Performance problems, a few questions.

[tyler2 10]

Image number 1: height = 1k, Video Settings = Low , OverallVisibility = 12k, ObjectVisibility = 12k, sceneComplexity = 1

Image number 2 : Height = 1k, Video Settings = Low , OverallVisibility = 12k, ObjectVisibility = 0.5k, sceneComplexity = 1

Image number 3 : Height = 1k, Video Settings = Low , OverallVisibility = 12k, ObjectVisibility = 12k, sceneComplexity = 1000000

Image number 4 : Height = 30k, Video Settings = Low , OverallVisibility = 12k, ObjectVisibility = 12k, sceneComplexity = 1

Image number 5 : Height = 30k, Video Settings = Low , OverallVisibility = 12k, ObjectVisibility = 0.5k, sceneComplexity = 1

Image number 6: Performance in the game justcause2, the range of drawing objects over 20k on a server with 500 people playing

First question: What is spent on the performance of image number 1 if the image number 2 I get 300% more FPS and graphics quality is the same ?

Second question: Why is the image number 3 with lots of objects I get 25 FPS, and the number 1 in the image without objects I get 30 FPS ? Productivity gains only 17 %.

Third question: Why is the image number 5 (the number of objects = 0 ) I get 169 FPS , and the image number 4 (the number of objects = 0 ) I get 125 FPS ? What consumes 26.5 % of the performance ?

Edited January 6, 2014 by tyler2

[Brisse 78]

Your settings are kind of ridicoulus. Just becouse you can set "ObjectVisibility = 12k", does not mean you should.

You can't blame the game when it's you who broke it by using bad settings.

[tyler2 10]

I do not understand what you mean. I'm interested in why an increase in visibility of objects (the scene has no objects) performance decreases. Perhaps the answer to this question will explain why all players have such low FPS with high visibility in the scene with objects. This is a very important issue, since the loss of performance is very high.

[Brisse 78]

Only the devs can give you and answer and I doubt they are going to give you one. I just got the feeling that you were trying to bash the game or devs for not optimising properly, which in my opinion is not fair. We have no idea what goes on under the hood. Arma has a lot going on besides graphics and just the amount of objects rendered is just a tiny part of the equation. There are more variables to consider.

Comparison to Just Cause 2 is probably not fair either. Not that I have played it but I can imagine they are very different games.

[tyler2 10]

I hope you do not think the image number 5 computer processes more information than the image number 6? I'm not just talking about the graphics, but also other complex things such as scripts. The fact that the image number 5 should only process computer graphics (by the way there is nothing to handle), and number 6 in the image processing computer graphics, scripts, and synchronization with 500 people. Type of game does not matter too huge difference in FPS. This difference is several times.

[ratszo 17]

Pointless thread.

Troubleshooting is where player help other players.

Are you looking for advice on settings, or what?

[lappihuan 178]

No he has some valid points.

Even if those settings are not usable references, there is strange behavior of fps regarding to what is rendered.

I think he just wants to find out what causes the fps loss without a visual change.

[windies 11]

One engine != another engine. RV is not the same as what JC2 is running on. Even still, Yeah there is something of a point in that if that engine can draw those kinds of distances with that many people in a server at one time on a multiplayer mod, what exactly is RV's problem since it's goal is pretty much the same thing. One thing is that the amount of objects seems to play hell with performance. It doesn't even seem to matter what LOD's are being displayed, just simply having an object being rendered destroys performance.

In your second question the difference is that you have object view distance at 12K in one and .5K in another, scene complexity if I'm not mistaken is just the LOD distance for objects.

[tyler2 10]

Discovered an interesting fact: Configuring visibility of 12,000 meters, then place the helicopter at an altitude of 56,000 meters, you get 200 FPS. After that the helicopter itself will fall. At an altitude of 55,000 meters below you get a sharp decrease FPS several times. Then do the same thing, only with a visibility range of 6000, at an altitude of 55,000 meters below you get a sharp decline in FPS.

If I understand correctly, this means that all objects in the game begin to be processed on the computer range 55000 meters instead of visibility that you configure in the game. In this regard, I daresay, if this option (which is responsible for 55,000 meters) to make customizable, and bind it to the range of visibility, you can get a huge performance boost. But this theoretical conjecture, I would like to hear the opinion of the developers on this.

[blackpixxel 53]

I don't know why People are questioning your work, you found some very interessting points.

I would like to add a similar thing with the shadows. When you have disabled them and adjust your shadow draw distance it will have an impact on your FPS. But why? Why sould something that is not there affect the FPS?

[Sneakers O'Toole 2]

Pointless thread.

Troubleshooting is where player help other players.

Are you looking for advice on settings, or what?

Pointless post

Trouble-shooting is where scenarios are brought to light that stress the engine in particular ways to shed light on problems/innefficiencies.

What are you looking for, besides displaying your rabid fanboyism, or what?

[kklownboy 43]

..

Trouble-shooting is where scenarios are brought to light that stress the engine in particular ways to shed light on problems/innefficiencies.

...

No ,this is a General section thread. And not a trouble shooting thread/post. Because there is no problem here.

But the short answer is Clutter. And the JC2 isnt really at 30k... its all make believe, where as the "ARMA Verse" is more in scale.

[ratszo 17]

Pointless post

Trouble-shooting is where scenarios are brought to light that stress the engine in particular ways to shed light on problems/innefficiencies.

What are you looking for, besides displaying your rabid fanboyism, or what?

Pointless flamebait.

[doveman 7]

Funnily enough, I was debugging DCS World trying to see why the fps was dropping to unplayable levels (16-19fps) and that turned out to be due to the number of "objects" as well. Now maybe ArmA "objects" are a completely different thing to DCSW "objects" but assuming they both losely refer to the thing then it would appear that DX9, at least in conjection with these engines and on certain GPUs (AMD seems to be worse than Nvidia from what I've heard), doesn't cope with large numbers of "objects".

See these screenshots I took in DCSW where the number of objects is shown in the yellow text top-left and as that increase, the fps drops. I forgot to take a screenshot at the start, when it was at it's best at around 9-10 objects and 40-44fps.

http://forums.eagle.ru/showthread.php?t=107551&page=6

[tinemem 11]

poor or no culling method. even underground, objects affect framerates in arma 3.

[jumpinghubert 49]

Discovered an interesting fact: Configuring visibility of 12,000 meters, then place the helicopter at an altitude of 56,000 meters, you get 200 FPS. After that the helicopter itself will fall. At an altitude of 55,000 meters below you get a sharp decrease FPS several times. Then do the same thing, only with a visibility range of 6000, at an altitude of 55,000 meters below you get a sharp decline in FPS.

If I understand correctly, this means that all objects in the game begin to be processed on the computer range 55000 meters instead of visibility that you configure in the game. In this regard, I daresay, if this option (which is responsible for 55,000 meters) to make customizable, and bind it to the range of visibility, you can get a huge performance boost. But this theoretical conjecture, I would like to hear the opinion of the developers on this.

this 56.000m thing is very interesting! Nice finding.

[tyler2 10]

this 56.000m thing is very interesting! Nice finding.

I think it makes sense

[tyler2 10]

Discovered another interesting fact:

The first time I put the camera in the far west of the island at an altitude of 55,000 feet above sea level, at an altitude of 54,899 meters has been a sharp decrease in FPS.

The second time I put the camera over the island at an altitude of 55,000 from the surface of the earth at an altitude of 54690 has been a sharp decrease in FPS.

Height of land above sea level in the place where I put the camera for the second time is 209 meters, so 54690 + 209 = 54899, is the same as in the first case. This means that the distance at which water begins to boot the computer is always 54899, because of this great loss FPS! What do you think about this? I think that it is unnecessary to load water at a distance of 54,899 meters from the player, it leads to poor performance :) This range of load water need to bind to the overall visibility.

[Brisse 78]

What do you think about this? I think that it is unnecessary to load water at a distance of 54,899 meters from the player, it leads to poor performance :) This range of load water need to bind to the overall visibility.

Now it's getting interesting. Good find...

[tyler2 10]

Created ticket: http://feedback.arma3.com/view.php?id=16779