How FXAA or MSAA affects to CPU&GPU loads

[chiliwili69 880]

I have been playing with the v5 benchmark mission (temporary) kindly created by SYN_Vander to know that.

I have put all in-game settings to low using a 1080p monitor, and I have been touching the AA settings in a number of runs of the mission.

These are the results:

noAA
Frames: 5934 - Time: 49000ms - Avg: 121.102 - Min: 101 - Max: 161
Frames: 5957 - Time: 49000ms - Avg: 121.571 - Min: 101 - Max: 163

 

FXAAx4
Frames: 6050 - Time: 49000ms - Avg: 123.469 - Min: 104 - Max: 160
Frames: 5969 - Time: 49000ms - Avg: 121.816 - Min: 101 - Max: 163

 

MSAAx4
Frames: 5910 - Time: 49000ms - Avg: 120.612 - Min: 100 - Max: 162
Frames: 5853 - Time: 49000ms - Avg: 119.449 - Min: 100 - Max: 159

 

MSAAx8
Frames: 5943 - Time: 49000ms - Avg: 121.286 - Min: 98 - Max: 159
Frames: 5863 - Time: 49000ms - Avg: 119.653 - Min: 99 - Max: 159

 

During all these runs the GPU load was well below 100% (each setting had different load), so we could say that it is not GPU constrained. So the fps limit was determine purely by the CPU/RAM performance. As we can see, there is not measurable diferences in all these runs. So we can conclude that AA alone doesn´t load more the CPU.

 

The impact it is seen in the GPU. I have logged MSI Afterburner %GPU load during these runs and I have composed an image that sumarize the load:

 

 

So the GPU cost of each option for a not constrained 1080Ti in a 1080p monitor is about:

+3% for 4xFXAA

+12-18% for 4xMSAA

+22-26% for 8xMSAA

 

This is the beauty of having a benchmark!

 

[chiliwili69 880]

Since many people run IL-2 in higher resolutions (4K monitor or VR) I also wanted to know what is the cost at those resolutions.

To test that I maxed out all in-game settings except clouds to Low using a 4K monitor. And also down-clock my CPU from 4.8GHz to 4.0Hz. I did this to run the benchmark at a low fps rate (to try having the CPU as the only limiting factor during the test, so GPU is always below 100% and gains can be measured).

 

The runs in 4K are:

noAA
Frames: 3345 - Time: 49000ms - Avg: 68.265 - Min: 58 - Max: 91

FXAAx4
Frames: 3194 - Time: 49000ms - Avg: 65.184 - Min: 55 - Max: 88

MSAAx4
Frames: 3193 - Time: 49000ms - Avg: 65.163 - Min: 56 - Max: 88

MSAAx8
Frames: 2967 - Time: 49000ms - Avg: 60.551 - Min: 50 - Max: 77

 

The last run is not only CPU limited since GPU reached 100%. But here the CPU cost of using AA is really marginal. (less than 3 fps).

 

The GPU% load for 4K are:

 

So the GPU impact in 4K with a 1080Ti are:

FXAA: negligible

MSAAx4: +25-30%

MSAAx8: +40-45% 

[41Sqn_Skipper 200]

Great information, thanks for the effort!

[Gambit21 5619]

Nice work chili.

[dburne 2590]

Did a lot of PWCG Campaign flying today as I was adjusting graphics for my new Reverb G2.

I ended up dialing out the MSAA ( Previously had that at x4), flew that for a bit, then added x4 FXAA.

If there was a hit on my performance from that I could not notice it.

 

[Bernard_IV 105]

FXAA blurs out distant contacts for me making them harder to see.  This is the last time I messed with it 4 months ago or so.