DCS 2.7 Detailed 4K Benchmark with Settings

[Hiob]

Update:

I made a new Benchmark Scene for the Marianas Map: (Which is now broken, since the trackfile doesn't work anymore after updates of the Map.)

 

 

Original Topic:

 

Hi Guys,

 

I wanted to make detailed and controlled Benchmarks to figure out the performance of the Syria map and to find my new comfort zone in the settings.

I made everything so that it is replicable, if someone wants to compare his own setup. The Benchmark was made with a Trackfile and defined start and end of benchmarking with Afterburner/RivaStatisticsServer.

The Scenery for the Benchmark was chosen with a focus on performance of the map. Therefore it's a flight in low altitude (4000 feet), BELOW the broken cloud layer over Haifa. Here is an example of the scenery:

(see end of post)

Attached is the trackfile, I used, aswell as the txt with the detailed results and screenshots of the baseline settings.

Note that I made this to test my personal Rig with a 4K-gsync-display and a 3080. There is probably no use in comparing it to a QHD or WQHD-Dispay-setup.

The Baseline for the tests are custom "medium" settings, which you can see here:

From there I walked my way up for each setting, that I feel useful and noted the change.

Those are my findings:

 

========================================================================================================

DCS.exe benchmark completed, 12138 frames rendered in 139.875 s (Custom "medium" settings, 5900x, 3080 stock clock)
Average framerate  :   86.7 FPS

DCS.exe benchmark completed, 12353 frames rendered in 139.968 s (Custom "medium" settings, 5900x, 3080 +50 Hz/+200 Hz)
Average framerate  :   88.2 FPS +1.5 FPS (+1.7% over stock)

DCS.exe benchmark completed, 12514 frames rendered in 139.500 s (Custom "medium" settings, 5900x, 3080 +100 Hz/+500 Hz)
Average framerate  :   89.7 FPS +3 FPS (+3.4% over stock)

DCS.exe benchmark completed, 12103 frames rendered in 139.812 s (Custom "medium" settings, 5900x, 3080 stock clock)
Average framerate  :   86.5 FPS (same settings as first test, 0.2 FPS margin of error)

DCS.exe benchmark completed, 12050 frames rendered in 139.609 s (Terrain Textures Low -> High)
Average framerate  :   86.3 FPS -0.2 FPS (within margin of error)

DCS.exe benchmark completed, 11478 frames rendered in 139.656 s (Water Medium -> High)
Average framerate  :   82.1 FPS -4.2 FPS (-4.9%)

DCS.exe benchmark completed, 11065 frames rendered in 139.640 s (Visible Range High -> Ultra)
Average framerate  :   79.2 FPS -2.9 FPS (-3.5%)

DCS.exe benchmark completed, 10919 frames rendered in 139.781 s (Shadows Medium -> High)
Average framerate  :   78.1 FPS -1.1 FPS (-1.4%)

DCS.exe benchmark completed, 10944 frames rendered in 139.891 s (Resolution of Cockpit Displays 512 -> 1024) (Important Note: I wasn't fully aware, what this setting did at the time. Since there no cockpit displays running in this benchmark scene (That is TV/FLIR/TADS/TPOD and Mirrors(!) it didn't impact the performance. In fact the performance hit is quite severe with actual displays running!)
Average framerate  :   78.2 FPS within margin of error

DCS.exe benchmark completed, 10929 frames rendered in 139.796 s (Lens Effects None -> Flare)
Average framerate  :   78.1 FPS within margin of error

DCS.exe benchmark completed, 10633 frames rendered in 139.875 s (Clouds Standard -> Ultra)
Average framerate  :   76.0 FPS -2.1 FPS (-2.7%)

DCS.exe benchmark completed, 9771 frames rendered in 139.985 s (SSAO Off -> On)
Average framerate  :   69.8 FPS -6.2 FPS (-8.2%)

DCS.exe benchmark completed, 9718 frames rendered in 139.640 s (Grass 1500 -> 0, Buildings 1 -> .7, Forrest 1 -> .7)
Average framerate  :   69.5 FPS within margin of error

DCS.exe benchmark completed, 12519 frames rendered in 139.937 s (Maximum Performance (with bearable looks))
Average framerate  :   89.4 FPS

DCS.exe benchmark completed, 7670 frames rendered in 139.750 s (Maximum Quality (within reason))
Average framerate  :   54.8 FPS

DCS.exe benchmark completed, 11453 frames rendered in 139.860 s (Final Settings for balanced looks/performance)
Average framerate  :   81.8 FPS

=========================================================================================================

 

Max Performance Settings (within reason):

Max Quality Settings (within reason):

My final "balanced" Settings:

Example of the Benchmark Scenery:

(start the Benchmark when the cockpit view slews to the right, end the Benchmark when the F-16 starts to bank to the left)

Benchmark.txt DCS Benchmark 4K.trk

Edited June 22, 2022 by Hiob

[flankerjun]

you can set MSAA off，because in 4K monitor,the MSAA will be no effect, i have an RTX2070 with 4K monitor, i also can get 60~90 FPS

[Hiob]

2 hours ago, flankerjun said:

you can set MSAA off，because in 4K monitor,the MSAA will be no effect, i have an RTX2070 with 4K monitor, i also can get 60~90 FPS

Are you out of your mind? 

When you are on the ground, the lines on the airfield will flicker like hell without msaa, same for every straight line in the cockpit. I can‘t stand that.

If you‘re fine with it - go ahead, but don‘t assert that there is no effect. 

 

Make sure you haven’t accidentally activated some kind of anti aliasing in the nvcp.

Edited April 25, 2021 by Hiob

[Hiob]

Comparison of 4K with and without 2xMSAA:

And those are stills. In motion, those staircases flicker like hell, which - even on a 4K - is very annoying.

 

EDIT: Obviously those are cropped images. I would give you the originals, but they are to big to upload....

Edited April 25, 2021 by Hiob

[Hiob]

25-04-2021, 11:23:53 DCS.exe benchmark completed, 11552 frames rendered in 139.735 s (with MSAA 2x)

(Note that this is with the same "final balanced" settings from above, which then gave 81.8 FPS on average - but different day different baseline!)
                     Average framerate  :   82.6 FPS
                     Minimum framerate  :   73.4 FPS
                     Maximum framerate  :   95.0 FPS
                     1% low framerate   :   73.0 FPS
                     0.1% low framerate :   71.6 FPS

25-04-2021, 11:29:18 DCS.exe benchmark completed, 13371 frames rendered in 139.781 s (without MSAA)
                     Average framerate  :   95.6 FPS +13.0 FPS +15.7%
                     Minimum framerate  :   86.2 FPS
                     Maximum framerate  :  109.6 FPS
                     1% low framerate   :   85.0 FPS
                     0.1% low framerate :   82.4 FPS

Conclusion: Dropping AA completely would give 15% more FPS and close to a hundred in absolute, but I won't do it anyway - because it looks like crap....

[rajdary]

Just want to say thankyou, very useful info.

[bandit648]

Thanks for doing this. It's always nice to see what kind of performance people are getting in 4K. Here's my results with your medium settings. System specs are-

 

i9-10900K @ 5.2Ghz

64 GB RAM CL 16 @ 3600Mhz

RTX 3090 Founder's Edition | +120/400Mhz

LG OLED48CX

 

26-04-2021, 21:54:06 DCS.exe benchmark completed, 14916 frames rendered in 148.969 s
                     Average framerate  :  100.1 FPS
                     Minimum framerate  :   88.8 FPS
                     Maximum framerate  :  116.8 FPS
                     1% low framerate   :   85.5 FPS
                     0.1% low framerate :   68.2 FPS

[Hiob]

Hi @bandit648,

 

that‘s interesting to see the comparison to a 3090, to be really comparable you need to start and stop the benchmark at the correct timestamps. EDIT: when doing it right, you‘re overall fps should be a little (little!) lower, but you‘re min 0.1% lows should be around a lot better.

 

The Benchmark should be 139.x seconds long if done correctly. If you want a howto, let me know.

You meant, that you used my „custom“ medium settings, which I used as a baseline, correct?

 

btw... cool monitor, this thing is a blast, isn‘t it? I‘m using the same...

Edited April 27, 2021 by Hiob

[Hiob]

@bandit648

If you want you, could you perhaps do the benchmark (139.xsec!) also in the „max quality setting“ (detailed settings above)?

I‘d like to see how the 3090 scales with higher settings.

 

Thank you!

[bandit648]

@Hiob

 

Here are my results. I had some background services for Corsair and Asus running that I think affected my 0.1% lows before. This is a great way to fine tune performance. I really like your benchmark track. 

 

With Hiob's "Medium" Settings

27-04-2021, 08:21:55 DCS.exe benchmark completed, 13682 frames rendered in 139.328 s
                     Average framerate  :   98.1 FPS
                     Minimum framerate  :   88.2 FPS
                     Maximum framerate  :  115.7 FPS
                     1% low framerate   :   87.1 FPS
                     0.1% low framerate :   77.5 FPS

 

With Hiob's "Maximum" Settings
27-04-2021, 08:32:58 DCS.exe benchmark completed, 9509 frames rendered in 138.844 s
                     Average framerate  :   68.4 FPS
                     Minimum framerate  :   62.1 FPS
                     Maximum framerate  :   79.1 FPS
                     1% low framerate   :   61.6 FPS
                     0.1% low framerate :   57.9 FPS

 

I like your balanced settings and might start there and tweak it to my liking. I normally run capped at 60 FPS to get the smoothest TrackIR panning.  

 

Oh and yes, the LG OLED48CX is glorious! I have a Reverb G2 but don't even bother with it these days. DCS looks so beautiful on the OLED48CX and I don't have to hassle with VR settings every day. Here's a picture of my rig.. 

 

[Hiob]

You‘re setup is marvelous! 

 

Yeah, those numbers are better (more relatable).

 

Notice how the relative performanceof the 3090 gets better with higher settings? I think this is where it shows.

I am also at 60 fixed at the moment. I also lowered the power target of the gpu. The gpu utilization is between 60-80% and temps went down by 5-7 degree with auto-curve. The benefit of this settings is, that it gives headroom for more demanding scenes....

 

The only thing I find irritating about your numbers is, that your 0.1% low (on medium) isn‘t that much better than mine... perhaps there are still issues?

[Hiob]

@bandit648

 

One thing, that came to my mind, which I missed in my OP is the nvcp settings:

 

 

The only things that are not default are:

- Power Management Mode: Maximum Performance

- Texture Filtering: Clamp/High Performance

- Vertical Sync: forced off

 

...and obviously for benchmarking the Max Frame Rate in uncapped.

 

I played with "MFAA" and "Antialiasing Transparency", but those hit quite hard on performance with mediocre effects.

Edited April 27, 2021 by Hiob

[bandit648]

6 minutes ago, Hiob said:

The only thing I find irritating about your numbers is, that your 0.1% low (on medium) isn‘t that much better than mine... perhaps there are still issues?

 

I'm still troubleshooting this as well. I had Resizable Bar enable previously but disabled it before running benchmarks. Other things to note-

 

NVIDIA Driver version: 465.89

Windows Pro 20H2

KB5001330 is installed

NVCP Settings are default except Perfer Max performance

 

4 minutes ago, Hiob said:

@bandit648

 

The only things that are not default are:

- Power Management Mode: Maximum Performance

- Texture Filtering: Clamp/High Performance

- Vertical Sync: forced off

 

...and obviously for benchmarking the Max Frame Rate in uncapped.

 

The texture filtering settings might explain the difference in 0.1% Low numbers. I was set at Allow/Quality. What Nvidia driver version are you on?

 

 

[Hiob]

I'm did all testing on the latest drivers (466.11) and up to date windows.

 

Oh and my other system specs are:

5900x

64 GB 3600

PCIe4.0 nVME SSD

ResBar enabled in the driver and Bios but not tweaked into DCS - so no effect.

 

I don't have much in the background when running DCS. The TrackIR was off during benchmarking.

Edited April 27, 2021 by Hiob

[bandit648]

I had Resize Bar enabled in BIOS and NVInspector for DCS but it didn't seem to do any good for me. I had downgraded to 465.89 while troubleshooting another issue so I'm going reinstall 466.11 and check some other things. I will re-run the benchmarks shortly.

 

 

[C3PO]

1 hour ago, bandit648 said:

@Hiob

 

Here are my results. I had some background services for Corsair and Asus running that I think affected my 0.1% lows before. This is a great way to fine tune performance. I really like your benchmark track. 

 

With Hiob's "Medium" Settings

27-04-2021, 08:21:55 DCS.exe benchmark completed, 13682 frames rendered in 139.328 s
                     Average framerate  :   98.1 FPS
                     Minimum framerate  :   88.2 FPS
                     Maximum framerate  :  115.7 FPS
                     1% low framerate   :   87.1 FPS
                     0.1% low framerate :   77.5 FPS

 

With Hiob's "Maximum" Settings
27-04-2021, 08:32:58 DCS.exe benchmark completed, 9509 frames rendered in 138.844 s
                     Average framerate  :   68.4 FPS
                     Minimum framerate  :   62.1 FPS
                     Maximum framerate  :   79.1 FPS
                     1% low framerate   :   61.6 FPS
                     0.1% low framerate :   57.9 FPS

 

I like your balanced settings and might start there and tweak it to my liking. I normally run capped at 60 FPS to get the smoothest TrackIR panning.  

 

Oh and yes, the LG OLED48CX is glorious! I have a Reverb G2 but don't even bother with it these days. DCS looks so beautiful on the OLED48CX and I don't have to hassle with VR settings every day. Here's a picture of my rig.. 

 

 

 

HOLY SH!$£TG!!!

[bandit648]

OK.. here are my latest results. I DDU'd 465.89 and reinstalled 466.11. I also set NVCP Texture Quality to Clamp/High Performance, re-enabled Resizable Bar and disabled some non-essential background services. I'm getting some better results now.

 

Hiob's Custom "Medium" settings

27-04-2021, 11:40:36 DCS.exe benchmark completed, 13832 frames rendered in 139.469 s
                     Average framerate  :   99.1 FPS
                     Minimum framerate  :   89.3 FPS
                     Maximum framerate  :  117.5 FPS
                     1% low framerate   :   88.9 FPS
                     0.1% low framerate :   83.6 FPS

 

Hiob's Custom "Max Quality" settings

27-04-2021, 11:29:04 DCS.exe benchmark completed, 9819 frames rendered in 139.953 s
                     Average framerate  :   70.1 FPS
                     Minimum framerate  :   63.5 FPS
                     Maximum framerate  :   81.6 FPS
                     1% low framerate   :   63.2 FPS
                     0.1% low framerate :   61.5 FPS
 

[Hiob]

That looks really good! No frame drops anymore.

I mean you never had any trouble in the first place with your beefy graphics card, but it‘s interesting to see, that „issues“ (drivers, background processes and so on, can still be sorted out to get rid of the 0.1% lows.

 

I don‘t(!) have resizable bar enabled for DCS, but I don‘t think they do anything anyway....

 

My next update step will be custom water cooling - first cpu then later gpu (when the block becomes available).

Now, that I have a reliable baseline, I will try to measure the effect of each step. (including driver updates, game updates and so....)

Edited April 27, 2021 by Hiob

[SharpeXB]

On 4/25/2021 at 12:42 AM, flankerjun said:

you can set MSAA off，because in 4K monitor,the MSAA will be no effect, i have an RTX2070 with 4K monitor, i also can get 60~90 FPS

I definitely prefer to have AA set to 4x in 4K. There is a very noticeable effect. 

On 4/25/2021 at 3:21 AM, Hiob said:

Are you out of your mind? 

When you are on the ground, the lines on the airfield will flicker like hell without msaa, same for every straight line in the cockpit. I can‘t stand that.

This

[Thumbs]

I really like this method of comparing numbers with a standardized test.

 

I want to play but I do not know how.  

 

1.  Do you need to buy the F-16?

2. What do you do with the trk file?

3. Would you mind walking me through the exact steps to test?

 

Thanks for your help.

 

[Hiob]

Hi, I'm sorry but you will need the F-16 module I think.

 

Problem is - a benchmark that really everyone could use, would be a TF-51 in caucasus,

but a hind over marianas would be more interesting. I'm thinking of doing the latter.

 

Otherwise it's simple.

Just save the .trk-File and play it using the "replay"-funtion in the main menu.

 

For the Benchmark you'll need Afterburner/Riva Statistics Server (the latter comes with the afterburner install).

Just define Hotkeys to start/end benchmarking.

Then replay the track, start the benchmark at the marking point (aircraft banks right) and end it when it banks left.

The go to the afterburner settings/options and search for "view logfile" or similar (will edit the exact options name and place later...).

You find the Results in a txt-file.

[Ozmandi]

NICE post! 

 

My computer specs:

CPU - AMD 5800X (tuned with - CTR 2.1 RC5)

GPU - NVIDIA 3080 (undervolted and overclocked)

RAM - 64GB Corsair

HD - Samsung 980 Pro 1TB NVMe  

LG ULTRAGEAR 2560x1440 165Hz

 

I ran your track file with my current settings and here are the results.

22-07-2021, 21:25:40 DCS.exe benchmark completed, 12279 frames rendered in 139.953 s
                     Average framerate  :   87.7 FPS
                     Minimum framerate  :   81.0 FPS
                     Maximum framerate  :  106.6 FPS
                     1% low framerate   :   80.8 FPS
                     0.1% low framerate :   80.0 FPS

 

Tomorrow I'll use your MAX settings and post the results.

[Hiob]

Thank you!

Seems you have your machine well optimized in every way.

0.1 FPS low being virtually identical to „min“ and min and average being so close is a very good sign!

[Hiob]

By the way, I‘ve been messing around with the whole head position fov matter and „discovered“ that the fov has a  noteworthy effect on the FPS.

I wonder if that adds to the problem of some people on the marianas map….

[Hiob]

Disclaimer: The following post is outdated, since the trackfile doesn't work anymore (probably due to updates of the marianas map).

 

I made a new Benchmark for the Marianas map.

This time I used a TF-51 for the flight so everyone (who is interested) can replay it.

 

Instructions:

Load the attached file as "Replay". Use the integrated Benchmark recording function in Afterburner/Riva Statistics Server.

At the beginning of the scene I do some adjustments to the head position and fov. Immidiately after I give the Manifold Pressure a little jump, that is when you start the recording.

After roughly 150 sec. (148+/-) and a 360+ degree turn I rev the engine again (when crossing the runway), this is when you stop recording.

 

Your results should look something like this:

 

23-07-2021, 20:25:11 DCS.exe benchmark completed, 9847 frames rendered in 148.141 s (Marianas TF-51)
                     Average framerate  :   66.4 FPS
                     Minimum framerate  :   56.4 FPS
                     Maximum framerate  :   74.9 FPS
                     1% low framerate   :   56.7 FPS
                     0.1% low framerate :   55.9 FPS

 

Those are the Settings I used:

Just for comparison, the results with the same settings in Syria:

 

23-07-2021, 20:17:41 DCS.exe benchmark completed, 11254 frames rendered in 139.922 s (Syria F-16)
                     Average framerate  :   80.4 FPS
                     Minimum framerate  :   71.5 FPS
                     Maximum framerate  :   90.8 FPS
                     1% low framerate   :   71.4 FPS
                     0.1% low framerate :   69.6 FPS

 

Here is the track-file:

 

 

 

That's just for the start.

Detailed benchmarking will follow.

Edited June 22, 2022 by Hiob