FPS drops in viewport that then returns after a few minutes.

Naeem_Girach · May 16, 2025, 2:48pm

Hi All, I have been using UE5 for a year or so but predominately for ArchViz. I started a VR project for a large-ish environment of a town and field that has game elements and UI systems and scene sounds etc. I got my FPS up to between 60-90 FPS depending on what the camera was viewing, however, my UE5 Viewport’s FPS goes through this almost cycle of running at 80fps and then dropping all the way down to 20fps. It stays at this low fps for sometime before eventually returning to the higher FPS. Then the cycle repeats after 5 minutes or so. This has been a common theme in my project even when the scene was significantly less populated with high poly mesh. I’ve tried many things to get my FPS up to what it peaks at now, such using lowering texture size, using LODs, culling, disabling ray tracing, lowering scalability settings, utilizing nanite etc. Doing this got my scene from peaks of 30fps up to 80/90 (current fps peak) I’m not an advanced user of UE5 so I was hoping somebody could let me know what the possible cause/s could be. Just an FYI, my hardware specs are ryzen 5900, RTX 3080 10GB, and 64GB RAM. It does say that I’m out of VRAM on some occasions but it still allows me to get to the higher FPS. Wondering whether my issues are purely due to hardware limitations.

Thanks

brs-sebascova · May 16, 2025, 4:07pm

Hello there @Naeem_Girach!

You are taking the right approach, first suggestion would be to use Unreal Insights, in order to better track your project’s performance. Try tracking memory use, checking for garbage collection events, increasing cleanup frequency, and reviewing object lifespans in your scene.

And regarding hardware, your setup looks powerful enough. If you are running out of VRAM, it can be resolved by optimizing your scene further. Please test implementing texture streaming, following the guides below:

github.com/JaredP94/Unreal-Development-Guides-and-Tips

Content/DevPipelines/TextureStreaming.md

master

# Texture Streaming

A summarised guide on the concepts of texture streaming, increasing the texture streaming pool size and disabling texture streaming.

## Texture Streaming
* This denotes the detail of the textures which are to be viewed.

* Texture streaming is responsible for handling the transition between different mipmaps as the camera distance is changed.

* As the camera moves closer to the texture, the texture streaming pool will become more full due to the larger mipmaps being streamed.
   
## Increasing Texture Streaming Pool Size
* Warnings may arise when attempting to render extremely high detail textures within the scene. This is typically common in ArchViz projects.

* This can be mitigated by increasing the texture streaming pool size in two ways.

* The first method entails using the *Console*, which can be opened with the **tilde** key, with the command:
	```cpp
	r.Streaming.PoolSize = [DesiredSizeInMB]
	```

This file has been truncated. show original

Naeem_Girach · May 17, 2025, 11:01am

Thank you, I will take a look at the UE5 insights. I’ve already looked at changing my texture streaming pool but after viewing that YouTube video I feel I’ve probably changed it to something too high (I did 10000) to match my graphics card memory size.

This may sound silly (and maybe obvious), but if I have google chrome open, with multiple paused YouTube videos in tabs, and one YouTube tutorial playing, is that what’s causing these issues? I typically use UE5 alongside a UE5 tutorial on my other screen, or sometimes I watch shows at the same time. Is that taxing my hardware?

brahmaforge · May 17, 2025, 11:58am

Hey @Naeem_Girach — welcome to the forums, and major props for diving into VR game dev from ArchViz! That’s a big leap, and you’re already tackling it with the right mindset.

The FPS drop cycle you’re describing (high FPS, then dipping every few minutes) is something we’ve seen a lot in large, asset-rich projects — especially in open environments with complex streaming or sound setups.

A Few Likely Suspects:

1. VRAM Paging / Texture Streaming Bottlenecks
If you’re occasionally getting “out of VRAM” messages on your RTX 3080, that’s a signal that Unreal is aggressively streaming textures or geometry in/out. Even if it recovers, those dips happen when it’s swapping large assets.

→ Check:

Console: r.TextureStreaming (try toggling it)
GPU Profiler: Ctrl+Shift+, → See what spikes during the dip
Disable high-res UI textures or lightmaps temporarily to isolate

2. Garbage Collection or Actor Ticking
Periodic GC can cause big hitches — especially in VR where consistent frames are critical. Similarly, sound cues or UI systems (especially widget tick/update logic) might be contributing.

→ Try:

Stat Game / Stat Unit to monitor CPU, render thread, and game thread
Ensure no unused actors or Blueprints are ticking every frame unnecessarily

When Learning Becomes Overwhelming (and How to Fight It)

Since you mentioned you’re still getting comfortable with UE5, here’s something we often see:
You watch 5–6 tutorials on YouTube — each showing a different optimization trick, Blueprint setup, or system. You try to follow along… and by the third video, your project’s already drifted from where you started. And worse? You forget which things you’ve implemented, and which you only meant to.

That’s exactly why we built Asset Optics.

It lets you:

Add comments like “Copied LOD setup from YouTube tutorial” or “Test Nanite toggle here” directly on the asset
Use checklists to track what you’ve optimized, tested, or still need to experiment with
Sync everything to a web dashboard, so even if you step away from your project for a few days, you come back knowing where you left off

We like to think of it as your “learning and debugging assistant”, especially when juggling a dozen changes across levels, Blueprints, and assets.

You’re clearly already thinking like a technical artist — balancing fidelity and performance. Stick with it, and keep narrowing in on the bottlenecks one layer at a time. Let us know if you narrow down the culprit — the whole community can benefit from those findings.

And if you ever want to share your performance tracking checklist, we’d love to see how you’re approaching it!