Optimal resolution for VR calculation / VR aspect ratio

[matemaster]

After tinkering with every kind of supersampling in and outside of game I came to definitive conclusion. It pretty much does not matter if you use SS in game or SS in VR or HMD quality in game... The only thing that matters is the final resolution.

In steam VR supersampling settings you can check the actual resolution lets say its 3640/3200 then you multiply both numbers by in game SS and in game HMD quality. Lets say SS ingame 0.85, HMD quality 0.75 the final resolution in game would be 2321/2187 by adding those values you get 4507 pixels this is the value that matters.
I found out that for my PC spec I am aiming for value around 5600. No matter how I achieved this value (tried steam VR SS only / only in game SS, in game HMD...) it does not matter the FPS is the same, same quality (if there are some differences those are negligible).

Now new question arised from this does aspect ratio matters ? Is it better to have more horizontal res for VR than vertical ?

 

[TorTorden_VR]

Aspect ratio simply matters that you do not want any optical distortions, tweaking vertical over horizontally will go no where good adjusting FOV on the fly is not something we are equipped with either since it closely ties with the actual lenses in the HMD and is hardcoded to the VR device drivers.

And at the end of the day, the displayed image is shown in the panels native resolution.
Rendering higher pixels than what can be displayed is merely a very costly, but high PQ antialiasing, and whatever benefits are merely percieved and not actually real.

The benefits is merely like having a picture of a certain resolution, then printed out on a lower DPI.
The end result is still a less resolved picture and on top of this we get Moire artifacting, the more wierd SS you apply, the more Moire you risk suffering from all the unnecessary steps added to the rendering process, now the panel resolution in Vive or CV1 is so low, you actually need a rather trained eye to detect moire artefacting. I happen to have that from my days as a printer (rotogravure, smeared paint on a lot of Kraft potatochip bags)
One reason I keep telling people to find ONE place to set their SS, and only that one place.
In game or with steamvr\oculus debug tool.
Pick one, and adjust that one to your suited preference.

On top of this, we still haven't confirmed the order of operations for supersampling, between runtime settings and in game, or even between the two settings in game for those that have multiple settings.

It is also for this reason you can actually find yourself getting a better result using less SS than constantly going higher, you aren't really gaining anything, but you are trying to match a grid of higher density to the grid formed by the panels.
When you also account for the added performance of rendering lower compared to higher, it's a fairly easy choice.

Yes going higher seems sharper, this is merely because the native edge detection going on in the HMD gets more pixels to work with, you don't actually get more detail, hence why I personally prefer a low to no SS rather and the framerate stability this gives rather than chasing details that will never actually show up.

Any SS adjustment for current panels outside of the marginal benefit of high pq AA is honestly a red-herring.
Either the native resolution is good enough, or not.
You can't go beyond the physical even though a perceived clarity increase is there. It is merely percieved, there are no actually new pixels shown in the details.

And I have tested SS ranging from 1x to 3x, granted the latter was simply not playable at all, but after a final SS of 1.33 any further benefits was neglible at best.

 

[loribee]

I'm still trying to figure out that sweet spot that does not peg the GPU on the Steam Frame timing to 100% in space stations .. is this a futile excercise? Are stations just that badly optimized? out in space, my frames are a constant 66 (which is the max for the Pimax)

Pimax Supersample set to 1.5
Steam SS to 1.0
ED SS to 0.50
HMD to 0.50

Shadows, occlusion and AA off

my CPU is an i7-3770k, Overclock to 4.3Ghz.

In stations, 1 core gets to about 80%, the rest are lower and GTX1080 is at 91% use inside Lave Station/George Lucas - I was flying between the two as a test between tweaking

any suggestions?

 

[TorTorden_VR]

I would simply set everything to 1x and go from there, again change it only one place at a time.

 

[loribee]

I would simply set everything to 1x and go from there, again change it only one place at a time.

Cool. I’ll try it tonight. It’ll be interesting to see what baseline SteamVR frames are with everything at 1x, although I know the quality will not be great as I checked out the candy rock star demo at 1.0 supersample in pimax and it was not all that sharp.

 

[matemaster]

Cool. I’ll try it tonight. It’ll be interesting to see what baseline SteamVR frames are with everything at 1x, although I know the quality will not be great as I checked out the candy rock star demo at 1.0 supersample in pimax and it was not all that sharp.

I have almost same rig as you
i7 2700K @ 4.6
GTX 1080


currently I use this settings
the only supersampling is done with my pimax 4k
check the resolution in steam VR here

With this settings I have around 50-55 FPS in stations, 55 FPS on planet surfaces and 60FPS everyvhere else

 

[loribee]

Thanks

I tried it and got about the same as you did with it, maybe slightly less in stations, but a constant 66FPS out there in space - which is the max FPS for my headset.