HISM + Nanite and still performance issues

[RecourseDesign]

The freeze rendering command doesn’t work with hardware or instance culling in UE5 though - you can test that by freezing the rendering and then moving around a level full of instances that have a distance cull, they still cull at correct distance from the camera.

Also there is still a large performance increase when things are occluded. I’ve been running some tests are these are the results so far:

rdBPtools Tutorial 22 - Page 3 of 6 - Recourse Design ltd.

But that wouldn’t explain UE4.27 not culling which I’ll test soon.

I do wonder if there are some situations where it isn’t working - it could explain some peoples tank in performance - it would be nice to get a complete picture.

[RecourseDesign]

nice! what hardware is that running on?

[Suthriel]

It´s a slightly older machine:

AMD Ryzen 7 2700X Eight-Core Processor with 3.70 GHz
16 GB RAM
NVIDIA GeForce RTX 2070, very first generation (no Super or TI or anything), just 8GB RAM

And Unreal Engine 5.2 ^.^

[ruigol (Deep Pixel Studio)]

Hey guys, quick update. It took me all day but I did rewrite all of my code (from spawning BPs first and removing elements as I go to setting the cell parameters as I go and spawning the maze at the end) and now have my system working perfectly. I add all of the HISM instances in 1 blueprint and now with nanite and HISM culling I get the exact same fps/ms performance no matter how many cells I spawn. The only limitation seems to be when it first loads, some large grid sizes cause it to crash but for what I wanted it for it’s perfect, thanks to everyone who commented and helped me get this far!

[Rev0verDrive]

You will get better initialization performance if you lower the grid coverage per BP.

Say a grid cell is 5m x 5m. A single BP would cover a 20 cell x 20 cell world area (100m x 100m).

A full maze (1Km x 1Km) would then be made up of 100 BP’s. You can use netcull distance of the pawn to load/unload BP Cells.