[REPORTED] Performance Issues

[Eldur]

I noticed one interesting thing.

 

When I look at the Island, my frames are low (<30 fps) and GPU is working only at ~60%

 

But when looking in the opposite direction, performance is a lot better (>70 fps) and GPU usage jumps to 100%

 

CPU & memory usage is roughly the same between two shots, so they're probably not my limiting factor.

 

Wrong. It's the CPU that is severely bottlenecking in DCS, literally everywhere. I rarely ever see more than 40% of GPU on my rig (3570k OCed to 4.4, 1080) even in DCS, expect for when sitting in the aircraft selection dialogue where it literally seems to melt, with the fans blowing at what sounds like 200% rpm. I've seen that in many games, so it's not just a DCS thing - for instance, I could raise the temp of my 3870 well over 130°C back in the day by simply sitting in the main menus of some games, it usually was at 40° on idle and at 60° under full load.

 

Simply put, as long as DCS doesn't do multi core or even multi thread, it would need a 15GHz single core CPU at least to run decent in VR and MP with the SC that has some assets on it, if you don't wanna drop your details. TBH, I'd love to see how this performs on one of those OC world record rigs with liquid nitrogen poured onto them, overclocked to the moon.

 

Fun thing about the shadows fps issue on the SC is that I absolutely had no problems with the first SC version at all while literally everyone was ranting about the low fps there, even with them set to high (terrain flat OFC), but then the first following update totally wrecked my performance and I haven't even touched the SC since then anymore except for some quick checks to see if it got any better. This really was one of the things I hoped to be addressed by those "we're going to do nothing but bug fixes for now" updates.

[imacken]

Wrong. It's the CPU that is severely bottlenecking in DCS, literally everywhere.

 

That is simply not true as a general statement. Depends on your GPU, CPU and settings.

 

In some scenarios, e.g. SC, it’s the CPU, in others, e.g. flying low in Dubai, it’s the GPU. With very high settings, it is more likely that the GPU is the bottleneck.

[statrekmike]

I still think the best solution here would be for ED to do some experiments for themselves and see where the greatest performance hits are with the Supercarrier. From there, figure out some kind of option that can go in the "special" settings that switches to a version that doesn't cause such a massive hit.

 

I strongly suspect that the changes required to get the SC module to run roughly similar to the rest of the sim would not be so great that it would cause a major issue anyway.

[...]

Exactly, it is not an exclusive problem of the Supercarrier. This also happens in Nellis (Nevada) in one of the predetermined missions. Looking in any direction gives more than 70 FPS and looking at the parking area reduces the FPS to 45/50. Nor is it a problem unique to DCS, as they want to imply. This problem is present in most flight simulators and I have tried all of them. If the Stennis gives 60 FPS and the Supercarrier with ultra-detailed textures gives 45, something that is basic math. All this fighting to beat you 5 FPS? Because you will agree with me that only a computing miracle could match the FPS of the Stennis and the Supercarrier. On the other hand, we also do not need studies and tests that tell us that eliminating the deck crew and deactivating the shadows improves the FPS because it is obvious, it is something we all know.

[Danny_P]

Whilst there probably is things they could do to 'improve' performance at the moment. I think that we'll be unfortunately waiting on the 'new' graphics engine to see anything major. Real time lighting is expensive, only recently have other engines made massive improvements on the number of lights and objects casting shadows. I feel that a large amount of the performance and graphics issues all fall into this pot. Only time will tell eh.

[Minsky]

Whilst there probably is things they could do to 'improve' performance at the moment. I think that we'll be unfortunately waiting on the 'new' graphics engine to see anything major.

 

If your engine is not up for the job, then maybe you shouldn't create such complex modules in the first place?

 

Once again I'm gonna agree with statrekmike. It could be something really obvious and easy to fix, but it's either not their top priority, or they just don't have time right now.

 

At least give us an update. It's been months. Why I must pester the developers themselves while their PR guys don't do public relations?

Edited August 11, 2020 by Minsky

[Danny_P]

If your engine is not up for the job, then maybe you shouldn't create such complex modules in the first place?

 

Once again I'm gonna agree with statrekmike. It could be something really obvious and easy to fix, but it's either not their top priority, or they just don't have time right now.

 

At least give us an update. It's been months. Why I must pester the developers themselves while their PR guys don't do public relations?

 

Oh I completely agree about the pestering, sometimes it just feels as though there's no true acknowledgment apart from [investigating]. And yes it may well be an easy fix, there are many of them in the list ;) But unfortunately time and time again we are seeing a clear ceiling that we are crashing into in terms of complexity and performance.

[TOViper]

I personally think in a way Minsky does.

 

To use my own words, I personally think that there is something "strange" happening in the calculation/drawing of the details of the SC, deck-crew, lights .. shadows. Maybe the frame drop is due to unneccesary "loops" in the programming, or some delays ... I don't know.

What this topic is all about for me is that ED should really take a deeper look OR say "there is nothing to check/optimize, sorry guys" (if that was the truth) or say "guys, we are looking into it very soon/now".

Or say "guys we know, but currently we have other tasks to complete before".

 

BUT for god's sake: COMMUNICATE!

Easy, hugh? :D

Edited August 11, 2020 by TOViper

[statrekmike]

Exactly, it is not an exclusive problem of the Supercarrier. This also happens in Nellis (Nevada) in one of the predetermined missions. Looking in any direction gives more than 70 FPS and looking at the parking area reduces the FPS to 45/50. Nor is it a problem unique to DCS, as they want to imply. This problem is present in most flight simulators and I have tried all of them. If the Stennis gives 60 FPS and the Supercarrier with ultra-detailed textures gives 45, something that is basic math. All this fighting to beat you 5 FPS? Because you will agree with me that only a computing miracle could match the FPS of the Stennis and the Supercarrier. On the other hand, we also do not need studies and tests that tell us that eliminating the deck crew and deactivating the shadows improves the FPS because it is obvious, it is something we all know.

 

 

I think you are completely mischaracterizing the issue and for the life of me, I can't figure out why you seem so aggressively committed to doing so. If this were simply a issue where a nicer looking thing runs a bit worse, I don't think this thread would be going as long as it has. Everyone here understands that you will see varying levels of performance depending on the visual complexity of a given module, map, etc. As you say, "It is obvious".

 

The problem here is that we are suddenly seeing a huge performance hit on one specific thing and there isn't any real justification for it in the visuals that we are seeing. The Supercarrier looks nice but it doesn't look -30 FPS nice. I could load the Stennis's deck with F-14's (the heaviest weight module visually) and still get better performance than I would on a completely empty Supercarrier deck. Does that seem normal to you? Does that seem justified considering the visual complexity of the Supercarrier module itself?

 

The core issue here is that out of everything you can do in DCS, the Supercarrier itself (again, to be VERY clear, with nothing on the deck but the player aircraft and the default deck crew) cuts the framerate in half without any real sign as to why in its basic visual design/complexity.

 

As I have said often in this thread, I am running a system that exceeds the recommended requirements for "high settijngs" as listed for the Supercarrier module itself and I see a 50% drop in overall framerates. I go from a nearly rock solid 60 FPS for everything else in the sim to 30 FPS on a empty Supercarrier deck. It doesn't make sense. There is something going on beyond "You need a better computer" or something to that effect.

[Lange_666]

Exactly, it is not an exclusive problem of the Supercarrier. This also happens in Nellis (Nevada) in one of the predetermined missions. Looking in any direction gives more than 70 FPS and looking at the parking area reduces the FPS to 45/50.

 

 

I second this, it's not a sole SC problem. As La Unión | Atazar says, it happens in other places and with different aircraft too:

 

 

Examples:

 

1: diff between A-10C and F/A-18, same location: https://forums.eagle.ru/showpost.php?p=3682733&postcount=1

2: difference in view direction, both F/A-18 now: https://forums.eagle.ru/showpost.php?p=3685246&postcount=9

[Minsky]

I second this, it's not a sole SC problem. As La Unión | Atazar says, it happens in other places and with different aircraft too:

 

No, I still think he's talking nonsense.

 

But your findings are really helpful, thanks.

 

If it's indeed a common DCS issue, then I guess the Supercarrier just made it more obvious and easier to trace. Since everybody is using the same boat, one of two planes, and it's usually far enough from the shore and irrelevant objects.

Edited August 11, 2020 by Minsky

[Minsky]

To use my own words, I personally think that there is something "strange" happening in the calculation/drawing of the details of the SC, deck-crew, lights .. shadows. Maybe the frame drop is due to unneccesary "loops" in the programming, or some delays ...

 

Here's a wild guess.

What if this is not GPU-related at all?

 

You see, before the upgrade I've had GTX960. 25 fps, 100% GPU usage, 50% CPU usage. So, probably my GPU was the limiting factor, right?

 

Now I have RTX2060. ~30 fps, 50% GPU usage, 50% CPU usage. A bit more frames, but now I'm bottlenecked somewhere else.

 

And since DCS can't properly utilize more than two cores, and it doesn't need all of my GPU horsepower (because, let's be honest, Supercarrier is not the prettiest and most complex model in the gaming history)... Well, what if CPU is the culprit?

 

To add to this theory, the Supercarrier is not your common DCS module. Everything related to the deck crew is encrypted. It's hundreds of files, including animations, sounds, models and even textures. I can't recall any other DCS module built like this. And what if CPU is simply struggling to decrypt all this stuff, or something like this?

Edited August 11, 2020 by Minsky

[S-GERAT]

Here's a wild guess.

What if this is not GPU-related at all?

 

You see, before the upgrade I've had GTX960. 25 fps, 100% GPU usage, 50% CPU usage. So, probably my GPU was the limiting factor, right?

 

Now I have RTX2060. ~30 fps, 50% GPU usage, 50% CPU usage. A bit more frames, but now I'm bottlenecked somewhere else.

 

And since DCS can't properly utilize more than two cores, and it doesn't need all of my GPU horsepower (because, let's be honest, Supercarrier is not the prettiest and most complex model in the gaming history)... Well, what if CPU is the culprit?

 

To add to this theory, the Supercarrier is not your common DCS module. Everything related to the deck crew is encrypted. It's hundreds of files, including animations, sounds, models and even textures. I can't recall any other DCS module built like this. And what if CPU is simply struggling to decrypt all this stuff, or something like this?

 

The problem are the shadows, these are calculated by the CPU and with only using one core there is a big bottleneck when there are a lot of objects casting shadows at sight. The only solution right now is or disable shadows or use it at low or with an edited shadows.lua with reduced shadow distance. You can check it with the DCS frame counter, it shows what in the scene is taking more time to render, and the shadows are the thing that most impact in the fps.

[...]

No, I still think he's talking nonsense.

 

But your findings are really helpful, thanks.

 

 

If it's indeed a common DCS issue, then I guess the Supercarrier just made it more obvious and easier to trace. Since everybody is using the same boat, one of two planes, and it's usually far enough from the shore and irrelevant objects.

 

 

You are also losing your tone and forgetting that this is a space for different opinions. Mine is worth as much as yours and you have no right to call it "nonsense"

 

 

 

 

We have a descendant of Einstein here to tell us that turning off shadows and removing deck gear improves FPS. What would we do without you? Anyway, wait for the (Russian) developers to fix your problem, we are not as smart as you. :megalol::megalol::megalol:

Edited August 11, 2020 by La Unión | Atazar

[Minsky]

The problem are the shadows, these are calculated by the CPU and with only using one core there is a big bottleneck when there are a lot of objects casting shadows at sight. The only solution right now is or disable shadows or use it at low or with an edited shadows.lua with reduced shadow distance. You can check it with the DCS frame counter, it shows what in the scene is taking more time to render, and the shadows are the thing that most impact in the fps.

 

Maybe. Maybe not. It could be shadows, but not just any shadows, but shadows from the deck crew (that also has to be decrypted by the CPU; theoretically - only once on first load, practically - we don't know).

 

As mentioned in the first post, you can partially resolve this issue by removing some of the animated people, and fully resolve it by removing the deck crew completely.

 

At the same time, placing dozens of other shadow-casting objects on the old Stennis won't bring your system to its knees. You will only see an expectable drop in frames.

 

Improper shadow casting (e.g. one object casting shadow multiple times) is a known DCS issue. Maybe the deck crew is just particularly sensitive to this bug?

Edited August 11, 2020 by Minsky

[Minsky]

You are also losing your tone and forgetting that this is a space for different opinions. Mine is worth as much as yours and you have no right to call it "nonsense"

 

In my opinion your arguments in this thread is nonsense.

And, unlike you, I'm not attacking others personally:

 

We have a descendant of Einstein here to tell us that turning off shadows and removing deck gear improves FPS. What would we do without you? Anyway, wait for the (Russian) developers to fix your problem, we are not as smart as you. :megalol::megalol::megalol:

 

I'm going to ignore your further posts. Have a nice day.

[...]

In my opinion your arguments in this thread is nonsense.

And, unlike you, I'm not attacking others personally:

 

 

 

I'm going to ignore your further posts. Have a nice day.

 

 

You said that 20 posts ago. You said you didn't want to argue and you have argued. I hope this time you will listen to yourself.

 

 

You have published nonsense, one after the other.

[TOViper]

...You have published nonsense, one after the other.

 

*LOL*

I really hope you don't mean Minsky's carefully collected data, and his huge efforts in testing all these things, and furthermore posting all this in a manner someone can understand and follow.

You can fight this s*** all day long, and type stupid things in your keyboard, but currently this leads to nowhere and is for nothing.

 

Please feed this thread with useful information, or something our beloved Einstein would be excited about ... :smilewink:

[...]

*LOL*

I really hope you don't mean Minsky's carefully collected data, and his huge efforts in testing all these things, and furthermore posting all this in a manner someone can understand and follow.

You can fight this s*** all day long, and type stupid things in your keyboard, but currently this leads to nowhere and is for nothing.

 

Please feed this thread with useful information, or something our beloved Einstein would be excited about ... :smilewink:

 

Huge efforts to prove what? How do you improve the FPS by eliminating the deck crew and turning off the shadows? Right from the start, comparing the performance of the Supercarrier with the performance of the Stennis has shown that no one can beat your way of making a fool of yourself.

 

I am not here to please you. It is not my fault that this matter takes away your sleep and you do not accept that there are people who do not see the problem that does not let you sleep. All that I have contributed, opinions and examples, are completely valid, however, you are here fighting against mathematics and anyone who does not see it as you.

 

Opinion number 1 The Supercarrier has high-quality modeling that justifies its high graphic demand.

 

Opinion number 2 That watching the island reduce the FPS is not an exclusive problem of the Supercarrier, it is something that also happens in Nellis, with other planes and also in other simulators.

 

Opinion number 3 Custom settings offer a worse quality / performance balance than presets.

 

Opinion number 4 That removing deck personnel and disabling shadows improves performance, is not something that adds much to the subject, because it is obvious.

 

If this seems stupid to you, it is because you have entered an infinite loop from which no one will be able to get you out ... Eagle Dynamics will manage to improve performance, but we all know that miracles do not exist and that the Supercarrier cannot match the FPS of the Stennis, pure mathematics.

 

 

About your attempts to hold Eagle Dynamics responsible for the Hornet SA page problem, caused by an NVIDIA driver, we'll talk another day if you want.

Edited August 12, 2020 by La Unión | Atazar

[statrekmike]

Huge efforts to prove what? How do you improve the FPS by eliminating the deck crew and turning off the shadows? Right from the start, comparing the performance of the Supercarrier with the performance of the Stennis has shown that no one can beat your way of making a fool of yourself.

 

I am not here to please you. It is not my fault that this matter takes away your sleep and you do not accept that there are people who do not see the problem that does not let you sleep. All that I have contributed, opinions and examples, are completely valid, however, you are here fighting against mathematics and anyone who does not see it as you.

 

Opinion number 1 The Supercarrier has high-quality modeling that justifies its high graphic demand.

 

Opinion number 2 That watching the island reduce the FPS is not an exclusive problem of the Supercarrier, it is something that also happens in Nellis, with other planes and also in other simulators.

 

Opinion number 3 Custom settings offer a worse quality / performance balance than presets.

 

Opinion number 4 That removing deck personnel and disabling shadows improves performance, is not something that adds much to the subject, because it is obvious.

 

If this seems stupid to you, it is because you have entered an infinite loop from which no one will be able to get you out ... Eagle Dynamics will manage to improve performance, but we all know that miracles do not exist and that the Supercarrier cannot match the FPS of the Stennis, pure mathematics.

 

 

Nobody has ever, ever said that they expect the Supercarrier to perform as well as the Stennis, that is a strawman argument of your own creation.

 

What we want is to know why we are seeing a massive loss of performance when there isn't a obvious visual reason for it. We want ED to do some tests and figure out if there are some things that can be done to improve the situation.

[...]

Nobody has ever, ever said that they expect the Supercarrier to perform as well as the Stennis, that is a strawman argument of your own creation.

 

What we want is to know why we are seeing a massive loss of performance when there isn't a obvious visual reason for it. We want ED to do some tests and figure out if there are some things that can be done to improve the situation.

 

 

In reference to one of your previous posts.

 

 

As I have said often in this thread, I am running a system that exceeds the recommended requirements for "high settijngs" as listed for the Supercarrier module itself and I see a 50% drop in overall framerates. I go from a nearly rock solid 60 FPS for everything else in the sim to 30 FPS on a empty Supercarrier deck. It doesn't make sense. There is something going on beyond "You need a better computer" or something to that effect.

 

 

I repeat it. You get rock-solid 60 FPS on the rest of the map, in any situation, because DCS maps are made up of low-resolution textures. I have already attached an example, tree texture and buildings so that you understand that they are textures that do not consume too many resources.

[Minsky]

Nobody has ever, ever said that they expect the Supercarrier to perform as well as the Stennis, that is a strawman argument of your own creation.

 

Save your breath. It's like talking to a wall. To a very thick one.

 

He'll just keep repeating this stuff about textures and polygons, ignoring the fact that it has nothing to do with textures or polygons.

Edited August 12, 2020 by Minsky

[...]

Save your breath. It's like talking to a wall. To a very thick one.

 

He'll just keep repeating this stuff about textures and polygons, ignoring the fact that it has nothing to do with textures or polygons.

 

 

Didn't you say you were going to ignore my messages? How can you have credibility if you don't have self-control?

[Lange_666]

and it's usually far enough from the shore and irrelevant objects.

Now this is not entirely true. If you look again my F-18C screenshots here, you see that in either direction there is nothing to be seen but water (except for a little island in the front view), object count for them is the same but still there is a big difference in FPS.

And an object is an object, irrelevant or not, it always counts. I first noticed this way back when the A-10C was released at Batumi airport. Looking left, high FPS, looking right FPS took a big dip. But on the right view, the object count shot way up because there is a city on that side. However, that city is not visible and therefore should have been irrelevant, still the dip was there.

In my F-18C screenshots, the object count is the same, still the dip is there.

 

 

On a side note: gentlemen, no need to get in a word fight, there are bigger problems in the world then arguing over an FPS dip in some game. Keep it polite!

[...]

On a side note: gentlemen, no need to get in a word fight, there are bigger problems in the world then arguing over an FPS dip in some game. Keep it polite!

 

I will always respond proportionally to people who derail and disrespect.

Edited August 12, 2020 by La Unión | Atazar