Search Issue Tracker
By Design
Votes
2
Found in
2022.3.5f1
2023.1.4f1
2023.2.0a23
Issue ID
UUM-42837
Regression
Yes
Performance drops when using a lot of Audio Sources
Reproduction steps:
1. Open the attached project “AudioSourcePerformance“
2. Open the Scene “Assets/Scenes/SampleScene“
3. In the Game View press the “Stats” button
4. Enter the Play Mode
5. Press the button “Play All”
6. Observe the Statistics window
Expected result: The Scene is running without performance issues
Actual result: The running Scene has a low frame rate, high DSP load, CPU usage is high
Reproducible with: 2022.1.0a5, 2022.3.5f1, 2023.1.4f1, 2023.2.0a23
Not reproducible with: 2021.3.28f1, 2022.1.0a2
Couldn't test with: 2022.1.0a3, 2022.1.0a4 (Unity crashes when trying to enter the Play Mode)
Reproducible on: Windows 11
Add comment
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
Latest issues
- Collapsing an empty VFX Subgraph Operator with contextual menu removes the arrow
- TMP_FontAsset fails to change when the Fallback Font Asset is changed
- Property values are not restored for all objects when re-enabled after reset
- "Sub Mesh Mask" field in VFX Graph has no padding
- Source Generation with GeneratePropertyBag fails when using reserved keywords as variable names with @ prefix
Resolution Note:
Thank you for bringing this to our attention, we can verify that there is indeed a performance hit in this case.
Going into 2022.2 we modified the audibility system internally so that it would work with send/receive on mixers. This modification gets quite expensive when the underlying audio graph is large (as is the case with this issue). We have plans internally to re-work the audibility system so that it is more performant, but it is sadly not a quick fix as this part of the audio system is complex - so any changes come with a high amount of risk.
If at all possible we recommend trying to lower the audio source count (2500 is pushing the audio system very far!) as well as lowering the voice limits from their maximum.