Nanite Performance is Not Better than Overdraw Focused LODs [TEST RESULTS]. Epic's Documentation is Dangering Optimization.

Because Nanite has an overhead, but once you get above that

yeah, base your logic on a UE5 dev meme that got spread around due to ignorance o the topic:

This admittedly poor description of Nanite’s performance is 100% wrong. The act of enabling Nanite only cost around .54ms at 1080p on a Desktop 3060. Mesh shaders are extremely inefficient & SW rasterizer is slow and scales with complexity because of how inefficient cluster culling is. None of your sources references properly prepassed content performance either. The content you shared used butchered content/rendering approaches for non-nanite to premote performance on 4080 hardware. You gave no side by side comparisons of mesh shader & SW raster performance for foliage either in any of your post.

SW raster grass will perform worse than HW shaded masked materials because my team proved how much faster mesh shaders are.

Nanite should generally be enabled wherever possible. Any Static Mesh that has it enabled will typically render faster, and take up less memory and disk space.

So the documentation is very clear to not use it if using only low poly geometry

No, it will not “typically” run faster & official presenters are saying the opposite of your conclusion: https://youtu.be/Cb63bHkWkwk?t=5496

It’s also manipulating for the docs to say a prepass is 75% faster than a basepass to jusifiy Nanite rendering. Epic should include non-vrs shading rates based on quad rates for prepass & non-prepassed content vs the visibility material resolves & vis buffer raster methods.
My team provided this for 5.6, but it should be standard for each version on the docs.
Epic should also fix the viewmodes for Nanite for anyone crazy enough to use it.

Nanite is a constant bottleneck for games, the documentation is a part of that problem but more so the people who think this topic is based off “opinions”. The facts have been provided, people