I was doing some testing AA settings,

So I from 8xmsaa to 4xssaa,

and noticed  That I went from 7.2gb vram using 8xmsaa down to 6.5gb of vram using 4xssaa
don’t know if this has been mentioned before but this is great.

mike

Edited June 18, 20205 yr by mikeymike

not sure if it's because I'm running i 4k.  But 4 SSAA is a little too much on my 1080ti.  

4 minutes ago, micstatic said:

not sure if it's because I'm running i 4k.  But 4 SSAA is a little too much on my 1080ti.  

Yes, with 2SSAA is more than enough for 4k and you will save VRAM

I'll be honest.  I wish I could use it (and will when the next cards are released).  My 43" screen 1ft from my face definitely could use the AA horsepower.  

10 minutes ago, micstatic said:

I'll be honest.  I wish I could use it (and will when the next cards are released).  My 43" screen 1ft from my face definitely could use the AA horsepower.  

I’m using 4xssaa at 4k 1080ti and it’s smooth 

Have any of you heard about setting "compatibility bits" using Nvidia Inspector to get better AA with lower settings like 2x? See this article / thread for IL-2:

https://forum.il2sturmovik.com/topic/61819-finally-almost-no-shimmering-for-nvidia-users-without-taxing-4xsgssaa/

I'm not sure if Inspector settings override those of Prepar3D anyway. They do set "enhanced mode" in the Nvidia Control Panel??

Edited June 18, 20205 yr by whitav8

I haven't.  Inspector died off with FS2004 for me.  So would only really want to use it if I had to

Still confused as to why 8msaa is consuming more vram than 4xssa?

mike

2 hours ago, Rob_Ainscough said:

8 is greater than 4.  2X, 4X, 8X, 16X indicate the number of color samples being taken.  The more color samples you take the more VRAM you use.  MSAA and SSAA are just different techniques based on the color samples taken.  MSAA is a faster technique than SSAA.  There are many different techniques that can be used to determine how to change a pixels color (which is essentially what one is doing with any AA technique).

MSAA - Multisampling

MLAA - Morphological (AMD thing)

TAA - Temporal (frame compare, not very efficient)

DLSS - Deep Learning Super-Sampling (nVidia thing that sadly wouldn't work well with Flight Simulators)

MFAA - Multi-Frame (nVidia thing)

SMAA - Enchanced subpixel Morphological (combines methods MLAA+MSAA+SSAA - post process)

FXAA - Fast Approximate (post process)

CSAA - nVidia more efficient version of MSAA

CFAA - AMD's more efficient version of MSAA

SSAA - SuperSampling

Ultimately, we don't want to use any AA at all as they are technically all "filters".

The need for sample count and type of AA used is based on these key factors but not exclusive:

1. Eye point distance to monitor(s)
2. Size of monitor
3. Resolution of monitor

For example, I have a 40" 4K monitor at 24" head distance and 4XMSAA look excellent.  However on my other PC, 55" 4K monitor and head distance 28", 4XMSAA doesn't look as good so I have to use 4XSSAA.

Cheers, Rob.

Fantastic info, I really never understood exactly what they where, 

What would you recommend for a 27" 4k monitor? I think right now im currently using 2XMSAA

12 hours ago, Rob_Ainscough said:

1. Size of monitor
2. Resolution of monitor

= DPI - dots per inch or pixel size.

Another thing to consider regarding AA is that during the assembly of the frame, objects are AA'd when they are scaled into their location in the frame like slotting in a jigsaw piece. The edges are transparent to combine with the edge pixels with their neighbours. Simply having more DPI on a monitor does not remove the use of AA in the scene.