with a sand bag also. and the result was the same you have the choice between unacceptable poly, or unacceptable topography mesh. so at this particular point i’m a bit mitigated
I agree. People don’t seem to understand the use cases. Everybody on the market is applying Nanite to everything regardless, like it’s some kind of magic pill. It’s not.
I downloaded a good asset the other day ( not naming names ), but things like CUPS had Nanite applied. And this is a dev who knows what they’re doing.
Unless I’m going to build a 5 million cup sculpture, this is useless to me.
Nanite’s not an excuse for bad modelling.
1 Like
i think i got how nanite manages so well complex scenes. there’s only 2 LODs and culling. and if not very close the mesh is LOD 1. anything that is not on screen is culled. this way you can have a nanite managed scene with much less polys than the same scene with 4 LODs.
I made a test with 10 M polys. with 10 x 1 M polys meshes on a small map. and i got 0 FPS drop.
for someone like me who comes from modding, and who is used to optimization, a 300k polys mesh sounds pure madness. but it’s not like that one should think of it. it’s the number of polys of the whole scene. if ¾ of your scene is culled, and in the last ¼ 98% is lodded, it’s ok.
Then like any systems it has limits too. But one has to use his/her brain a little bit too