Anonymous performance data per play session is now sent to the server. The client-side overhead of this is very low, but it will allow us to use the gathered data to try and help optimise settings for specific users, and attempt to find the causes of large numbers of "stutter" or "spike" frames for a small subset of the community.
To give an example of what we can learn: In initial testing (Cutting Edge) we are seeing that roughly 92% of play sessions have zero stutter frames. 5-6% of the remaining sessions have only one stutter frame.
If you are curious what we are sending, the raw data is saved to Logs/session.log (each play). We will be publicising a real-time aggregate feed of this data in the coming weeks!
To better track frames which are long enough to affect gameplay, spike frames are now considered 30ms or higher. This is still at a point where it is considered "recoverable, but annoying". We will be using this as our benchmark going forward.
What we learned from these: That we can't trust feedback from users. There's too many variables, subjectivity and placebo involved. Going forward we will be tracking and publicising (anonymous) performance information for each osu! update so you can be sure the update has not changed osu!'s performance profile.
As part of our effort to improve performance for a small group of affected users, and to remove any subjectivity, a very simple A/B testing framework has been added. This will allow is to test various micro-changes and observe if they help with performance.
We will do our best to not make this noticeable in normal play. If any negative effects are notices we will immediately disable a test server-side.
Support further development of osu! and become an osu!supporter today!
Not only will you help speed development, but you will also get some extra features and customisations!
do you have it turned off in settings?
Comments80