Forum Discussion
Anonymous
10 years agoNot applicable
5.1.2p2 -> 5.2.1f1 - Large Performance Loss
Long post short (i.e. tl;dr;): Avoid 5.2.1 for now. Major performance loss. I'm sure nobody will read the rest of this since Oculus Connect is happening now, but if you want more details, read on! ...
Anonymous
10 years agoNot applicable
Dave B - thanks for responding. I'm glad you guys are on it. Hopefully some of this data will help track down what the issue is. 5.1.2f1 (and 5.1.1p3) actually have performance issues as well that I reported about a while back that were fixed in 5.1.2p2. I have included more data on this below. Here is the bug fix related to that performance fix (seen in 5.1.2p2 release notes):
(none) - VR: Improved MSAA performance by only antialiasing eye textures, not the composited back buffer.
In addition 5.1.2p2 included a ton of other VR fixes. Based on that, I'm surprised you guys still recommend 5.1.2f1.
I have improved my benchmark a bit by trying to make it more consistent and I now run it in built application instead of in editor. Then I spent more time retesting most of the latest versions of Unity to track down what works and what doesn't. Here are the results (higher ms is better, indicating more CPU time for your application):
5.1.2f1 7.1ms
5.1.2p2 9.2ms (What I have been using for almost 2 months now as I found it most stable)
5.1.2p3 9.2ms
5.1.3f1 crash every time - can't benchmark
5.1.3p2 9.4ms (BEST! I will probably move to this version)
5.1.3p3 6.8ms
5.2.0p1 6.5ms
5.2.1p1 6.7ms
Based on this, it looks like a major rendering performance issue was introduced in 5.1.3p3 and beyond. I don't see it noted in 5.1.3p3 release notes, but it appears this is when Unity moved to 0.7.0.0 dependencies because the performance monitor started tracking app CPU/GPU time in this version. In previous versions, this data was still N/A. Hopefully the performance drop is fixable on Unity side and not some indication of performance loss with 0.7.0.0 API and runtime.
edit: note all of the tests above were run with 0.7.0.0 runtime, 0.1.0 utils, and Nvidia 355.98 drivers. Only variance was Unity version.
ccs
(none) - VR: Improved MSAA performance by only antialiasing eye textures, not the composited back buffer.
In addition 5.1.2p2 included a ton of other VR fixes. Based on that, I'm surprised you guys still recommend 5.1.2f1.
I have improved my benchmark a bit by trying to make it more consistent and I now run it in built application instead of in editor. Then I spent more time retesting most of the latest versions of Unity to track down what works and what doesn't. Here are the results (higher ms is better, indicating more CPU time for your application):
5.1.2f1 7.1ms
5.1.2p2 9.2ms (What I have been using for almost 2 months now as I found it most stable)
5.1.2p3 9.2ms
5.1.3f1 crash every time - can't benchmark
5.1.3p2 9.4ms (BEST! I will probably move to this version)
5.1.3p3 6.8ms
5.2.0p1 6.5ms
5.2.1p1 6.7ms
Based on this, it looks like a major rendering performance issue was introduced in 5.1.3p3 and beyond. I don't see it noted in 5.1.3p3 release notes, but it appears this is when Unity moved to 0.7.0.0 dependencies because the performance monitor started tracking app CPU/GPU time in this version. In previous versions, this data was still N/A. Hopefully the performance drop is fixable on Unity side and not some indication of performance loss with 0.7.0.0 API and runtime.
edit: note all of the tests above were run with 0.7.0.0 runtime, 0.1.0 utils, and Nvidia 355.98 drivers. Only variance was Unity version.
ccs
Quick Links
- Horizon Developer Support
- Quest User Forums
- Troubleshooting Forum for problems with a game or app
- Quest Support for problems with your device
Other Meta Support
Related Content
- 4 years ago
- 4 months ago
- 5 months ago