DCS 2.7 Graphics options recommendations

[Benchboy]

Hello! I've been benchmarking recently and here's my guide to maxing out image quality and performance.

My benchmarking right now is in the A-10 flying in the Caucasus and Vegas, singleplayer only. I'll include Strait of Hormuz and Syria later. I will update this thread with more information as I encounter new situations causing performance issues. Here's what you can expect in the end:

 

 

(After having played all included A-10C Caucasus scenarios, I consistently get 60 FPS +/-15 at worst.)

 

Step 1: Set all the graphics options to absolute minimum.

 

Step 2: Set anti-aliasing to match your desired initial performance, ideally about 50% above your desired final performance. For instance, if you currently have 144 FPS on minimum, set the level of anti-aliasing that matches the desired initial performance of 90 FPS if your desired final performance is 60 FPS. You can measure this in any appropriate scenario. It's all you gotta do on your own!

SSAA 1X MSAA 1X (-0% performance)
SSAA 1X MSAA 2X (-10% performance)
SSAA 1X MSAA 4X (-20% performance)
SSAA 1.5X MSAA 1X (-30% performance)
SSAA 1.5X MSAA 2X (-40% performance)
SSAA 1.5X MSAA 4X (-50% performance)
SSAA 2X MSAA 2X (-60% performance)
SSAA 2X MSAA 4X (-70% performance)

Finally, in the NVIDIA Control Panel set "MFAA" to "On" for significantly improved anti-aliasing (-2% performance)!

Note: the performance values above are based on how they affect your final, not initial, performance.

Note: I intentionally skipped SSAA 2X MSAA 1X. I recommend never using it as performance matches the superior image quality SSAA 1.5X MSAA 4X.

Alternate method: if you set "Antialiasing - Mode" to "Enhance the application setting" and "Antialiasing - Transparency" to 2X, you substantially improve anti-aliasing at substantial performance change, which is interesting as it allows you to use SSAA 1X with image quality comparable to SSAA 1.5X and +16% performance; however, the image may not appear as sharp. Also, screenshots will appear aliased for technical reasons.

 

Step 3: Set all graphics options that don't affect performance.

Civ. Traffic: Personal preference
Heat Blur: High
Resolution: Personal preference
Aspect Ratio: Personal preference
Monitors: Personal preference
Res. of Cockpit Displays: 1024 Every Frame
Lens Effects: Personal preference
Preload Radius: 100
Chimney Smoke Density: Personal preference
Gamma: 2.2
Messages font scale: Personal preference
Scale GUI: Personal preference
Rain Droplets: Personal preference
Vsync: OFF (Use NVIDIA Control Panel)
Full Screen: Personal preference
Cursor Confined to Game Window: Personal preference

Step 4: Set the graphics options that are extremely situational.

Depth of Field: Personal preference (Affects performance greatly only in third-person view, -50%?)
Clouds: Ultra (Still searching for situation where it makes difference)
Clutter/Grass: 1500 (Affects performance weakly in cockpit, -2.5%/-5% at 750/1500)
Forest Visibility: 100% (Affects performance greatly in woods, I recommend 67-80%)
Forest Details Factor: 1 (Still searching for situation where it makes difference)
Scenery Details Factor: 1 (Still searching for situation where it makes difference)

Step 5: Set the graphics options that measurably affect performance. Upgrade the graphics options in any order until you're at your desired final level of performance.

Textures: High (-1% performance)
Terrain Textures: High (-1% performance)
SSLR: ON (-3% performance)
Cockpit Global Illumination: ON (-3% performance)
Shadows: High (-3% performance)
Water: Medium (-5% performance) or High (-10% performance)
Motion Blur: ON (-10% performance)
Anisotropic Filtering: 2X (-2% performance), 4X (-4% performance), 8X (-8% performance), or 16X (-12% performance)
Terrain Objects Shadows: Flat (-5% performance) or Default (-15% performance)
Visib. Range: Extreme (-17% performance)
SSAO: ON (-25% performance)
There's also one option in the gameplay menu:

Mirrors: ON (-10% performance)

My recommendation (click to view graphics options, Imgur): I have a 3070 graphics card and these graphics options give me an image quality matching the maximum graphics options with generally 60+ FPS. If you have set all of the above and want higher performance I recommend setting these in order until you achieve your desired final performance: Anisotropic Filtering 8X, 4X, or even 2X, Motion Blur OFF, Mirrors OFF, SSAO OFF, Terrrain Object Shadows Flat, Water Medium, Water Low, SSLR OFF, Cockpit Global Illumination OFF, and Shadows Low.

 

Comparison screenshots: All Minimum, My Recommendation, All Maximum. My recommendation has about 4X the performance of all maximum and you can hardly tell, right?

 

A few of my key insights here are:

• Performance should always be objectively measured in %!
• There are graphics options that don't affect performance and for that reason should always be on.
• Ideally for performance you should have 95%+ GPU usage, 95%- CPU usage, and 95%- VRAM usage. I recommend checking this on occasion using Afterburner.
• The anti-aliasing combination SSAA 1.5X MSAA 2X MFAA On is extremely effective at reducing aliasing artifacts. MSAA reduces aliasing of 3D objects (e.g. houses) and SSAA reduces aliasing of sprites (e.g. trees). These are examples of spatial aliasing, while MFAA reduces temporal aliasing only observable in moving images.
• It's generally a good idea to set Visib. Range Extreme and Forest Visibility to 80% (matches Visib. Range Ultra), 67% (matches Visib. Range High), 45% (matches Visib. Range Medium), or 35% (matches Visib. Range Low) instead of lowering Visib. Range, at least if the forest is causing the performance change such as in the Caucasus terrain. This significantly increases the geometric quality of objects.

Edited July 21, 2021 by Benchboy

[jack333]

Thank´s

 

[71st_AH Rob]

Benchboy

 

There is another option in the Gameplay menu in addition to mirrors; Wake Turbulence which can have a significant impact, particularly with propeller driven aircraft.

 

Edited July 19, 2021 by 71st_AH Rob

[Hiob]

Good effort, thanks.

The only thing I find strange is the big impact of anisotropic filtering.

In my own tests it had only negligible effects on fps (if any).

I will review.

 

I missed the resoultion you did those tests in?

Edited July 19, 2021 by Hiob

[Fastbreak]

Well structured, easy-to-follow, well done! 

 

Thank you, too.

[71st_AH Rob]

1 hour ago, Hiob said:

Good effort, thanks.

The only thing I find strange is the big impact of anisotropic filtering.

In my own tests it had only negligible effects on fps (if any).

I will review.

 

I missed the resoultion you did those tests in?

I found the same with anisotropic filtering, no impact.  I also have found, surprisingly, that resolution is min impact.  I run in 4K with a GTX 1070 at 50fps normally.  I have found that I can leave MSAA off at 4K and it looks as good as at lower resolutions and gain the performance.

 

Also, VSync is a major resource hog, if you have a GSync monitor and NVidia card, I recommend you use that.  If you can live without VSync you can increase the eye candy.

 

[Hiob]

2 hours ago, Hiob said:

The only thing I find strange is the big impact of anisotropic filtering.

In my own tests it had only negligible effects on fps (if any).

I will review.

 

 

See, good I researched it. My memory played tricks on me. In my own tests, I didn't cover AF.

 

Those are my results (different Scene than above):

19-07-2021, 20:08:27 DCS.exe benchmark completed, 9745 frames rendered in 147.937 s (16x AF)
                     Average framerate  :   65.8 FPS 94,4%
                     Minimum framerate  :   55.3 FPS 90,8%
                     Maximum framerate  :   74.0 FPS 95,1%
                     1% low framerate   :   55.9 FPS 91,0%
                     0.1% low framerate :   55.1 FPS 90,6%
19-07-2021, 20:12:21 DCS.exe benchmark completed, 9888 frames rendered in 148.000 s (8x AF)
                     Average framerate  :   66.8 FPS 95,8%
                     Minimum framerate  :   57.0 FPS 93,6%
                     Maximum framerate  :   74.9 FPS 96,3%
                     1% low framerate   :   57.4 FPS 93,5%
                     0.1% low framerate :   56.8 FPS 93,4%
19-07-2021, 20:17:38 DCS.exe benchmark completed, 10328 frames rendered in 148.078 s (0x AF)
                     Average framerate  :   69.7 FPS 100%
                     Minimum framerate  :   60.9 FPS 100%
                     Maximum framerate  :   77.8 FPS 100%
                     1% low framerate   :   61.4 FPS 100%
                     0.1% low framerate :   60.8 FPS 100%

[Benchboy]

On 7/19/2021 at 5:28 PM, 71st_AH Rob said:

Benchboy

 

There is another option in the Gameplay menu in addition to mirrors; Wake Turbulence which can have a significant impact, particularly with propeller driven aircraft.

 

 

 

I'm not disabling anything that affects the gameplay!

 

On 7/19/2021 at 5:37 PM, Hiob said:

Good effort, thanks.

The only thing I find strange is the big impact of anisotropic filtering.

In my own tests it had only negligible effects on fps (if any).

I will review.

 

I missed the resoultion you did those tests in?

 

 

I did it in 2560x1440, maybe that's why we see a difference. Also differs by situation.

[Vintageflyer01]

Could it be, that this is the difference between benchmarking at a real mission and benchmarking by using a video recording of an mission?   

 

[Hiob]

4 hours ago, Vintageflyer01 said:

Could it be, that this is the difference between benchmarking at a real mission and benchmarking by using a video recording of an mission?   

 

There is no difference.

The graphics AND physics engine have to render a replay scene the same way they need to render gameplay.

The replay doesn't "record" the frames but the (input-) parameters of a flight and then simulates it again.

That is the reason that sometimes the replays divert from the actual flight at some point....

 

A replay therefore works perfectly as a benchmark.