[Suggestion] Add a "UI scale target resolution" feature with the advent of DX11
With the advent of DX11, we FINALLY get some really HQ SSAO in the form of HBAO+ Which is VASTLY superior to the really low quality SSAO from the DX9 client in a number of ways.
_________________________________________________
---------------------------------------------------------------------
-Physically accurate occlusion taking into account light
-Far less prone to aliasing
-Far more samples per pixel
-Full resolution buffer
-Much less prone to artifacts in motion and clouding which make IMO DX9 SSAO completely unusable.
__________________________________________________
----------------------------------------------------------------
This is a HUGE plus! Making the game world feel even more alive.
The only problem now, is that we are left with only FXAA for Anti-Aliasing.
This is very bad for a number of objective reasons. And this leaves the DX9 client for Nvidia users with a major advantage.
==Higher quality Anti Aliasing.
Rather than discussing why it's vastly superior in every way. I'll just show you a video comparison of FXAA and 4xSGSSAA in motion.
https://copy.com/wYWPwuimvWI68ZRN
The video encode isn't very great, but it works well enough to showcase most of the biggest differences which should be easily visible to anyone who knows what aliasing looks like.
______________________________________________________
-------------------------------------------------------------
Now, we could get decent AA in the DX11 client via OGSSAA (downsampling) alongside using FXAA.
The only problem is that the UI doesn't scale well at very high resolutions. Even at it's largest settings it is either too small,unreadable or very ugly looking and blurry due to the scaling from rendering *Everything* at a much higher resolution.
The solution to this is to simply code in a set of options to fix this problem.
==The first being Ordered Grid Super Sampling Anti-Aliasing.==
This is a piss easy solution to implement into games these days that is seeing more and more use. It is essentially the same as downsampling with a utility like GeDoSaTo (Google it) or driver downsampling or DSR/VSR.
You simply take every buffer, and render it to a fixed scale proportional to the display/final render resolution.
Ex- 1920x1080 with 1.5x1.5 OGSSAA multiplies each axis by 1.5, resulting in a resolution of 2880x1620. Which in turn is "2.25xSSAA" Because there are 225% as many pixels as in 1920x1080.(The base value of 100%)
The reconstruction or resampling filter when you downfilter these buffers to display resolution is often a Box or Bilinear function. Nvidia's DSR uses an adjustable Gaussian function (Which is not the same as Gaussian blur. Gaussian Blur is a blurring technique using a Gaussian function).
Choosing one of these would be simple and they could even offer multiple choices for users.
The options for SSAA would fall in line to something like
==SSAA Scale==
[*]2.25xSSAA
[*]2.25xSSAA + FXAA
[*]4xSSAA
[*]4xSSAA + FXAA
==SSAA Filter==
[*]Box
[*]Bilinear
[*]Bicubic
[*]Lanczos
[*]Gaussian
==Display Resolution==
[*] List of detected resolutions - This is what the SSAA is scaled back to
==UI Display/Render Resolution==
[*] Fixed to Display Resolution
[*] Not fixed to Display Resolution
This would render the UI pass AFTER the downfiltering stage of all the other rendering buffers before final output to the Primary Flip Chain.
And thus fix the scale to your display resolution and not the rendering resolution when SSAA is used. Preventing unsightly tiny or hard to read text and UI elements.
__________________________________
_--------------------------------------------------
What do you think? Any suggestion on whom to send this to for consideration?
