General Discussion (dev branch)

[dale0404 5]

;2676208']The problem is that people have seen way larger numbers of AI work fine in OFP' date=' A1, A2 and OA - both FPS wise and behavior wise.

What people criticize is not the state of things alone, but the [b']massive degradation[/b] in various areas over time and sadly A3 has reached another low point here.

People play the series not to have small engagements. The scale of terrain, AI+player numbers and gameplay made this series unique.

The sad part is that A3 has not put emphasis on the key areas people care about but others of little relevance for the most part.

Hi mate,

Gonna disagree with you here, having small engagements can be just as intense as having large ones. Both small and large engagements can be just as good. And please do not assume that all players think the same as you :D

[Alpha-Kilo 36]

I think this discussion about the performance of the game clearly shows the different concepts of the devs and the users. BI says: We give you a lot of toys (and advertises the game as feature-rich and stable) but don't overuse them or accept the responsibility for taxing your systems. The players say: If I have the toys and own a better-than-avarage computer I want to use them at least to the same extent as I was able to in the previous iterations of the game. Bad Benson's posting #11825 says this very well and the other posts on this thread provide many practical examples and explanations.

It is particularly interesting that the community can come up with many solutions to most of the problems (ALive, weapon resting, female characters, useful tactical glasses to name just a few) Unfortunately running many addons together is complicated and may lead to more problems which in turn makes it difficult to find servers which accept them. But despite the shortcomings these addons show that it is still possible to enhance the game. Shouldn't professionals who know the engine inside-out be able to achieve even more than hobbyists no matter how enthusiastic they are?

My own personal way of handling the situation which probably isn't for everyone is quite simple: I think the majority of the players prefers COOP but I see the AI mainly as a training tool. I play SP-missions which are carefully designed to mitigate the problems and my online gaming is mostly TvT maybe with some AI added as accessories and for decoration.

I am glad that the devs are following this thread and it is good to have pettka's comments about the ongoing effort. I say there is still hope and we should always remind ourselves not to take the game too seriously.

Edited April 27, 2014 by Alpha-Kilo

[vegeta897 13]

Hi mate,

Gonna disagree with you here, having small engagements can be just as intense as having large ones. Both small and large engagements can be just as good. And please do not assume that all players think the same as you :D

Agreed. The idea that huge maps are only a benefit for large engagements is just ridiculous. Engagement size has very little to do with the map size. It's not like you could "run out of room" (without tens of thousands of units).

One of my favorite missions to play with my friends is a small TvT in which we have to capture a randomly chosen target building. Each team spawns in a random location on the map, without GPS, and must land nav to find out where they are and how to get to the target. So, we're talking about engagements that usually only involve 1v1, yet the map size is what makes the mission fantastic and so open-ended.

[Furret 0]

Just to make it clear, we are working on the optimization and it is an ongoing task (as may be seen in the changelogs and all the BROREPs). I have just clearly stated that using resources depends on mission makers too, our missions are optimized even in this way. But throwing hundreds of AIs to some frenetic urban battle and expecting no performance hit would seem funny to every game developer :icon_twisted:

Everyone can appreciate that AIs are performance hungry, I think the problem people have with it is they don't understand why they see their FPS drop due to AI (CPU)calculations, people see AI and graphics/FPS as separate entities.

[.kju 3245]

Dale its not about everyone. Of course a good junk of people also like 5on5 pvp, 5-10 humans vs AI or SP/campaigns. That is not the point.

---

To the above people: You need to take more time to actually read more closely. Of course it is not only about size. Please read the list again.

pettka even confirmed/reinforced the notion that path finding uses most/a lot of CPU cycles.

Hence more complex terrain, less FPS. Complex terrain means number of objects, the density of objects per area,

the precision is set to move and analyze the terrain, the amount of hiding positions, the number of building positions,

the number of objects to consider for AI vision, etc, etc, etc.

So to make a specific example - whats the point of making all buildings enter-able and very complex, when even AI is hardly

capable at all to make use of it (aside from put there by the mission designers, via scripting or player commanded).

Even for players fighting within buildings is mostly a clumsy, non rewarding and even frustrating experience.

Open world combat is where the series shine - hence micro terrain, more elaborate mechanics of cover and to interact with vegetation,

objects and the related gameplay features would have been the area to expand where the series is strong.

[ProfTournesol 956]

;2676361']Open world combat is where the series shine - hence micro terrain' date=' more elaborate mechanics of cover and to interact with vegetation,

objects and the related gameplay features would have been the area to expand where the series is strong.[/quote']

That sums it up perfectly.

[bad benson 1733]

;2676208']The problem is that people have seen way larger numbers of AI work fine in OFP' date=' A1, A2 and OA - both FPS wise and behavior wise.

What people criticize is not the state of things alone, but the [b']massive degradation[/b] in various areas over time and sadly A3 has reached another low point here.

People play the series not to have small engagements. The scale of terrain, AI+player numbers and gameplay made this series unique.

The sad part is that A3 has not put emphasis on the key areas people care about but others of little relevance for the most part.

....

;2676247']The real answer in terms of terrains would have been:

1) Make 3 mid size terrains like in OFP instead of one huge.

2) Reduce complexity in terrain where its just a gimmick and put emphasis on gameplay elements.

3) Reduce complexity in terms of visuals to free up resources from both GPU and CPU.

4) Not look into diving and underwater as its really a non element - if you want to have a focus on water' date=' it should have been above water and for vehicles and not infantry.

5) To me the whole replicate a real location as the prime focus is not helpful. Interesting, diverse gameplay and good performance should been have the key driving factors - to use real world data is just a means to achieve that.[/quote']

truth. all of it.

also. terrain size is a relative term. sure Altis without objects would run smooth but with size comes more object density in one rendered frame since when you see the ocean on Stratis already you will see two more cities on Altis because the terrain size is what allows so many objects to be in your scene. same goes for trees or pretty much anything. not to mention that cities and towns are too big for the engine to handle on Altis.

and to interact with vegetation,

tree hugging? ;)

Edited April 27, 2014 by Bad Benson

[Bouben 3]

;2676208']

People play the series not to have small engagements.

Don't speak for me' date=' mate, I enjoy small engagements very much.

---------- Post added at 21:41 ---------- Previous post was at 21:36 ----------

;2676247']The real answer in terms of terrains would have been:

1) Make 3 mid size terrains like in OFP instead of one huge.

2) Reduce complexity in terrain where its just a gimmick and put emphasis on gameplay elements.

3) Reduce complexity in terms of visuals to free up resources from both GPU and CPU.

4) Not look into diving and underwater as its really a non element - if you want to have a focus on water' date=' it should have been above water and for vehicles and not infantry.

5) To me the whole replicate a real location as the prime focus is not helpful. Interesting, diverse gameplay and good performance should been have the key driving factors - to use real world data is just a means to achieve that.[/quote']

1) tolerable

2) don't agree

3) can agree but only to some degree

4) agree

5) agree but I love believable locations

Moreover, the terrain does not have such a huge performance impact (with rational visibility distance setting) so there is no need for your points. The AI would lag the game even on the desert island from OFP.

Also, I agree with you that there is no need for every building to be enterable. Only the strategic ones. The rest of the buildings should be placeable via the editor so that the mission maker can make what he/she wants.

Edited April 27, 2014 by Bouben

[.kju 3245]

Bouben sorry to say but you don't seem to get it. It is not about visuals or GPU use.

More complexity means more CPU and RAM use, and more demanding AI calculations.

Most community terrains have considerable better performance compared to BI terrains - this was even true for OA/A2/A1.

While some of them had comparable terrain complex and object density.

I think better FPS, thus more enjoyable play, more intelligent AI behavior and room for higher AI or player numbers is preferable.

[mistyronin 1181]

If I remember properly some BI said dev in this forums that there is also morale involved in the AI management, so if the morale is really low the AI just run away.

You could check past posts ( if I'm not wrong it was here in the dev forum, probably in the AI thread ).

Edited April 27, 2014 by MistyRonin

[Bouben 3]

;2676665']Bouben sorry to say but you don't seem to get it. It is not about visuals or GPU use.

More complexity means more CPU and RAM use' date=' and more demanding AI calculations.

Most community terrains have considerable better performance compared to BI terrains - this was even true for OA/A2/A1.

While some of them had comparable terrain complex and object density.

I think better FPS, thus more enjoyable play, more intelligent AI behavior and room for higher AI or player numbers is preferable.[/quote']

I get it and I don't want to reduce terrain complexity so that AI has less to calculate. I believe AI code clean-up and catching would be a better solution.

Also, I play on a notebook and performance is good. So I don't know what missions you guys play that you have performance problems. Isn't it also a mission design problem - not effective use of AI? Lots of scripts?

Edited April 27, 2014 by Bouben

[Bouben 3]

If I remember properly some BI said dev in this forums that there is also morale involved in the AI management, so if the morale is really low the AI just run away.

You could check past posts ( if I'm not wrong it was here in the dev forum, probably in the AI thread ).

I believe this is not the case. Repro-missions proves that.

[Alwarren 2767]

So I don't know what missions you guys play that you have performance problems.

I had a mission located around Oreocastro with just around 100 AI, no scripts or anything, that caused horrendous performance problems. Why? Because most of the AI was in the town. That shouldn't do anything, since it was multiplayer and the bad performance was on the client, but there you have it.

Any mission with large-scale engagement needs a supercomputer from space in order to work. Denying that doesn't help.

[bad benson 1733]

I believe AI code clean-up and catching would be a better solution.

you say that like there's just a switch for that in the engine. it's not that simple i'm afraid.

i agree though that arma 3's graphical fidelity in itself (stuff like improved lighting and shaders) is not really the problem. i think that having a huge town that is enterable just has no gameplay benefit if it can never be used in really large engagements. on the contrary most urban PvP missions i played degrade to a clunky hide and seek with a few players in big ghost towns. i'd prefer some nice points of interest instead and a lot of not so complex cover that the AI can actually use as opposed to buildings.

speaking of cover. since houses and walls already have predefined cover points, i always ask myself why BI even go the route of calculating all this on the fly when it would make much more sense to prepare the environments in a way that AI merely have to interpret paths and hook points. a lot of other games do the same in their classic "map editors". it would be extra work but nothing compared to a total rewrite of the AI or an AI that works great and runs great (not quite there yet :p) for that matter. it seems to me that it's too much of a code only solution.

they already did this with building positions for ages now which then got used by modders to make AI actually search houses. i don't see why this can't be done on a bigger scale. you could even make AI react according to the terrain better as a terrain maker.

at least some more helpers could lift some weight of the "on the fly stuff" and make it more effective as a result. i'd be willing to put the extra work as a modder into custom buildings or my map if it means that AI can use it better with less performance cost.

[k0rd 3]

;2676247']The real answer in terms of terrains would have been:

4) Not look into diving and underwater as its really a non element - if you want to have a focus on water' date=' it should have been above water and for vehicles and not infantry.

[/quote']

This is a really, really, really good point that I almost missed skimming through the thread. I thought I'd quote it again for emphasis :)

[Bouben 3]

Mines used to be a serious threat in former versions but right now they are completely useless. They are unable to kill a soldier even in multiple quantities. A soldier will simply heal and continue his journey.

Mines should be deadly again, because in Arma, there is no way of simulating serious injuries they cause in real world.

I did a test with multiple squads running through a dense minefield with APERS mines and none of the soldiers was killed anywhere on the field. Not a single one.

[lexx 1392]

Just tried it. Placed a few APERS mines, had someone walk to it and boom, died in the first blast. AT mines are pretty strong too. A single one was enough to fully take out my Kamysh.

/Edit: Bounding mines are even worse. Two exploded and 3/4 of my squad was dead. Even the ones that looked to be in a save enough distance.

Edited April 28, 2014 by Lexx

[eymerich 11]

you say that like there's just a switch for that in the engine. it's not that simple i'm afraid.

i agree though that arma 3's graphical fidelity in itself (stuff like improved lighting and shaders) is not really the problem. i think that having a huge town that is enterable just has no gameplay benefit if it can never be used in really large engagements. on the contrary most urban PvP missions i played degrade to a clunky hide and seek with a few players in big ghost towns. i'd prefer some nice points of interest instead and a lot of not so complex cover that the AI can actually use as opposed to buildings.

speaking of cover. since houses and walls already have predefined cover points, i always ask myself why BI even go the route of calculating all this on the fly when it would make much more sense to prepare the environments in a way that AI merely have to interpret paths and hook points. a lot of other games do the same in their classic "map editors". it would be extra work but nothing compared to a total rewrite of the AI or an AI that works great and runs great (not quite there yet :p) for that matter. it seems to me that it's too much of a code only solution.

they already did this with building positions for ages now which then got used by modders to make AI actually search houses. i don't see why this can't be done on a bigger scale. you could even make AI react according to the terrain better as a terrain maker.

at least some more helpers could lift some weight of the "on the fly stuff" and make it more effective as a result. i'd be willing to put the extra work as a modder into custom buildings or my map if it means that AI can use it better with less performance cost.

If I haven't misread your opinion, if this mean better code with the same open building I'll agree with you; otherwise, i disagree.

To be honest I really likevery much the open building in Arma3 even if this is most probably due to my sp experience.

I would prefer work on the code instead than removing the open building.

[jgbtl292 0]

when arma use only 40 % from 4 cores than is a better multithreading better , than remove object's form the island. the most Pc have the power but arma 3 use it not ^^

[Variable 322]

How about getting rid of the automatic "Return to formation" message for units that are far from the leader? It was introduced in Arma 2 and It's really annoying, especially in coop. You send a player to do something away from you and you need to hear "yourself" telling it to get back. it's pretty senseless. As a workaround, we just give the unit a WP or tell it to stop, but that can cause players to unintentionally disembark from vehicles. Can we find a single player that thinks that this feature has any value at all?

[nomisum 129]

How about getting rid of the automatic "Return to formation" message for units that are far from the leader? [...] Can we find a single player that thinks that this feature has any value at all?

things like this shouldnt ever be automatic. neither SP nor MP

[fennecus 10]

Might be difficulty settings related.

[Greenfist 1863]

Just tried them too and found that bouncing mines are nearly 100% deadly up to 10m and very likely to kill even at 25-30m. That's the width of the 8 man line formation, so once it wiped out the whole team. Sounds almost overpowered.

Normal APERS is far less powerful and usually kills just the person who triggered it. And because they detonate on the ground there's a much bigger change some of their power is deflected by bumps on the terrain, trees, rocks or other people.

[oldy41 61]

To be honest I really likevery much the open building in Arma3 even if this is most probably due to my sp experience.

I would prefer work on the code instead than removing the open building.

Yes and no:

I am also primarily interested in SP, and really appreciate the open buildings. In A2 all those locked houses were a bit immersion breaking.

BUT on the other hand I mostly avoid entering the buildings because to me it always feels a bit like cheating on the AI. Though it is true that AI now enter buildings occasionally (at least with AI mods) they still can not cope with them reasonably well. :(

[Ezcoo 47]

There's a difficulty setting that enhances the protection of infantry, don't remember what it's called (unable to check it now). Check your difficulty settings and make sure that the enhanced protection is disabled.