Performance impact of every graphics option

[chiliwili69]

For a PC, which is not constrained by GPU, and running in a FullHD (1920x1080) monitor, it is calculated the fps impact of every graphic using a benchmark.

More info in this thread: https://forum.il2sturmovik.com/topic/56485-benchmark-for-cpuram-performance-remagen-40024003/?do=findComment&comment=878212

 

 

[von_Tom]

 

Good job.

 

I wonder if the fps hit for various options differs with different pc components.

 

von Tom

[RedKestrel]

This is very useful for showing how to get some extra FPS without seriously impacting the look of the game.

How does one tell if the game is constrained by GPU or CPU?

[=TBAS=Sshadow14]

12 minutes ago, RedKestrel said:

This is very useful for showing how to get some extra FPS without seriously impacting the look of the game.

How does one tell if the game is constrained by GPU or CPU?

If you GPU is pegged at 99% usage while gaming and you are getting low FPS or Stutters the cpu is barely being loaded up then the GPU is the bottle neck in most cases
(sometimes other components, like bad motherboard with slow pci-e lanes, bad ram, settings can effect this) 

If your CPU is pegged at 100% usage while gaming and you are getting stutters but your GPU is not being used properly (like 25-75% core usage) Then in most of these cases the cpu is the bottleneck. (tho again, poor motherboard, and other components can effect this,) 

(Note some cpu usage "Meters" that show a single cpu core % (like MSI afterburner) 
will read cpu usage wrong by averaging out the usage over all cores then showing that number, eg.
It may show the cpu is at 25% but its really Seeing this and going 100% Divided x 4 = 25% load.
core 0 = 100%
core 1 = 0%
core 2 = 0%
core 3 = 0%
 

[LLv24_Zami]

On 1/8/2020 at 2:28 AM, chiliwili69 said:

For a PC, which is not constrained by GPU, and running in a FullHD (1920x1080) monitor, it is calculated the fps impact of every graphic using a benchmark.

More info in this thread: https://forum.il2sturmovik.com/topic/56485-benchmark-for-cpuram-performance-remagen-40024003/?do=findComment&comment=878212

 

 

Good work. What's that Remagen benchmark you have used?

[Heckpupper]

For all it's issues, BoX is pretty well balanced I think. As long as you have a medium-decent PC, running it at stable 60 is hardly a problem. 

Same definitely cannot be said for other titles in the genre.

[chiliwili69]

7 hours ago, LLv24_Zami said:

Good work. What's that Remagen benchmark you have used?

Thanks.

The benchmark is described with instructions in the below link. Everyone can run it if they want to know if their PC is delivering the expected performance.

https://forum.il2sturmovik.com/topic/56485-benchmark-for-cpuram-performance-remagen-40024003/

 

[FTC_ChilliBalls]

Thx, though I‘m a bit confused. Are the costs or gains in FPS or in %? 

[chiliwili69]

On 2/21/2020 at 6:41 PM, So_ein_Feuerball said:

Are the costs or gains in FPS or in %? 

 

They are in fps.

[Velxra]

Can you post another example where the base line is the lowest settings as well? Overall a very nice spread sheet. 

[kestrel79]

Great find! I imagine this took a while to do. Would be curious to see if it's any different after a couple updates on a modern build of the game. I'm going to use this to help tune my VR settings.

 

I still need to get the courage to overclock my cpu, and try Open Composite for my Rift S to get a couple more free fps out of the VR setup. Thanks!

[Dutch2]

I did do some testings a month ago on most of the adjustments as here being published here and I can confirm thats mostly the same when searching for improvements. Only the amount of FPS is a bit different, like for a sample SSAO is on my rig only 1.5 FPS after 4 testruns. 

[SCG_motoadve]

On 1/7/2020 at 4:28 PM, chiliwili69 said:

For a PC, which is not constrained by GPU, and running in a FullHD (1920x1080) monitor, it is calculated the fps impact of every graphic using a benchmark.

More info in this thread: https://forum.il2sturmovik.com/topic/56485-benchmark-for-cpuram-performance-remagen-40024003/?do=findComment&comment=878212

 

 

What if I am flying a plane with no mirrors like a 109, then its the same as having mirrors off, correct?

We only loose the 14.7 fps with mirrors at medium setting and only with planes with mirrors I assume?

[chiliwili69]

2 hours ago, SCG_motoadve said:

What if I am flying a plane with no mirrors like a 109, then its the same as having mirrors off, correct?

Those test are two year old, since then some changes have happen in IL-2 like referred rendering (which load a bit more GPU) and also new clouds and new sky (which load more GPU). Also, those tests were done with the Remagen benchmark which is a 3 minute flight with a Spitfire with a variety of terrain and landscape.

 

Answering to your question, yes, I assume having them off is the same than flight a plane with no mirrors. But I have not tested that with a 109.

2 hours ago, SCG_motoadve said:

We only loose the 14.7 fps with mirrors at medium setting and only with planes with mirrors I assume?

All those tests were done with High Settings preset (base case is with High Settings preset)

[IckyATLAS]

GPU's have improved, CPU's have improved and even the hardware RAM, SSD etc have also improved. IL2 has gone from 4.003 to 4.704 and the graphic engine has been changed at least two to three times.

My PC setup was made last year with Intel I9 11900K, Win10, and a RTX 3090, 125GB Fast DDR4 RAM, SSD NVMe Gen 4. Display is 43" full 4K (3840x2160).

Here is a small benchmark I made just for getting an idea of the impact on my setup.

Map is Kuban Summer, 12 fighters flying and fighting,  there are trains, trucks, tanks, ships and AAA batteries.

My plane is a LA5FN Ser2. I choose that one because it has the largest mirror surface.

All game graphic settings are to the maximum/ultra/extreme. Mirrors graphics settings are set to Complex.

Canopy reflection are set to High. Visibility is 150km and the Distant Buildings parameter is checked.

Antialiasing is MSAA X4.

Atmospheric conditions are with clouds of the Cumulonimbus type which are large ones with vertical high vertical structure. Cloud level is 2000mt.

Clouds graphic settings are set to Extreme.

Vertical sync is deactivated.

My plane flies at 2000mt.

 

The FPS are measured in the cockpit (full mirror in view and clouds can be seen in the mirror) with planes, and landmass, and clouds in the field of view. This is to ensure we have a well loaded image.

 

In those conditions as mentioned above, the difference the mirror makes is 5 fps on sustained 120-125 fps without mirror.

If the sky is completely clear then the mirror still makes a 5 fps difference but this time on 130-135 FPS without mirror.

I would say for a setup like mine it is negligible in the sense that I do not feel any difference when flying a mission and I have to look at the FPS.

Now just for the record if I hit F2 and go outside of the cockpit with an external view of my plane, in clear sky FPS jumps to 180 if the field of view is cluttered with other planes and landmass and up to 280 if my plane is alone in the field of view. Clouds + Mirror will bring that down to 140-150 and 250, so the impact is larger here but the values change a lot as you move around so does not give much information here.

 

Now the combination of clouds and mirror with the maximum settings eats up at least 10 FPS. This means that if your gaming rig setup is in the same conditions delivering 50-60 FPS then 10 FPS is a big hit. But you can improve that by lowering your graphic settings overall and compensate for it.

 

 

 

 

 

Edited May 10, 2022 by IckyATLAS

[moustache]

hello @chiliwili69, are these tests carried out only in planes or with tanks too?

 

On 1/8/2020 at 1:28 AM, chiliwili69 said:

 

for me, the antialliassing and the quality of the grass have an impact...

[chiliwili69]

14 hours ago, moustache said:

are these tests carried out only in planes or with tanks too?

 

You bring three good points to this table. Before answering that I will need to state some points:

 

- Those test were done using the Remagen track benchmark from Nov-2019. That benchmark was a 3 minutes flight in the new Rheinland-Summer map at 19:00 around the famous Remagen bridge. Just two SpitfireIX vs 4 Ju88 and 1 He-111 H16 at 600m altitude. Wheather average and clouds at 1200m. As far as I know there is no video about that so it you can not know how it was looking like.

 

- At that time there were one antialising option technique. Which it is our current FXAA. So, not MSAA at that time.

 

- In the last two years, the IL-2 engine has suffered two important graphic engine update: new Deferred shading with version 4.006, new clouds 4.701 and new sky/light 4.703. The Deferred change was loading a bit more the GPU and unloading the CPU IIRC.

 

So, answering to your points:

 

- I think there were not ground vehicles IIRC, the last part of the track was a low flight over remagen town and flight below remagen bridge. I don´t remember if there was any ground unit.

 

- I think the grass settings affects when you are in the ground. At 600m there is not effect. Do you know if grass settings affects CPU or GPU?

 

- Antialising: Since the MSAA technique appeared after I made this comparison, I made an specific analysis of that here.