My 3090 Settings for my G2 (2.7.11)

[ShaunX]

25 minutes ago, Ready_ENP said:

Would you mind explaining the test you are running and how to get to the end results? I would like to jump on board and run exactly the same tests, so I get a clear understanding of where I am at with my rig in relation to yours. I ran my own tests before based on something else and then it is hard to compare. I have pretty decent hardware specifically bought for DCS and ran through a lot of guides, but have the feeling something is seriously off when I see your stats.

Thanks a bunch,

You need to install FPSVR from Steam. FPSVR has a hoy key - right CTRL + F11 which will start recording stats, pressing it again stops recording and writes the output to a .csv file in documents\FPSVR\.

So, leave you G2 hanging on its stand (or wherever you can place it so that it won't move), switch on the G2, launch FPSVR (this will launch SteamVR) launch DCS. You should see the menu centered on your mirror display. Launch the track file performance track 276.trk as it launches press the hot key. When it finishes press it again.

Open the CSV file, there'll be thousands of rows. global find/replace All " characters (or I find calculations won't work). Remove the first few and last few rows (It should be obvious how many to remove as the CPU values will be either far to high or far too low.) Then take the average of the FPS column and copy that formuia to the GPU/CPU framerate and GPU usage columns. And there's your results. We usually perform the test twice and take the average of the results between the two.

A little long winded but easy once you've done a few

 

..oh and don't foget to disable motion smoothing/vector first..otherwise FPS will just be 45 throughout (I forget that regularly!)

Edited October 26, 2021 by ShaunX

[Ready]

1 minute ago, ShaunX said:

You need to install FPSVR from Steam. FPSVR has a hoy key - right CTRL + F11 which will start recording stats, pressing it again stops recording and writes the output to a .csv file in documents\FPSVR\.

So, leave you G2 hanging on its stand (or wherever you can place it so that it won't move), switch on the G2, launch FPSVR (this will launch SteamVR) launch DCS. You should see the menu centered on your mirror display. Launch the track file performance track 276.trk as it launches press the hot key. When it finishes press it again.

Open the CSV file, there'll be thousands of rows. global find/replace All " characters (or I find calculations won't work). Remove the first few and last few rows (It should be obvious how many to remove as the CPU values will be either far to high or far too low.) Then take the average of the FPS column and copy that formuia to the GPU/CPU framerate and GPU usage columns. And there's your results. We usually perform the test twice and take the average of the results between the two.

A little long winded but easy once you've done a few

Thank you very much. I am using FPSVR, but always make quick screenshots to get a rough idea of where I am at. I just play a stock mission and record a clip with OBS, then take a screenshots from within OBS at exactly the same moment of each clip. Never knew about the csv. Will follow your instructions and see what it gets me. Much appreciated!

Here is the last time I made a comparison with the mods. 

 

[ShaunX]

3 minutes ago, Ready_ENP said:

but always make quick screenshots to get a rough idea of where I am at

..exactly what I did until I found this thread..and very glad I did

[peachmonkey]

some new reports from the 2D world about Low/No Shadows causes the performance drops, i.e. it's better to set them to LOW... 

Haven't tested it myself yet, just FYI:

 

 

[speed-of-heat]

nope, I cant replicate the data 

I get approximately 75FPS with Default shadows and  80 something without shadows

 

[dureiken]

16 minutes ago, speed-of-heat said:

nope, I cant replicate the data 

I get approximately 75FPS with Default shadows and  80 something without shadows

 

75 for you and 72 for me, seems better

[Ready]

On 10/26/2021 at 8:48 PM, ShaunX said:

..exactly what I did until I found this thread..and very glad I did

Okay got it working and running the same tests. Compiling them into a table. 

Thanks!

Edited October 27, 2021 by Ready_ENP

[speed-of-heat]

good for you

[Ready]

9 hours ago, speed-of-heat said:

good for you

Thank you and thank you for creating the baseline track with all the data. It really gives a strong and clear benchmark reference.

I ran 3 quick tests and see there is some potential performance to be gained if I take a 5-10 % performance difference between the 3080 and 3090 into account.

FPS	GPU frate	CPU frate	GPU Usage	-
61,23	12,60	9,15	79,89	DCS 2.7.7 Simplex 2.7.0
50,01	13,38	9,08	70,84	DCS 2.7.7 No shaders
83,73	8,61	7,75	78,34	DCS 2.7.6 IC Pass Shaders, Reshade, FSR

Just a quick recording of my fourth trial. (based on nr3 above with added MSAA2X, visibility Ultimate, Anisotropic filtering ON and forrest/scenery details max.

Great track file SOH!

 

Edited October 27, 2021 by Ready_ENP

[dureiken]

1 minute ago, Ready_ENP said:

Thank you and thank you for creating the baseline track with all the data. It really gives a strong and clear benchmark reference.

Could you please point me to the Simplex 2.1 and Clearwater 2.7.7 mods? I can't find them in the thread nor the ED Forum downloads.

I ran 3 quick tests and see there is some potential performance to be gained if I take a 5-10 % performance difference between the 3080 and 3090 into account.

FPS	GPU frate	CPU frate	GPU Usage	-
61,23	12,60	9,15	79,89	DCS 2.7.7 Simplex 2.7.0
50,01	13,38	9,08	70,84	DCS 2.7.7 No shaders
83,73	8,61	7,75	78,34	DCS 2.7.6 IC Pass Shaders, Reshade, FSR

 

Could you post your DCS settings please ?

 

You should try FSR too

 

thanks

Edited October 26, 2021 by dureiken

[Ready]

52 minutes ago, dureiken said:

Could you post your DCS settings please ?

 

You should try FSR too

 

thanks

 

Will post a link to the table I am making with all my settings. Am running FSR.

 

[speed-of-heat]

Okay after some re-testing on the shadows I can confirm that when you ask a GPU to do more work it does make it go slower:

			Frametime in ms		Delta From Baseline
Ver	Test	Baseline	GPU	CPU	GPU	CPU	Changes
2.7.6.13133	82	Test 75	9.52	8.55			Average for Baseline 2.7.6, FSR1.1, Windows 11, Hotfix Shaders
2.7.7.14727	97	Test 82	10.06	8.08	-0.54	0.47	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water No terrain shadows New baseline
2.7.7.14727	98	Test 97	10.44	8.33	-0.38	-0.25	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water, Flat Terrain Shadows
2.7.7.14727	99	Test 97	10.62	8.71	-0.56	-0.63	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water, Default Terrain Shadows

Edited October 27, 2021 by speed-of-heat

[dureiken]

7 minutes ago, speed-of-heat said:

Okay after some re-testing on the shadows I can confirm that when you ask a GPU to do more work it does make it go slower:

			Frametime in ms		Delta From Baseline
Ver	Test	Baseline	GPU	CPU	GPU	CPU	Changes
2.7.6.13133	82	Test 75	9.52	8.55			Average for Baseline 2.7.6, FSR1.1, Windows 11, Hotfix Shaders
2.7.7.14727	97	Test 82	10.06	8.08	-0.54	0.47	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water No terrain shadows New baseline
2.7.7.14727	98	Test 97	10.44	8.33	-0.38	-0.25	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water, Flat Terrain Shadows
2.7.7.14727	99	Test 97	10.62	8.71	-0.56	-0.63	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water, Default Terrain Shadows

 

is that possible to add mean/min fps ?

 

thanks

[speed-of-heat]

i also note that this is pretty consitent with what i tested back with test 13

13

Test 4

15.14

8.13

0.63

0.46

Terrain Object Shadows to Flat

1 minute ago, dureiken said:

is that possible to add mean/min fps ?

 

thanks

No because its actually not a helpful measure... FPS is an aggregate of CPU and GPU frametime, it can be fairly misleading. The important thing that as GPU and CPU frametime increases observed FPS decreases.

All of my tests and data points are based around trying very hard not to talk about FPS because of this and solely looking at measured performance deltas in hardware

[dureiken]

14 minutes ago, speed-of-heat said:

i also note that this is pretty consitent with what i tested back with test 13

13

Test 4

15.14

8.13

0.63

0.46

Terrain Object Shadows to Flat

No because its actually not a helpful measure... FPS is an aggregate of CPU and GPU frametime, it can be fairly misleading. The important thing that as GPU and CPU frametime increases observed FPS decreases.

All of my tests and data points are based around trying very hard not to talk about FPS because of this and solely looking at measured performance deltas in hardware

ok thanks, it was just to have an idea of what it could be in term of fps

I have a 5.1Ghz 8600k, do you think a 5600X or better could improve that ? or GPU is almost everytime the bottleneck ?

 

thanks

[ShaunX]

34 minutes ago, speed-of-heat said:

Okay after some re-testing on the shadows I can confirm that when you ask a GPU to do more work it does make it go slower:

			Frametime in ms		Delta From Baseline
Ver	Test	Baseline	GPU	CPU	GPU	CPU	Changes
2.7.6.13133	82	Test 75	9.52	8.55			Average for Baseline 2.7.6, FSR1.1, Windows 11, Hotfix Shaders
2.7.7.14727	97	Test 82	10.06	8.08	-0.54	0.47	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water No terrain shadows New baseline
2.7.7.14727	98	Test 97	10.44	8.33	-0.38	-0.25	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water, Flat Terrain Shadows
2.7.7.14727	99	Test 97	10.62	8.71	-0.56	-0.63	2.7.7, FSR 1.2 Clear Water Shaders, Medium Water, Default Terrain Shadows

 

LOL, yes, like we knew it would !

Have to tested FSR 1.1 v 1.2. Your GPU framerate seem high considering..

..which was also presumably on high water I got a much bigger GPU reduction when I added FSR 1.1.. though maybe we're using different settings  ? // AMD presets:     //   Ultra Quality => 0.77     //   Quality       => 0.67     //   Balanced      => 0.59     //   Performance   => 0.50     "renderScale": 0.80,     // tune sharpness, values range from 0 to 1     "sharpness": 0.0, "radius": 0.75, ..and  I have clouds on Standard

[speed-of-heat]

ahh  my clouds are on ultra

yep 82 was also on high water 

[ShaunX]

9 minutes ago, speed-of-heat said:

ahh  my clouds are on ultra

yep 82 was also on high water 

Makes sense, just quickly tested with clouds ultra, CW, medium water

..and there's the additional GPU second

Still don't quite understand why my CPU is so low. Must be the 5900X I guess, which interestingly only ever boosts to 4.8

 

[speed-of-heat]

makes sense! well there is more to frametime than just Hz... but yes that might be a factor

[Lurker]

Yes the 5900X is a beast, and that whole series of CPUs are consistently better than (current) Intel CPUs in most single or multithreaded apps. 

[dureiken]

Just to be sure :

 

Quote

GPU Frametimes:
Median: 11 ms
99th percentile: 14.2 ms
99.9th percentile: 14.9 ms
frametime <11.1ms(vsync): 60.6%
CPU frametime:
Median: 8.6 ms
99th percentile: 10.8 ms
99.9th percentile: 11.7 ms
frametime <11.1ms(vsync): 99.6%

When I read this, I can admit that my GPU is the bottleneck ? no use to change the CPU ?

thanks

[speed-of-heat]

at the moment correct... you might get some additional performance on things like AI ...  so it may help my guess is not much 

[Baltic Pirate]

Your guide has proved to make my Reverb G2 reach new heights. Improvement is visible. I cannot thank you enough. Really great

[zildac]

Just curious, where people are running FSR (1.2 in my case) are you running steam SS at 100% when running these benchmarks?

[dureiken]

@speed-of-heat

I don't understand resolution in your guide.

 Custom Resolution Modifier for DCS to 80% 2836x2768

With G2, native resolution isn't 2160*2 = 4320 and 2160 high ?

I have this for G1 :

Render resolution per eye: 2200x2160(by SteamVR settings, Max.) (HMD driver recommended: 2201x2158)

Thanks