Hacker News new | past | comments | ask | show | jobs | submit login

Godot is using a more traditional graphics pipeline, and sending mesh data to the GPU after having the CPU do initial calculations for LOD. Nanite is designed as a "GPU software renderer" and therefore performs fine-grained LOD calculations "during" its rasterization, within the shader code. That's oversimplifying the exact steps, but it explains the orders-of-magnitude difference between Nanite and traditional approaches since doing more on the GPU means there aren't as many data transfer bottlenecks.



If data transfer bottlenecks are the issue, would using a hybrid CPU/GPU (Intel IGP, Apple Mx) setup not solve the issue?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: