Given that the Reflex 2 of Nvidia GPUs appears to be a form of frame extrapolation, can it potentially be used as a method of providing VR users a hardware-based motion reprojection?
Although motion reprojection is already present in some/most HMDs the implementations are often severely artifacted and suffer from frame timing issues etc. (I think they rely on nvenc encoders?).
From what I can see from the Nvidia slides, the only difference appears to be that it’s responsive to mouse movements rather than HMD movements, which are closely related.
Since MSFS shares the motion and depth buffers - is there an opening here? since DLSS frame-gen doesn’t benefit VR users?
Perhaps @mbucchia can shoot this thought balloon down before it gets too high.
I know I’m reaching for something that is not at all there, but I at least wanted to start a discussion.
Thank you for your post! Your topic has been moved to a sub-category of the User Support Hub
The General Discussion category is meant for discussions that fall outside of our other sub-categories.
If you would like other users to help you with an issue you are experiencing in the sim, consider these User Support Hub categories for your future post:
I don’t know if this reflex tech is what VR needs as much as the new 3 frame FG. That would be a VR savior as theoretically you get very close to flat screen frames in VR if that worked. I am sure it won’t, but just imagine if it did.
You might have missed my point, and that’s probably my fault. Frame gen is exactly what reflex 2 does, but in a way that is amenable to VR because it is extrapolation. Meaning it is predicting frames based on movement. Exactly what ASW, motion smoothing, etc do. Exactly what VR needs. And based off their “75% reduction in latency” it appear that this extrapolation maybe be 3 frames for every generated frame - multi-frame gen in the forward direction. I realize Reflex2 is not at all intended for VR, but its a step in that direction. Maybe Reflex 3 will be for VR.
I posted on the 5900 thread that it just might be possible for Reflex 2 to be used in VR for motion smoothing frame generation. Right now it uses input from the mouse to predict where the view will be in the future. If that input can instead be changed to the headset movement, then it should work in VR. Whether it is possible to make that change in the driver or in the sim I have no idea.
