Why DCS can't support large scale warfare

[bkthunder]

Testing a mission with about 600 ground units and 8 human aircraft.

Well, in the very little time we've been able to fly without the game crashing, we experienced HUGE FPS DROPS whenever MLRS rockets were being shot.

 

Now this is interesting:

 

-the ground troops were about 30nm away, behind a mountain and completely out of sight.

 

-All of the 8 players (3 of us were flying F-15s at high altitude, 3 of us on the ground ramp starting, 1 ka-50 in flight, 1 playing Combined Arms) experienced FPS drops to about 10FPS whenever the MLRS were shooting.

 

- we all have mid to high end systems, normally flying at 40, 50 or 60 FPS.

 

 

Unfortunately, as it stands, we'll never be able to simulate warfare beyond a few dozen troops.

 

 

Thanks

[Kuky]

Too much CPU power required (lot of things being calculated even if its not visually visible), hence the FPS drop.

[bkthunder]

Of course it's using too much CPU, that's why I said it will not support large scale warfare, which is a shame for a modern sim. Trying to calculate every single event happening in the map is clearly the wrong thing to do.

We should all fire up the good old 1998 Falcon 4.0 then...Bubble logic anyone? That works great.

[falcon_120]

I really think something of that kind should be implemented,the falcon bubble is not perfect,

and have a lot problems,but maybe is better than not having the possibility of playing a large scale war at all. Things like lets down to 5-10 fps everytime a MLRS fires or a CBU is dropped are a pity.But it's the way it is right now, I think that ED is trying hard to get a better simulation with every iteration,we just need to wait a bit more.

[msalama]

...is clearly the wrong thing to do.

 

Not necessarily if/when you can simultaneously utilize, say, 4 processor cores.

[bkthunder]

Not necessarily if/when you can simultaneously utilize, say, 4 processor cores.

 

....maybe we'll see that by the end of....2030 :doh:

[BTTW-DratsaB]

Arnt the FPS drops caused by MLRS and bombs just bugs that need to be fixed?

[Rusty_M]

As MLRS and CBU's are currently a known issue, perhaps it would be worthwhile trying out a similar scenario that avoids using that type of weapon. It could at least give us a handle on how many units can be used.

[maturin]

ArmA can simulate hundreds of AI with vastly more fidelity than DCS' basket cases. They need to find some way to scale things. When a convoy is doing nothing but driving along a straight road for 30mins, the CPU load should be equivalent to a game of minesweeper.

 

But what was your FPS when the MLRS wasn't firing? If it was fine, then I retract what I said, and ffs use howitzers instead.

[Exorcet]

I haven't tried yet with DCS, but FC2 could handle quite a bit of stuff, and my computer was not high end.

 

I think the real issue here is MLRS and CBU's. Other than that, it would probably be fine. Either way I'll found out. I'm planning on making a A-10C mission that will put the A-10 in its original role, a plane for taking out a full scale Russian tank invasion.

[almonds]

maybe it will improve with the nevada map?

[PFunk1606688187]

The problem is probably in large part because of a lack of multi threading. Individual core speeds haven't increased all that much in the last few years, its the increase in the number of cores in average CPUs thats been the real improvement, yet if the game can't use more than 1 of them then its basically operating on 5 year old tech.

 

Its a bottleneck that until its solved will permanently hamper people's ability to push the mission making end of things.

 

Also, I would imagine a lack of dedicated server support increases the client load, and revelations about differentiation between two clients' apparently impact point on something like WP rockets might indicate that there isn't much of an efficient way for clients to actually share information about whats happening in the game world.

 

There are lots of things that this game needs to fix if we're ever going to outshine Falcon's dynamic campaign.

[Silver_Dragon]

There are more important issues to worry about in a dynamic campaign at the moment. And not return to the multicore, and rivers of ink have been on this.

[c0ff]

Taking current obvious and hidden bugs of DCS aside, the question of 'large scale warfare' is far from being simple.

 

Note that there's NO complex large scale games in the market at all.

The simulated world of MMO RPG's is far from a complex modern warfare scenario.

BF3 - 64 units max on a small map.

 

The problem is that the complexity of the task is so large, that it's impossible to efficiently solve in a generic manner. All existing solutions are based on a tremendous trade-offs.

And these solutions don't scale. We've hit this with Combined Arms.

 

About multicore: efficient use of it requires complete rearchitecturing and rewriting of the code. As in: throw everything to trash can and start from scratch.

So in reality it's done one step at a time.

[msalama]

....maybe we'll see that by the end of....2030 :doh:

Maybe :)

 

But the sim still runs pretty well for me even in its current state. The most crowded mission I have has some 250 ground units and several flyables, and I'm seeing a framerate of approx. 45-60 f/s in the cockpit without any stutters at all - and the only thing actually bringing my computer to its knees ATM are the MRLS launches (and those, as we know, are broken at the moment). So a relatively happy camper here at least, although I've admittedly had to tweak the game a fair bit to achieve all this. But still :)

[PFunk1606688187]

Note that there's NO complex large scale games in the market at all.

The simulated world of MMO RPG's is far from a complex modern warfare scenario.

BF3 - 64 units max on a small map.

 

Actually I think for the most part the proliferation of console gaming has hindered the progressive development of mainstream large scale FPSs.

 

Even so, Project Reality mod for BF2 has a largely stable and playable 126~ player server code. Arma has seen some tests by communities where they dumped the AI load for a massive 100+ person COOP onto a separate server. And lets not forget EVE which features basically the largest PVP space combat ever conceived in a game world that live and breaths and propagates based on the insanely massive combined efforts of players, soon to be directly linked into an FPS game which will show combat on worlds in the EVE universe.

 

The potential is there, its just that the market dictates mostly that the way to make money is to cater to the short attention span XBL gamer who thrives in the 24 player max server size.

 

Planetside has ambitions of hosting 6000 player servers which feature 3 continents and 3 teams fighting a persistent battle.

 

 

There may not be a current massive ground warfare game out, but the potential and the ambitions appear to be materializing.

[bkthunder]

I'll sound repetitive BUT Falcon 4 has had a dynamic, large scale warfare since 1998. Yes, it has limitations and it's had some issues, but all in all the thing just works.

BMS is the primary choice of many if not most serious online squadrons because it's just easy to grab a plane and go fighting the Korea war. No weeks spent creating a single mission, no stutters, no pain.

 

Apart from the dynamic campaign, how is it that an old sim can handle thousands of AI troops and aircraft 24/7, and DCS can't run a script mission with a couple hundred troops? Can't DCS scale effects with distance? Does it use LODs?

 

Ok, there's a bug with CBUs and MLRS, but FPS drop just looking in the direction of troops / aircraft that are 60nm away and completely out of sight, it even stutters when some ka-50 enter the game in a far away FARP....

[sorcer3r]

without a game save feature such large scale scenarios wouldnt make any sense imo.

 

If we had already such a feature we could already host "medium scale" scenarios and save their progress and then combined arms would be really very very interesting. :-D

[sobek]

Apart from the dynamic campaign, how is it that an old sim can handle thousands of AI troops and aircraft 24/7, and DCS can't run a script mission with a couple hundred troops? Can't DCS scale effects with distance? Does it use LODs?

 

Of course it uses LODs. This is not about graphics, it is simply about things that you can't turn off if you wan't the sim to stay realistic, even if nobody is watching.

 

Falcon aggregates units that are not inside a player bubble, that means that those units do not exist in the 3d world, instead they exist as entities on a 2d map where the outcome of any ancounter is based on statistics. Saves a lot of resources but also negates things like terrain playing an important part in the outcome of a battle.

 

It was a) a matter of design choice for ED to NOT use such an approach because it means cutting corners and b) is next to impossible to retrofit now, so please leave Falcon out of this. Yes, you can build much larger wars right now with Falcon, but to do so it uses compromises that should not so easily be put aside.

[Puddlemonkey]

It depends what you really mean by large scale warfare. I would think that the important step forward that ED could make is to be able to have a large-scale scenario and for smaller missions to affect that scenario. Hopefully online dynamic campaign functionality will achieve this.

 

I imagine a situation where you have a server which offers up a large environment with troops of all sorts throughout it. People could then launch and enter individual missions which impact that large environment. Those missions could be hosted by other servers that communicate back to the main one. You could compare it to the SimCity regions and cities approach.

 

I think that being practical, we would not have the computing power to simulate a whole battlefield unless the processing is delegated across many different severs / clients.

[sobek]

I imagine a situation where you have a server which offers up a large environment with troops of all sorts throughout it. People could then launch and enter individual missions which impact that large environment. Those missions could be hosted by other servers that communicate back to the main one. You could compare it to the SimCity regions and cities approach.

 

How would you handle the edges of one mission? Would you simply negate the fact that units from one region could interact with units from another?

[c0ff]

I don't want to sound like a broken record, but it's doubtful there's a market for this.

We are speaking about complexity >BF3 with budget <<BF3, which was ~$50M on marketing alone, and most probably the same amount for development.

 

People in general want stories and immersion, not 'realism'.

With 'realistic warfare' we are talking about niche product which will never have the budget of a blockbuster. And there's no technological silver bullet as of now.

[Cyb0rg]

Yeah, still things are getting better - step by step. We would just like to be faster.

[bkthunder]

Falcon aggregates units that are not inside a player bubble, that means that those units do not exist in the 3d world, instead they exist as entities on a 2d map where the outcome of any ancounter is based on statistics. Saves a lot of resources but also negates things like terrain playing an important part in the outcome of a battle.

 

Brilliant!

 

It was a) a matter of design choice for ED to NOT use such an approach because it means cutting corners and b) is next to impossible to retrofit now, so please leave Falcon out of this. Yes, you can build much larger wars right now with Falcon, but to do so it uses compromises that should not so easily be put aside.

 

Yes, that is clear. Some would say that "something is better than nothing"... Anyways, this thread was just to say that at this moment in time, DCS is not a warfare sim, and that became really apparent when we tried to simulate a bigger scale battle. We'll find another way to have fun.

 

...After all, it's called digital COMBAT simulator...One day we'll have computers powerful enough to accurately simulate every bullet shot hundreds of miles away, then ED might make a Digital Warfare Simulator :joystick:

[BRooDJeRo]

Shouldnt the CPU load go down when you use options in AI on/off and use bubble activation techniques with zones/movingzones in combination with triggers when thinking of full mapsize missions or with alot of units to spread the load.

A unit very far away out of sight can simply be on pauze or activated to do its thing when any player comes close or mission relative movement is needed.

Edited November 15, 2012 by BRooDJeRo