Good point, I was up late and had a hairbrained idea. Feel free to close KK

Tonic's VVS is pretty reliable for exploding vehicles, especially if you remove his attempted fix for it: http://www.armaholic.com/page.php?id=23020

It also seems like AI groups stall, even with WPs to sectors. Check sectorstalled mission and notice after 2 sectors captured, AI groups just stand around.

Derp.. thanks! Please close

I love you.

Possibly related: Commands do not work on AI units in vehicle (such as enableSimulation/hideObject/disableAI)
http://feedback.arma3.com/view.php?id=21702

Issue in this ticket may of been "Driver" discovered by Wolfenswan:

• if(_x != leader _this && {!(isPlayer _x)} && {!("Driver" in assignedVehicleRole _x)}) then { //doesn't work, driver is cached

+ if(_x != leader _this && {!(isPlayer _x)} && {!("driver" in assignedVehicleRole _x)}) then { //works, driver isn't cached

13-05-2014
EXE rev. 124294
Size: ~83 MB
http://goo.gl/4KAfxB
...
"Minor PhysX optimization"
...
"ZBE 144 Groups, 1145 Units, 17 NoCacheFPS, 51 CachedFPS, 59 DeletedFPS"

Test system: i5-2500K @ 4.8GHZ/Nvidia GTX 670

1.18 Stable Branch (5/12/2014)
"ZBE 144 Groups, 1145 Units, 15 NoCacheFPS, 48 CachedFPS, 60 DeletedFPS"

12-05-2014
No new EXE
Size: ~199 MB
http://goo.gl/EN8ciF
"ZBE 144 Groups, 1145 Units, 13 NoCacheFPS, 47 CachedFPS, 60 DeletedFPS"

Will update from time to time with new dev branch builds.

Repo mission added. I could simplify however I have a busy week ahead. Fank may look at it later.

I was speaking more of the BIS FpsManager.fsm which Highhead found in the modules.pbo and modified for general use:
https://www.dropbox.com/s/wikzxjn40sk9lzf/fpsManager.fsm

Your FSM would be very interesting to test, thank you!

Will add you to Steam!

Didn't mean to Fireball, I hit Edit to fix some wording issues in which you may of edited before I hit save, feel free to modify.

This ticket was written while I was at work and in a rush, apologize for the abuse of the english language ;)

Previous ticket (before confirming Physx,Sounds slowdown/cause)
http://feedback.arma3.com/view.php?id=18451

YOU GUYS ROCK!!

13-05-2014
EXE rev. 124294
Size: ~83 MB
http://goo.gl/4KAfxB
...
"Minor PhysX optimization"
...

FPS before spawning "C_Offroad_01_F": 59
FPS after spawning "C_Offroad_01_F": 59

I then went though and SPAMMED the script and spawned a few thousand vehicles in (random class names):
59 FPS

Shameless plug of AI "caching" which is somewhat related to this issue but may be a different beast entirely: http://feedback.arma3.com/view.php?id=18734

As for THIS specific issue, it seems resolved with only a few FPS drop which is understandable and a MASSIVE improvement. It's a shame that how this issue was fixed cannot be disclosed (for curiosity sake) but it is resolved nonetheless.

Update/patch 1.18

Issue seems to be resolved. FPS only drops a few which is understandable.

Was confused with enablesimulationgloabl vs enablesimulation. enablesimulationglobal only works from SERVER (dedicated) machine, not from editor.

Anyway, enablesimulation before patch 1.18 was dropping FPS and as of now FPS only drops a few frames which is a massive improvement.

Bravo BIS for fixing this issue so quickly! Please keep ticket open while Fank and others can duplicate results to make sure issue resolved.

Very cool, good work everyone!

This also seems to happen with Units as well!
http://forums.bistudio.com/showthread.php?176389-enablesimulation-false-and-hideobject-true-what-else-is-simulating&p=2670115&viewfull=1#post2670115

For example change in spawnpos.sqf:

_civcarpool = [
"C_SUV_01_F"
];

_indcarpool = [
"C_SUV_01_F"
];

_milcarpool = [
"C_SUV_01_F"

One could modify the calls to be all _civcarpool instead:
[_civcarpool,_industrial,true,"ColorYellow"] call _car_spawn;
[_civcarpool,_military,true,"ColorRed"] call _car_spawn;
[_civcarpool,_civ,true,"ColorGreen"] call _car_spawn;

And then just plop classname in:
_civcarpool = [
"C_SUV_01_F"
];

Not a very specific repro script I admit however it works to duplicate the issue.

Machine info:
2500K @ 4.8GHZ
8GB RAM
Nvidia GTX 670 (OC)
2x Samsung 840 EVO 250GB in RAID0

I've done some experimenting with this addon: http://www.armaholic.com/page.php?id=24517 (for easy modification of config.bin)

FPS no vehicles: 53
FPS spawned vehicles (stock config.bin): 27
FPS stripped config sounds: 34

7 FPS gain by stripping class sounds from the config.bin which would mean that sounds for vehicles are not disabled when enablesimulation false. Don't know if sounds are handled by an eventhandlers system or otherwise. Will be a good place to start.

Will update as I experiment more.

Nice, very much appreciated Frank!

Hmm.. Starting at 56FPS
C_Boat_Civil_01_police_F only drops FPS to 53
Box_NATO_Wps_F doesn't drop FPS = 56
B_MBT_01_arty_F drops FPS to 27!
B_Boat_Transport_01_F doesn't drop FPS = 56
C_Offroad_01_F drops FPS to 28!
B_Heli_Light_01_F drps fps to 53

This is using the same repo mission, spawn script.

I don't think this is Physx, maybe so but could it be the event handlers for vehicle sound? Maybe it has to do with Physx complexity?

Why not disable Physx and sound eventhandlers (or whatever is used) when enablesimulation false?

Forum topic
http://forums.bistudio.com/showthread.php?176389-enablesimulation-false-and-hideobject-true-what-else-is-simulating

Additionally Manalog doesn't work with A3 (https://dev.withsix.com/issues/27636)

I assume that A3 also doesn't have diag_captureSlowFrame, diag_captureFrame,diag_logSlowFrame? RPT reports (missing ;) tried on stable and dev, command(s) listed in A3 wiki page https://community.bistudio.com/wiki/Category:Scripting_Commands_Arma_3

Would be nice if the community could submit helpful dumps/info, but such may be easier for internal BIS teams to do (to control environment).

A workaround would be to spawn/delete vehicles during mission runtime (based if player near) but would instead impact MP and netcode from packets spammed during vehicle create/delete..

Would love an update :D

Since we have shiny new HC built into the server and it works fine, this ticket can and should be closed.

Related: http://feedback.arma3.com/view.php?id=14814

Was not fixed in http://feedback.arma3.com/view.php?id=4662

It's pretty much undocumented and unsupported. Take a good look at the BIS mission using the module. Notice the triggers and the units synced.

I simply extracted the missions pbo in the Arma3/addons folder using eliteness. It's a great way to get insight on how BIS does missions. BIS loves using FSMs so be sure to download a FSM editor.

Annnd. this is how you use it. Uploaded to ticket and
https://www.dropbox.com/s/gevs4x2f4aefki9/MP_COOP_m04.Stratis.zip

Still waiting for the HC DLC (Would be good solution)

@Xeno
Point taken. That has been an issue for the longest time (https://dev-heaven.net/issues/6963).

Although with A2 BIS has made the final statement on the issue (https://dev-heaven.net/issues/6963#note-28) maybe things will change for A3.

I feel as if such dialog were to be opened again, a separate ticket would need to be created as this ticket focuses on a stop-gap or addition to your solution.

Link updated for AI pref suggestions in general and native HC Fix:
https://drive.google.com/file/d/0B3VdHnk7BX1BUGtMRmMyRDRza1U/edit?usp=sharing

I have created a whitepaper of sorts that goes beyond the issues currently plaguing HC that could solve the issue of AI performance within the core game engine. It has fancy graphics and doesn't drone on and on. Most importantly it offers a solution that should be accepted by both parties as it resolves all concerns.

https://www.dropbox.com/s/y5ourdvidcxw6zf/Arma3ServerWorldspaces.pdf

My original (apparently unrelated) note stated:
/*

1. Add a switch to the dedi server binary -client or create a new dedicated client that is cdkeyless or universal (as the dedi server operates in this fashion)

2. Only allow the HC/Dedicated client to connect to localhost (127.0.0.1) to nullify security issues.

As issues exist with remote HC (https://dev-heaven.net/issues/62500) limiting HC to localhost only shouldn't have a negative impact.

Sure, the admin may hack their own server via this method using an "unbannable client" but such easier ways as proving grounds locked to UID exist already.
*/

Irony as forceheadlessclient=1 made it into Dev the same time this did (http://feedback.arma3.com/view.php?id=4662).

*EDIT by Dwarden*
Dwarden has suggested buying another copy should be necessary for HC w/o compromising security layers.

You can use SteamCMD to run multiple times per machine,
as long as you use login to STEAM account which owns arma 3 alpha
each new SteamCMD session of course needs new unique account (with arma 3 ownership)

rest is unrelated ...

[9:39:42 AM] David Foltyn (Dwarden): ye but the Dev branch has the OA fixes
[9:40:00 AM] David Foltyn: it will take some time before the OA merges will be in stable branch (maybe next week)
[9:40:10 AM] David Foltyn: there is way too many changes

"has the OA fixes"

The evidence of forceheadlessclient=1 working in editor and partially in MP leads me to believe that the functionality of the OA code merge has been completed in this case, the fact that the code hasn't been massaged into A3 working code should of been the point of your statement.

My reported information stands correct as the "code" has indeed been "merged" into the dev A3 tree for this specific function; reported information may help further dev conversion of code.

Fireball, I take it as you being helpful in providing formation but the perception is that I should stop reporting on this issue. I take it as the code has been merged into Dev and there is an issue that it should be reported on.

Please close as forceheadlessclient=1 was put in Dev (along with cd key checking that breaks HC completely http://feedback.arma3.com/view.php?id=7082)

Still broken with the (oa beta patch) merge. forceheadlessclient=1 doesn't get deleted by editor BUT on server the HC slot is unjoinable (nothing can join HC slot including HC) and I get this:

Received 130, expected bool
Unexpected message data (message struct NetworkMessagePlayerRole, item leader)
Before (0x00000000): 11 00 00 00 00 35 ff ff ff ff 0f b0 ed 8f 0e be f6 dd 8d 06 00 00
Current (0x00000016): 82

Also my old school (manually move HC into slot or don't execute) method is broke:

if !(player == HCSlot) exitwith {};

Error position: <HCSlot) exitwith {};
Error Undefined variable in expression: hcslot

File mpmissions\__cur_mp.Stratis\scripts\Init_UPSMON.sqf, line 27

I got a nice leatherman here that I've used to cut bailing wire, no engineer required!

Didn't know it was that bad until I started researching to see if a "ZBE_Unstuck" module would be worth the effort

Video: http://youtu.be/dcBAPpKxaNw

It seems it would be worth it.

Its revolves around object instancing and physx. Decrease your object draw distance and notice the improvement (500m or so).. High polycount models with physx simulation shouldn't be rendered out to 1k if you aren't going to separate the physx rendering into it's own worldspace. We are so spoiled by Arma 2 but the fact is our CPUs simply cannot keep up with the frame preprocessing with the level of detail Arma 3 provides. At the very least, object LOD by distance is up for work.

I personally get 50-60FPS with 60-99% GPU load w/ a gtx 670 & 2500K @ 4.8GHZ @ 600M object draw distance with all else on ultra. However, the second you start putting in more and more AI or player models; the situation rapidly worsens.