Nope. VR headsets aren’t monitors, you have a lens between you and the screen which creates distortion...
Like I said above the headset is designed to work with a render resolution of over 3000x3000 per eye... that’s ~9 million pixels each or 18 million total vs 8.3 million of a 4k screen.
In a really quick high level explanation - in order to have an image look correct after it has passed through the lens it needs to be pre-distorted (called barrel distortion as the end result looks barrel shaped). Since there is currently no mainstream rendering pipeline that outputs a predistorted image, you have to render the scene traditionally to a buffer. That image is then subjected to barrel distortion which involves stretching and enlarging the image. That process requires you to render at a higher resolution otherwise you will end up sub sampling the parts of the image that are being stretched.
Forget what you know about panel resolutions from flatscreen. There is no “native” resolution for VR with current pipelines, if you set 2160x2160 per eye then you are subsampling the entire frame after distortion is applied and clarity suffers significantly. The “optimal” setting is enough that you get a roughly 1:1 rendered to displayed pixel in the most distorted area of the image, so gives the best image quality but still isn’t “native” as you are supersampling all the less distorted areas of the image. On the G2 that optimal setting is a little over 1.4x the linear panel resolution (changes per subpixel, but technically it would require on average around 1.5x to get every last pixel at least 1:1). That is what steam will set as the 100% target for the G2. I forget exactly but it’s something like 3160x3072 per eye.
It’s not quite as straight forward a comparison as that as you can use hidden area masking to reduce (but not eliminate) render load for areas that will be off the screen anyway after distortion which helps a little, but start including that and it gets pretty confusing pretty fast. Then there are certain other technologies that can complicate direct comparisons further like Nvidia’s SMS or the upcoming sampler feedback in DX12U. As a basis for general comparison it’s close enough though - point is you can’t necessarily precisely compare the 18 million vs 8.3 million numbers.
You can run lower render resolution by setting less than 100% resolution but you aren’t then getting what the headset is capable of. Anything less than about 75-80% becomes very noticeable to me in terms of clarity fall off. Between 80-100 it’s still noticeable for me but more subtle. Above 100 you just get improvements to aliasing really rather than more detail.
As mentioned the normal course of action for a game that you can't get to run consistently at 90fps would be to use motion vector reprojection to go from 45fps to 90fps in the headset. To work most effectively it does require some headroom, so you'd ideally be able to hit a minimum of 60fps with it disabled and then it'll give you a solid 90fps and do a pretty damn good job in most games of giving you good quality interpolated frames.
Anything else, totally depends on the individual. As mentioned I lock to 30fps with no motion smoothing in MSFS and while I'd like it to be faster it doesn't cause me any sickness... but I've never suffered VR sickness anyway. Others might find that very uncomfortable. If you are going that route of sub 90fps but no motion vector smoothing then it's important to lock the framerate to a sub division of 90 - ie 45 or 30. At 30fps I get each frame displaying for 3 refreshes, so while it's crap (especially looking out the side window) it is at least consistent. If I uncapped with my current settings I'd get around 35fps which would result in some frames displaying for 3 refreshes and some frames displaying for 2, which leads to a much more stuttery result despite the higher FPS.
Last edited:
