Open getrasa opened 6 days ago
I think a RTX 4070 should be able to handle this complexity without many issues. I must admit that i did not so far try looking super close into things, but i did not notice such a breakdown in framerate. So thats not expected.
Two things come to mind:
hey, same issue here but first of all thanks a lot for providing your URP VR fix with this repo. works great and is such a help for our prototypes. much appreciated.
I noticed the heavy performance hit when getting close to a model as well - but already in your demo scene (without any changes!) when moving a body part close to the view/headset - more or less independent of the quality.
the issue is the same with our own lightweight models (single layer, only 100k splats).
tested with quest3 standalone but also noticeable via oculus link (4070ti). haven't profiled it yet nor tested with the original repo by aras but it's on my list.
Sure. I’ve uploaded my model here: https://drive.google.com/file/d/10nfCFngpiSQnOdqyaVCU4azD36duYDRG/view?usp=sharing Feel free to use it for testing, although I believe that any larger model could be used to reproduce this issue or perhaps even smaller ones like fx-lange has mentioned.
I haven't tested it with the original repo yet, but judging by this issue I found on the original implementation, this doesn't seem to be VR-related. It describes the same problem: https://github.com/aras-p/UnityGaussianSplatting/issues/10
Hi,
First, I want to thank you for this amazing project. I've encountered a performance issue that I'd like to report.
Setup: Unity Version: 2022.3.51f1 Scene: "Test" scene Changes Made: Replaced the default gaussian model with my custom model (350k splats) Settings: Tested on both High and Medium quality of gaussian splat import
Issue: I'm experiencing significant performance drops when the VR camera gets close to my custom gaussian model. This occurs regardless of whether I use High or Medium quality settings.
Additional Information:
Is this a known issue? How do I resolve it?
Any help is much appreciated!