Jump to content
Sign in to follow this  
redarmy

better perf, big groups/small groups?

Recommended Posts

So making a scenario,and there are alot of ai,iv had to cut back on some.

One thing i cant find on the performance that someone who played arma2 or prior to that might know...

Is it better to have a few large groups of infantry,or many small groups?

How would the cpu cope with the difference?

Share this post


Link to post
Share on other sites

In tests I have ran.

I have found no difference between group sizes and/or individual AI as far as the server is concerned.

Just the total amount.

I did find that the amount of AI directly affects the clients fps.

Even if the client cannot see the AI.

Share this post


Link to post
Share on other sites

As far as I have been playing with the editor, AI works in groups.

So a lot of soldiers in big groups is more efficient than the same amount in small groups.

The other thing is the graphic impact, than then obviously as more stuff you have, more heavy is against the hardware.

Share this post


Link to post
Share on other sites
As far as I have been playing with the editor, AI works in groups.

So a lot of soldiers in big groups is more efficient than the same amount in small groups.

The other thing is the graphic impact, than then obviously as more stuff you have, more heavy is against the hardware.

Think your rite mistyRonin,

Was testing again with 10 small groups,then with 5 large groups(same amount infantry). Giving them waypoints clearly showed that 5 WPs to follow was better than 10 on the cpu.Ill base it on that and keep my groups large from now on.

The only draw back is the combat mode bounding,the more units following the leader the longer it will take to reach destination.Though i guess,this is ArmA,not COD ;) be patient and absorb the beauty of the battlefield

Share this post


Link to post
Share on other sites

I've done some work in this area on populated servers ...

The code essentially spawned a load of AI and assigned them waypoints immediately.

On populated server with lots of AI, here were results:

With 12 groups of 10 units, there was only about 20% compliance to the waypoints. Server would desync and some groups would not follow the waypoints.

When I reduced the number of groups from 12 to 6, and shifted units accordingly (20 per group), there was closer to 100% compliance to issued waypoints.

My analysis is that less groups (more units per group) yields greater performance, noticeable when the CPU is already stretched thin.

Share this post


Link to post
Share on other sites

Guys,we do need to post any and all even if small results/tests for ANYTHING that can increase fps in this game.BIS may never address the cpu issue so always post your findings.

Possibl a thread should be started for "little things" to increase performance.

Share this post


Link to post
Share on other sites

Big groups are more efficient but "dumber". Smaller groups are better able to seek cover and move more efficiently through built up areas than big ones. So which is better for you depends on what you are looking for. I personally like small platoon sized firefight kind of missions so my groups are usually the size of a fireteam. But if I were making an all out war kind of mission I would make groups squad size or more to save on performance/reliability.

Share this post


Link to post
Share on other sites

We use a group size of 6 for the AI.

With a cache system in place for all dynamically created None AO AI population. (generally around 150-350 AI, containing 2-6 vehicles + trip mines and IEDs + roadblocks + Static defensive camps and garrisoned+patroling groups of AI inf)

Then none cached AI for the AO. (generally around 90-100 AI containing 2-6 vehicles + static defence camps + reinforcement spotter scripted AI + Mortar spotter scripted AI + static mortar camps)

The AO has a cap of 150.

The dynamic AI have a cap of 350.

15 players with around 300-350 AI active on the server, It gave results of 49fps and 47cps with drops to 35fps and 34cps when large fights occurred.

Normally after 4-6 days of playing the server performance degrades enough that we restart the mission.

even though performance could be seen ingame to degrade server fps and cps are still around the >45fps with cps being 5 less than fps.

the performance degrade we believe is because of the amount of destroyed map objects after 6 days, resulting in AI being created to slow along with desync chains and rubber banding.

Although it could well be something mission related we are trying to figure this bit out atm.

Because of the AI behaviour Patrol scripts we use UPS and edited UPS.

We found that smaller groups gave better fights.

We use no editor placed AI or waypoints at all.

Our server creates 750 ai using the AIserverbenchmark mission on default params.

TLDR :- 15 players and 350 active AI using group size 6 gave us a server performance of 49fps and 47cps.

Share this post


Link to post
Share on other sites

350 active AI in MP with 15 players at 49fps?

Does that mean 350 cached AI or full sized groups on map?

Share this post


Link to post
Share on other sites
350 active AI in MP with 15 players at 49fps?

Does that mean 350 cached AI or full sized groups on map?

Full groups on map. (350 AIL in Freds ASM) split into various types of groups sizes.

We don't use the reduce squad to SL type cache system.

We create locations at mission start with AI amounts for each location and spawn/de-spawn AI based on player distance/time deactivated from location.

Depending on the amount of activity on the server though the mission will start to degrade after 4-6 days of persistence.

We think its a mixture of looping scripts and Map object damage over time.

I should say though its rare that we hit 350+ AI on map its normally around 150-250.

Edited by BL1P

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  

×