gamekiller48.bsky.social
@gamekiller48.bsky.social
Someone forwarded this world to me: vrchat.com/home/world/w...

Maybe that might be worth checking out for testing?
VRCMark v2 by Hellcat|へるねこ
[VR Only] VRChat用のベンチマークワールドです。 CPUとGPUに意図的に負荷をかけ、その性能を測定します。
vrchat.com
September 8, 2025 at 11:58 PM
Any test setup would likely have to be run in VR + in instances with other players, but at that point the chaotic nature of other player's movements would make a comparative test impossible.
I don't really have any good ideas for how one could work around that limitation.
September 8, 2025 at 3:33 PM
I unfortunately have to conclude that my testing setup likely is flawed, and that in these circumstances (single person in a world, running on desktop) the differences are more likely explained by measuring error and me as a player not being able to replicate movements 100% exactly in both runs
September 8, 2025 at 3:31 PM
I've retested, this time on the fairly popular Complex 7 map (vrchat.com/home/world/w...), which has a lot of textures and a few animated NPCs.
The good news? the visual bug in my menu disappeared, so it might have been a random glitch.

Less good: my test results are still all over the place.
Complex 7 by Fins
The sequel to District Roboto‚ Complex 7 features a world 7 times larger‚ full of NPC‚ collectible‚ unlockable avatars‚ and hidden secrets․ Discover the dark history behind this technicolor world of l...
vrchat.com
September 8, 2025 at 3:27 PM
Unfortunately no, I'm using MSI Afterburner for logging, which doesn't seem to have an option for logging frame latency.
I'm still trying to figure out what would make for a good test world. A photo world to test texture impact and an MMD world to approximate a number of avatars on screen?
September 8, 2025 at 10:36 AM
Also I wrote "frame time improvement" in that excel cell, it should obviously be "frame time change", as indicated by the red colouration (the frame times increased, not decrease)
September 7, 2025 at 11:52 PM
I think I'll try to find a better map for testing tomorrow, but it's hard to find a map that pushes the fps below 144 on desktop (and testing in VR is very difficult, since it is a lot harder to control character movement to be compareable between runs).
If you spot any mistakes in my test setup lmk
September 7, 2025 at 11:50 PM
With the settings recommended by you for the profile inspector, it causes a observable visual difference in VRchat: The VRchat menu should be transparent, it is now opaque.
Looking back through the list of changes, the change to texture filtering seems like the most likely culprit.
September 7, 2025 at 11:48 PM
As you can see, frame time data and fps data for some reason don't seem to line up. I'm not entirely sure what could be causing this issue. That's however not the biggest problem.
September 7, 2025 at 11:44 PM
I just gave it a testrun on this map: vrchat.com/home/world/w...
The results were, for this particular test, not good.
It's an unfair test, since gaussian splatting maps have almost no textures, but many transparent elements. I didn't think the test through, but I'll share the data anyways:
- 川北東橋 ⁄ Kawakita-higashi Bridge - 3DGS by DEKA_KEIJI777V
このワールドはテスト用です。【場所˸徳島市 川北東橋・撮影日˸2023年5月2日・Splat数˸300万・撮影枚数˸約780枚】当時撮影データをフォトグラメトリに使用する予定だったものを試しに3D Gaussian Splattingに使用してみました。ノイズが多いですが楽しんでいただけたら嬉しいです。
vrchat.com
September 7, 2025 at 11:41 PM
Thank you <3
July 27, 2025 at 9:39 PM