Prediction Settings Feedback
After 6 months of playing with the prediction slider like a small child I'm confused that setting prediction to zero always seems to provide me with a smoother experience. This just doesn't make sense, but it has applied to DK1 and now to DK2.
For anyone unaware, Oculus has a wonderful article here explaining their approach to prediction. After reading it I was really stoked, as it is convincing and feels like another stone in the path towards working VR.
On the forums I often hear people say the default of 30ms is perfect for them, yet I also wondered if they're compelled to tweak the slider until it proves its purpose like I am. But when I started developing in VR I didn't find prediction to be as beneficial in current practice though. A default setting has a swimming quality, making me and others feel ill. When set to zero the effect disappears. It is noticeably faster with head movement without a hit on performance.
After watching the 2014 Oculus Unity talk I saw that prediction was being handled in a new way with SDK 4.0. This 'dynamic prediction' requires VSync and is tied into Time Warp.
Recently I built a scene with hundreds of objects orbiting another object. This scene is getting 250fps in DK2. When using prediction of 30ms it is judder-y with a delay in head movement. Same scene, same run, turning prediction to zero, the scene blends into presence (shocking me enough that I dodged one of the objects).
Given that it looks like VSync is causing issues in Unity, is this a VSync thing that I'm blaming on prediction? Is it Time Warp's judder? Or some unholy amalgam of all three? Why did this occur on DK1 when I wasn't using VSync?
Love to hear people's experiences to gather more data. Thanks!
For anyone unaware, Oculus has a wonderful article here explaining their approach to prediction. After reading it I was really stoked, as it is convincing and feels like another stone in the path towards working VR.
On the forums I often hear people say the default of 30ms is perfect for them, yet I also wondered if they're compelled to tweak the slider until it proves its purpose like I am. But when I started developing in VR I didn't find prediction to be as beneficial in current practice though. A default setting has a swimming quality, making me and others feel ill. When set to zero the effect disappears. It is noticeably faster with head movement without a hit on performance.
After watching the 2014 Oculus Unity talk I saw that prediction was being handled in a new way with SDK 4.0. This 'dynamic prediction' requires VSync and is tied into Time Warp.
Recently I built a scene with hundreds of objects orbiting another object. This scene is getting 250fps in DK2. When using prediction of 30ms it is judder-y with a delay in head movement. Same scene, same run, turning prediction to zero, the scene blends into presence (shocking me enough that I dodged one of the objects).
Given that it looks like VSync is causing issues in Unity, is this a VSync thing that I'm blaming on prediction? Is it Time Warp's judder? Or some unholy amalgam of all three? Why did this occur on DK1 when I wasn't using VSync?
Love to hear people's experiences to gather more data. Thanks!
