Hi everybody.

 

In UserCfg.opt, there is line SecondaryScalingVR 1.000000. This is actually standard upscaling method for VR. It can be used with DLSS to have great results in performance. It basicly upscaling factor. Here is little bit info about DLSS:

 

DLSS quality modes with downscale prosent:

- Quality - 66.6% (2/3) per axis, 45% resolution.

- Balanced - 58% per axis, 33% resolution

- Performance - 50% per axis, 25% resolution.

- Ultra Performance - 33% (1/3) per axis, 11% resolution.

 

DLSS is designed to work more efficient with lower resolutions. 

 

Here is DLSS presets

Preset A: Intended for Performance/Balanced/Quality modes. An older variant best suited to combat ghosting for elements with missing inputs, such as motion vectors.

Preset B: Intended for Ultra Performance mode. Similar to Preset A but for Ultra Performance mode.

Preset C (most preferred): Intended for Performance/Balanced/Quality modes. Generally favors current frame information; well suited for fast-paced game content.

Preset D (2nd preferred): Default preset for Performance/Balanced/Quality modes; generally favors image stability.

Preset E: Slightly sharper image, improved fine detail stability, reduced ghosting and better temporal stability

Preset F: Default preset for Ultra Performance and DLAA modes.

Latest nvidia driver and DLSS 3.8.10 only supports presets E and F!

 

Here is example:

- you put SecondaryScalingVR 1.200000 and it will upscale your resolution by 20%

- you select DLSS Balanced in MSFS/MSFS24. It will downscale your resolution by 58% but uses DLSS magic to make image sharper and better.

- Simple mathematic example.

     * I have Pico 4 used with Virtual desktop God like mode (3120x3120).

     * I put SecondaryScalingVR 1.50000, so Post rendering resolution is 4680x4680.

     *  DLSS balanced downscales it to final rendering resolution 3042x3042.

     * Performance is far better than you would use TAA, but it looks pretty good. There is offcourse artifacts from DLSS as general there is allways.

     * if you do benchmark with only DLSS Ultra quality mode, to match final render resolution as with SecondaryScalingVR+DLSS balanced, you wont see the diffrence and performance is way more better with SecondaryScalingVR+DLSS balanced.

 

Changing DLSS modes

- You can change DLSS modes downscale prosent with DLSS tweaks or Nvidia profile inspector. Guides are all over the place (youtube, avsim)

- I use DLSS balanced with 65% downscale and SecondaryScalingVR 1.120000 and im having constant 72fps with 72hz mode, WITHOUT MOTION SMOOTHING/ASW/SSW/MOTION REPROJECTION!

 

OXR Toolkit

OXR toolkit´s override resolution does same thing that SecondaryScalingVR does.

 

Here is comparison about those two methods (same final render resolution):

This is DLSS Balanced 0.6 (VD Godlike 3120x3120). So final rendering 2028x2028

 

This is DLSS Balanced 0.6 + SecondaryScalingVR 1.50000  (VD Godlike 3120x3120). So final rendering 3042x3042. Post rendering 4680x4680 from wich DLSS downscales to 3042x3042.

 

This is DLSS Balanced 0.6 + OXR Toolkit resolution override  (VD Godlike 3120x3120). So final rendering 3042x3042. Resolution override to 4680x4680 from wich DLSS downscales to 3042x3042.

Edited December 2, 20241 yr by Corsten
added comparison images.

Interesting thanks!

I wonder if I can try the same with FSR2 on my AMD's.

I wanted to edit first post, but not able to do it anymore. Here is addon for it:

DLSS version

both version of MSFS still uses old version of DLSS (not frame generation, it is the newest).
There is huge lift in performance and visual with latest 3.8.10
you can download it here: https://www.techpowerup.com/download/nvidia-dlss-dll/
make backup of the original nvngx_dlss.dll and paste this new one to the core.
you can also use DLSS swapper ( https://github.com/beeradmoore/dlss-swapper )
 

Edited December 2, 20241 yr by Corsten