2.5 Settings Comparisons and GPU Impact

Yes, as stated above, the resolution is not 2160x1200 but 1080x1200 TWICE. In DCS settings, next to monitor choose "3 screen". You will have the same resolution as before but it is being split and rendered for 3 different views. You will notice an impact to your FPS. This is also similar to what happens with Mirrors and Res of cockpit displays.

No, there is a difference between rendering a scene ONE time on a big frame buffer which is later splitted, and rendering a scene TWO times on - no matter - one or two buffer(s)...

However, I know Nvidia was supposed to work some black magic with up to 16 viewports (viewpoints really) with the current 1000 series but I don't know how effective it is or what is required for it to work.

Viewport is not necerssarly "point of view"... there is no "black magic"... rendering from two points of view mean, calculating perspective, geometry, lighting, etc, two times for the same frame...

In fact with your settings and PD = 1 I obtain quite solid 45 fps and not always 90, maybe that's because 1080x1200 pixels x 2 is different from 2160x1200 pixels, even if the mere pixel count is the same.

Rift render target is 2700x1600 because image must strecht and cut to fit to optics. So even 1.0 PD its rendered much higher resolution than panel native.

Sent from my iPhone using Tapatalk

In fact with your settings and PD = 1 I obtain quite solid 45 fps and not always 90, maybe that's because 1080x1200 pixels x 2 is different from 2160x1200 pixels, even if the mere pixel count is the same.

This is simply two differents things... VR is about drawing the scene twice, from two different cameras, one for the right eye and one for the left one... Frame buffer size is a secondary problem here.

In any case this topic has been very enlightening to me... I used to play VR in default "VR settings", just increasing cockpit res to "high"; I've just played a-10c "sitting ducks" mission for a quick test, and pushing graphics high, quite similar to sideslip settings except antialiasing set only x2 (I do not see much visual difference in pushing it higher) allows a very big boost in visual quality (trees, sea, far distance and so on), without lowering fps: with PD 1 I'm 45 fps rock solid and the game is smooth. I don't know why, but a lot of micro-stutterings I experimented in 1.5.8 have gone. And the game has still to be optimized! I'll try with PD 1.2, but then according to the above table: PD 1.2 and MSAA 2 I should have a big boost in quality without lowering too much performance... I'll try.

Quality 10 lowers to much performance; aiming at quality 8 with best performance (8 ) in need to put PD to 1.2, but now (PD 1 and AA 2x) I have quality 6 and performance 10, so the best quality for maximum performance; I have to see if I want to trade a bit of performance (10 -> 8 ) for a bit of quality (6 -> 8 ).

PS strongharm can you link me the specific topic of performance and settings in VR?

I just read this (I did not know that MSAA works on this way):

MSAA basically calculates the edge pixels of your 3D models which is comprised of several polygons with the pixels adjacent to it. With 2x settings, it's calculated by doubling the resolution and then scaling it to fit your resolution. 4x quadruples it etc. So this process overall puts a lot of performance impact.

If this is true, then I understand why we have a problem in performance especially people who play at higher resolutions and VR users.

PS strongharm can you link me the specific topic of performance and settings in VR?

You can get there with the 'originally posted by' link:

VR PD = 1.2

MSAA = 2x

Anisotropic Filtering = 2x

Deferred Shading = ON

Terrain Shadows = FLAT

Cockpit Global Illumination = ON

 

 

[TABLE]

__PD__

|

_MSAA_

|

_Quality_

|Performance

1|

0

|

0

|

10

1.2

|

0

|

2

|

10

1.4

|

0

|

4

|

10

1.6

|

0

|

4

|

10

1.8

|

0

|

4

|

8

2.0

|

0

|

6

|

6

1

|

2

|

6

|

10

1.2

|

2

|

8

|

8

1.4

|

2

|

8

|

6

1.6

|

2

|

10

|

4

1.8

|

2

|

10

|

2

2.0

|

2

|

10

|

2

1

|

4

|

8

|

4

1.2

|

4

|

8

|

4

1.4

|

4

|

8

|

2

1.6

|

4

|

10

|

0

1.8

|

4

|

10

|

0

2.0

|

4

|

10

|

0

1

|

8

|

8

|

2

1.2+

|

8

|

10

|

0

[/TABLE]

 

No, there is a difference between rendering a scene ONE time on a big frame buffer which is later splitted, and rendering a scene TWO times on - no matter - one or two buffer(s)...

But it's not splitting the frame. It is 3 different views, that's why there is less distortion with 3 screens set to 170 degrees but 1 screen at 170 degrees is crazy distorted. The scene is rendered 3 times, once for each FOV. The only difference between that and VR is there is some overlap in viewed area, but it all has to be recomputed anyway.

You are confusing the 3 screen setting in DCS with actually using 3 monitors. You can use 1 monitor with the 3 screen setting or 3 monitors with the 1 screen setting.

Excluding the fact that VR will have X degrees FOV rendered 2 times and the 3 screen setting is A+B+C=X degrees FOV, they are the same thing in principle.

https://en.wikipedia.org/wiki/Viewport

In 3D computer graphics it refers to the 2D rectangle used to project the 3D scene to the position of a virtual camera. A viewport is a region of the screen used to display a portion of the total image to be shown.

3 screens = 3 viewports

VR = 2 viewports

It does not matter if they overlap or not. In VR they are facing similar directions (not the same mind you) from 2 separate points in space (each eyeball). With the 3 monitors setting they are each facing different directions but are all from the same point in space. Doesn't matter much except for total FOV.

I just read this (I did not know that MSAA works on this way):

It uses an algorithm to find edges, and then essentially increases the resolution (sample points) in that area to get a more accurate pixel color. Instead of a black line on a blue sky being all either blue or black pixels, they are different shades of blue and black as determined by the additional sample points.

This is a good video from Nvidia explaining their MFAA, which is a new way of doing MSAA at reduced performance cost:

If this is true, then I understand why we have a problem in performance especially people who play at higher resolutions and VR users.

But it's much easier to scream "ED don't know what they are doing! They broke MSAA!"

Keep in mind for VR; MFAA and other nVidia hardware level custom settings don't work with headsets that are currently available. A good compromise is MSAAx2 in the DCS settings, but this requires that PD be lowered to 1.2 to balance performance (and it looks great). MFAA would be a huge improvement for VR.

If PD has any influence on game when I'm not using VR?

biggest thing for me is the defferedshading on/off thing

dammed if its on, dammed if its off

hope this gets sorted in official 2.5 release

Keep in mind for VR; MFAA and other nVidia hardware level custom settings don't work with headsets that are currently available. A good compromise is MSAAx2 in the DCS settings, but this requires that PD be lowered to 1.2 to balance performance (and it looks great). MFAA would be a huge improvement for VR.

I'm not even sure MFAA works with DCS period, never mind VR.

Valium, the Pixel Density setting in DCS only effects VR. Here's a good explanation.

Sideslip, I've read conflicting reports about MFAA working with DCS. I personally only fly VR, so I haven't tested it. Sounds like a worthy addition to your outstanding analysis.

I would, but I think I'll leave that to the AA thread. There are so many people that swear it's working with override this or enhance that or use Nvidia Profile Inspector CVSAAACMAFS64 V1337... I don't think any conclusion that I come to would be accepted anyway.

Agreed Sideslip, I like the purist approach to settings you did in the OP analysis. Truly outstanding work. Thanks again.

Valium, the Pixel Density setting in DCS only effects VR. Here's a good explanation.

 

Sideslip, I've read conflicting reports about MFAA working with DCS. I personally only fly VR, so I haven't tested it. Sounds like a worthy addition to your outstanding analysis.

This article is more about the pixel density of a physical display as it relates to the human eye. The setting in game is just another name for supersampling, which is to say that the game renders at a larger resolution than what the screen can display and then scales it down, adding an effect similar to anti-aliasing.

I've been asked for my NCP settings and have added them to the My" Recommended Settings" tab.

This article is more about the pixel density of a physical display as it relates to the human eye. The setting in game is just another name for supersampling, which is to say that the game renders at a larger resolution than what the screen can display and then scales it down, adding an effect similar to anti-aliasing.

MSAA and MFAA don't have to scale entire frames as FSAA did. MSAA has an algorithm that upscales only the necessary pixels, which makes it less expensive than FSAA. MFAA has a much more efficient algorithm to perform a similar operation, which makes it even less expensive than MSAA. This is the 30k ft view summary. There's more info in the video Sideslip linked or the nVidia site for those who want the NOE details.

Any pics on what FXAA does to the GUI and to the cockpt textures blurring wise?

Any pics on what FXAA does to the GUI and to the cockpt textures blurring wise?

Just test it your self , im not tested it in VR yet but least in normal 1080p monitor it softens small fonts in GUI , but they are still ok.

In game i find that it takes jaggies away from pit (i dont use msaa , only fxaa) but there is some shimmering in far away. But nothin too bad.

Performance is good with fxaa.

VR test later , finally last Night sift for this week.

Sent from my iPhone using Tapatalk

There is a perceivable difference that I can see between terrain textures on High vs Low, it's most prominent on the Winter terrain

On Low there is this really ugly effect of terrain tiles close to the surface

On High the blending is much more pleasing to the eye with no noticeable tiling

I wasn't aware how resource hungry "default" terrain shadows are in 2.5 (they were not really in 1.5.8 ). Thanks Sideslip, if it wasn't for your investigation, i wouldn't tinker with them, as I always preferred to have them on "default".

Turned them back to "flat" and not only I gained 10-20 fps on Caucasus, but I got rid of almost all microstuttering there and on Nevada as well (Normandy remains unoptimized wreck as usual :D). Surely they will look bad during sunrises and sunsets, but I think that's the price I'm willing to pay (microstutters are a pet peeve of mine)

They shouldn't really, my understanding is that once you're far enough from the ground you'd get flat shadows anyway.

There is a perceivable difference that I can see between terrain textures on High vs Low, it's most prominent on the Winter terrain

 

On Low there is this really ugly effect of terrain tiles close to the surface

 

 

 

On High the blending is much more pleasing to the eye with no noticeable tiling

 

Indeed. I didn't experience that at all on normal terrain, but it seems that while there is only 1 set of textures for the base terrain, not so with snow. It's funny because I can't imagine there are many different snow textures that would really necessitate a reduction in quality.

Good find.

Any pics on what FXAA does to the GUI and to the cockpt textures blurring wise?

My screen shots of my settings and the comparison of Cockpit Global Illumination on/off both had FXAA on IIRC. That said, the cockpit photo was resized in Photoshop.

Blurring exists, but is extremely minor. YMMV.