Performance Guide For DCS World

[TZeer]

After a lot of hours with testing and trying out different settings, I want to make this guide to help all of you out there who are looking to get the most out of your DCS World installation.

 

PS: I will not cover the broad specter of mods out there that deal with optmizing, since that is a whole chapter for itself. This will simply be about getting the

the most out of your game with a default installation. This should then give you a good foundation to build upon if you so wish.

 

 

Before we dive into the actual settings and parameters I want to touch on the subject about GPU's and SLI/Crossfire.

 

There is a lot of people asking this on a regular basis, and the answers they get range from no to yes.

Many of the answers are pure guessing and rumors that someone picked up from some uncles cousins sister, if you know what I mean :)

 

- SLI

 

Yes, yes and yes. It works. I have had 2 x GTX 580, and now 2 x GTX Titan in SLI. In both setups I have been able to make SLI work without any problems.

I'm in the process of testing out the GTX 690 with a friend, and will update when this is done.

 

So, if you are looking for a new card or maybe multiple GPU's, I have no problems recommending a SLI setup.

 

- Crossfire

 

I have yet to see anyone posting evidence of crossfire working @ 100%. With evidence I mean a screenshot that shows both GPU's working at more then 90% or around that area. Alot of crossfire users have reported

better performance with only 1 card activated in DCS. And there have been some people that have reported working crossfire, but the performance has not been what to be expected.

I'm not saying AMD cards are bad, but not my first choice if I where going for multiple GPU's.

 

 

Testing your FPS

 

When people are testing their FPS, many do the mistake of doing it in the middle of citys, airports etc. I'm not saying it's wrong. But it's not the first place you want to test.

There are so many things that can interfere with your results, and you can never know for sure if it's your GPU or the CPU that is holding you back by just looking at FPS.

 

- Your maximum FPS possible

 

To test this you need to do your testing far, far, far away from anything that can resemble a building, road, tree etc. To do this you want to place your aircraft of choice somewhere

to the north where you are "outside" of the modeled world. Reason we do this is to be sure we are not restricted by the CPU. 3d models consists of multiple number of objects, the number of those

objects determines how much load you will get on your CPU, see this thread for a more in depth investigation on objects and their influence on FPS

http://forums.eagle.ru/showthread.php?t=101491

 

 

Another thing to do is to turn off your vsync in game and in the control panel of your GPU. This removes the lock that is put in place for the FPS to not go over the refresh rate of you monitor.

If it is active, your FPS will never go above 60 FPS (for 60Hz monitors). And if the scene is easy to render, your GPU will be strolling along with 40-50% GPU usage, and only using what it needs

to give out those 60 FPS. Thats not what we are interested in. We want to find out what your maximum FPS is when we are not restricted by the CPU or other stuff. This is also a good way to test

your SLI/crossfire setup. You also want to go into your graphic.lua file to line 185 (if you are using notepad++) and change the maximum FPS to 400, or something.

 

Below you can see me running one card and the GPU usage in that scenario. As you can see, my GPU is running at close to 100% and my FPS is ~105 FPS

 

 

 

Second screen you can see me running my GPU's in SLI, and my FPS is now 181, with both GPU's running at >90%

 

 

 

Reason we get these good FPS numbers is that the CPU has almost nothing to do, and is not a bottleneck for the performance. So in other words, the FM or avionics are not the massive resource hogs that some people think.

This does not mean they don't use resources. But if you are running a newer CPU like i7, i5 or something, FM and avionics is not the reason for bad performance.

 

Important: DCS World is not able to make use of multiple cores. In it's current state it uses 2 cores. 1 for sound, and 1 for the rest. So an i7 2600k @ 4GHz, will be better then a 6 core Sandy Bridge-E clocked at 3,7GHz for example.

 

 

Now we have established your maximum achievable FPS. Now is the time to start playing with your settings like cockpit shadows, TSSAA, MSAA, etc. You will now be able to see how much each option affects your performance.

 

My settings for the tests above was the following .

 

 

NB: There is a few options that will not be any point in testing at this location, like trees, visibility range. Since you have no trees and no buildings, roads etc to render at this point.

 

If you have a second monitor it's almost a must to have some sort of monitoring software running on that one. This will give you live data on the performance of your components.

I'm running nvidia inspector to monitor my GPU usage while testing. I really like this software as it gives live info with no lag in feedback.

 

If you don't have a second monitor, use some sort of logging function and look at the data between each test.

 

 

Testing your CPU:

 

We will do this at the same spot in the editor. Only change now is that we will check and see how a good model versus a bad model will interfere with your FPS.

For this we will first use ground vehicles. We will use 2 types to show you the difference between optimized 3d model and a old non-optimized model.

 

First you can start placing this unit, SAM SA-6 Kub STR 9S91 radar on the map. Place like 10 of them and check your FPS and object count (to see your object count, hit ctrl+pause/break two times). You should have about 700 objects in that scene now. Now go back to the editor and add some more units and monitor your FPS and object counter.

 

Now we can test out an optimized and well made 3d model. The M1A2 abraham. Change all your Kub radars to the M1A2. Now enter the game, check your FPS and number of objects.

Number of vehicles is the same, and the M1A2 is also a lot nicer then the Kub. But it's giving us a lot better FPS :)

 

At the end of both tests place a A-10C or any other aircraft of your choice on a direct approach to the vehicles. Now you can check your performance in cockpit in the two different scenarios.

 

(After some feedback I'd like to point out that the radar on the Kub etc, has nothing to do with the FPS drop we get. I have tested this with multiple models. And my reason for using the Kub, is that this model has one of the highest number of objects of what's available in the editor.

I could change to a static 3d model, but it would mean you would have to find the EDM file, move to a different folder, rename it, back up the original file you are replacing, and the reverse it when you are done. As I saw no difference in the performance, I saw no point in complicating things.)

 

Bad model, giving us 75 FPS.

 

 

Good model giving us 150 FPS.

 

So there you have it. Difference between a good model and a bad model :)

 

Playing around with all of this will give you a rough estimate of what your rig is capable of.

 

 

Graphic options:

 

I wont go into every option in the settings, but I will touch on a few that can really impact your performance.

 

- Visib Range: This decides how far out buildings and other stuff start popping up. Longer view = more 3d models = more objects = lower FPS

 

- Res. Of Cockpit Displays: Alot of people are running this at 1024 every frame. Not necessary. And it takes alot of performance. Personally I'm running it at 512.

In the scenario above with 20 M1A2 I'm getting ~120 FPS with the TGP on the tanks on a flyover. If I change to 1024 every frame this drops to 75 FPS if I go with the black hot/white hot mode.

 

- MSAA: Some people like to take this to the max. I have taken 2 screen shots. One with MSAA at 4x and one at 16Q.

 

 

 

Not much difference between them is it? Except 170 FPS versus 88 FPS.....

 

- TSSAA: This also takes a lot of performance. People has to decide if it's worth it.

 

Here I have a screenshot of a flyover with the TGP on the tanks. I have turned on TSSAA, MSAA 16Q and 1024 every frame. My FPS is now down to 52 FPS. Do I really gain

anything by turning that stuff that high?

 

 

And the last one I have switched on cockpit floodlights. I'm now down to 34 FPS.

 

 

What I'm trying to show here is that even if you have the fastest GPU on the market, you can still kill the performance if you are not careful with your settings. Turning everything to the max

doesn't alway give great returns in quality when you compare to the reduced performance.

 

 

Hope this helps a little on how to test your system and find your sweet spot when it comes to performance and eye candy :)

 

 

Cheers

Edited June 1, 2013 by TZeer

[Mustang]

I'd like to add that altering line 146 in graphics.lua can help substancially with frame rates - change the default to the following (always use notepad++ when performing these edits)

 

structures = {20, 10000};

 

this should actually be the correct setting for this parameter, if you compare to the High.lua line 56 you can see that the districts/cities are set to render at 10000 meters, the same should apply to the graphics.lua.

 

Just incase you're wondering there is no perceivable loss of image quality doing this and the FPS gains are well worth it.

 

(Good guide BTW :))

Edited June 1, 2013 by Mustang

[Blooze]

I'd like to add that altering line 146 in graphics.lua can help substancially with frame rates - change the default to the following (always use notepad++ when performing these edits)

 

structures = {20, 10000};

 

this should actually be the correct setting for this parameter, if you compare to the High.lua line 56 you can see that the districts/cities are set to render at 10000 meters, the same should apply to the graphics.lua.

 

Just incase you're wondering there is no perceivable loss of image quality doing this and the FPS gains are well worth it.

 

(Good guide BTW :))

Can you provide a path to this file?

[Nealius]

I'll have to give this a shot tomorrow. Would you suggest anything other than SA-6s, that will give the same results, so I can fly around without being shot at? :D

[TZeer]

I'll have to give this a shot tomorrow. Would you suggest anything other than SA-6s, that will give the same results, so I can fly around without being shot at? :D

 

It has no missiles, it's a radar.

[GeorgeLKMT]

Can you provide a path to this file?

 

It's in the Config folder

[Nealius]

It has no missiles, it's a radar.

 

Ah, just the radar. All the pictures of SA-6s I have seen only showed the missile launchers :doh:

[slowhand]

Great Post:thumbup:

[Kuky]

Reason we get these good FPS numbers is that the CPU has almost nothing to do, and is not a bottleneck for the performance. So in other words, the FM or the avionics has nothing to do with your low FPS. Myth busted ;)

 

While most of the stuff you have correct this one you don't, reason you think FM or the avionics has nothing to do with your low FPS is because with nothing around, like you said CPU has not much to calculate... it however does need to calculate that FM and avionics its just that the video cards are the only thing at that moment that is maxed out, not the CPU.

 

Also, optimized model vs unoptimized model... I am not suggesting one is optimised and other is not, it might as well be because SAM has radar working (many of them in same scene) it needs more to calculate. The tank has no radar hence nothing to calculate (and hog the CPU) which is why you see lot more FPS.

[TZeer]

While most of the stuff you have correct this one you don't, reason you think FM or the avionics has nothing to do with your low FPS is because with nothing around, like you said CPU has not much to calculate... it however does need to calculate that FM and avionics its just that the video cards are the only thing at that moment that is maxed out, not the CPU.

 

Also, optimized model vs unoptimized model... I am not suggesting one is optimised and other is not, it might as well be because SAM has radar working (many of them in same scene) it needs more to calculate. The tank has no radar hence nothing to calculate (and hog the CPU) which is why you see lot more FPS.

 

I was waiting for this one ;)

 

Yes, the FM and avionics still need CPU power, but not in the quantity that some people think.

 

But we can look at it like a glass of water.

 

If you have an old and slow CPU, the glass is already 3/4 full, before adding the tanks.

 

If you have new and fast CPU, the glass is maybe only 1/5 full, before adding the tanks.

 

When the glass is full, the glass is full, it will not matter what's in the glass, but we know it's not able to contain any more.

 

So for people with old CPU's, yes the FM and avionics might take up much of their CPU.

But I'm leaning more towards the actual cockpit taking the resources, since going external boosts my FPS up to 300.....

 

But I will rewrite that part so people don't get confused.

 

Regarding the SAM. Negative. I have done this with multiple other objects with same results. The difference is that the SAM is easy accessible in the editor, and one of the models with the most objects.

 

It's radar has no impact, as I should have had a significant FPS drop entering it's radar coverage, which I don't.

 

I could have used other models with similar objects, but those are static objects from the game world, and would need people to move EDM files and renaming them to get them into the editor.

 

And since I saw no difference between the Kub and a static object, I decided to use the Kub so people didn't have to start moving around files, renaming, and maybe breaking something in the process.

 

 

Thanks for the feedback :)

Edited June 1, 2013 by TZeer

[Kuky]

Waiting for it aye... LOL :D

 

About DAM radar... to confirm if radar working taks a CPU hit, try that same mission but set SAM's to be of same country as aircraft you fly... I think they won't be working in this case, and if they don't the FPS might be higher. If not then I'm wrong :)

[TZeer]

As requested :)

 

Here is the default one, you can see the radar is active and they are painting me good.

 

 

 

Here is your requested test, as you can see, no radars are painting me :)

 

 

Both showing the same amount of FPS.

 

No performance hit due to radars :)

Edited June 1, 2013 by TZeer

[Kuky]

Nice one... good testing, need to REP ya' for this one but can't :( so I'll owe you one ;)

[TZeer]

Nice one... good testing, need to REP ya' for this one but can't :( so I'll owe you one ;)

 

I'll rep your for your awesome increased visual range mod ;)

 

 

Edit:

LOL, I cant' rep you. I have to spread some reps around. Haha.

Edited June 1, 2013 by TZeer

[Aries]

I can be your mutual rep buffer... I'll take good care of them reps for ya... :music_whistling:

 

EDIT:

THX guys!!! :thumbup:

Edited June 2, 2013 by Aries

[drummer]

A really great and informative post, thanks for your work!

[_Dredd]

Nice post, thanks for the testing write up.

You've got me interested in SLI with Titans now. Though my only question now is the performance increase on multi-monitors. Don't suppose you run multi-monitors as yet?

 

Is there anyone out tree with SLI cards that can add to this for any benefit gains?

[BIGNEWY]

As an ATI user who is resigned to the fact that until we get EDGE the chances of getting Crossfire to work correctly is poor.

 

I have tried all different ways and failed, closest I got was using crossfire and SoftTH, but even then it was not great.

 

Since then I have opted for 1 card ( yes still ATI LOL ) I have taken a gamble hoping EDGE will benefit ATI cards.

 

When I run the test mentioned above I get a good 120 FPS which is great, but of course as soon as you are on a mission and there is a lot going on it drops 60 > 40 > 17 depending on what is going on.

 

DCS needs optimizing big time, and I have said in other threads I do not think they will optimise it, why would they ? an out dated outgoing graphics engine would take up to many hours to put right.

 

I just hope we do not have to wait to long for EDGE.

[winz]

"You must spread some Reputation around before giving it to TZeer again..."

 

Great post.

[_Dredd]

Went through some of this testing tonight :joystick:; and have found that its best for me to leave fps unlocked and Vsync Off to get the best benefit when frames aren't impacted so fps is 60 or higher.

 

I'm quite happy now, and may very well look at getting a second Titan to SLI.

This should give me a 20% boost to keep frames at 60.

 

With attached settings(screenshot) is on triple monitors(plus a 4th for Helios) at @ 6696x1920

I so far saw the worst fps at 39-41fps when low, and the best at 130fps(map screen).

 

Avg flight with good fps in A10C was 57-62, and F-15C 61-81.

Interested thing to note was MSAA didn't impact much at all :)

Grass had a huge impact when above 0m, frames drops to 31 when zoomed in close to vehicles.

 

Another Rep your way when it's avail to add. :D

[aceflier]

I've said this a hundred times but here goes again.... Its not ED's fault ATI crossfire doesnt work right its ATI's job to fix this.

[xtaffy]

Hello,

 

What is the best way to measure to measure FPS in DCS in BS

 

Cheers

 

Nigel

[Kuky]

Hello,

 

What is the best way to measure to measure FPS in DCS in BS

 

Cheers

 

Nigel

 

easiest is use built-in FPS counter (press LCTRL+Pause key, press same again and you get more details, press again and it turns off)

[doebro2010]

Hello

I have a question regarding my setup

I have 4 cores processor 3770k and a gtx680 2.0gb

Is it better default with 3.5 + or to overclock it to 4.6 which I did for other games.

I am asking this because I read that only 2 cores are needed.

I tried a couple of months before to play but I had constantly stutters

[Prophet]

DCS seems to be more CPU dependant. I don't know how much, but you should see FPS improvement from overclocking. I have heard there are GTX6xx series problem though, not sure if that was resolved.