Multiplayer not in playable state

[NeuroFunker 11]

you are on my ignore list. but since you tend to be all over threads that deal with issues of the game, it's hard to ignore you to be honest, since people tend to respond to your nonsensical one sided view. so yea. if you'll try to be less "loud" in threads that interest me then i might be able to respond to you less, if that bothers you.

i only take a look in this thread once in a while to see stuff like Terox posted. you know actually relevant facts and maybe positive progress i might have missed.

i just was on one of the a3rc servers. the coop one, since the PvP ones are either empty or locked (will test there later).

the result was a constant jumping between 15 and 18 FPS which funnily enough the game itself shows as a red number in the video options. while that's better than the 15 FPS max on some servers, i have to say i'm not convinced that your claims are anything but subjective sugar coating. on the hostile take over servers (PvP) i get 30 FPS max in quiet areas and 20 FPS max at the AO. that's not amazing but playable (30 is the first number the game doesn't show as yellow ;))

i agree that the very lite PvP missions tend to have better performance but that's not only obvious but also ignoring the (much) bigger part of arma 3 multiplayer (or any arma for that matter), which is coop or complex PvP missions with RPG elements or base building. if the thread title is too bold for you then maybe try to focus on the things that are actually being dicussed here. the title and OP is only the start of the discussion. and about 20 pages in you are kinda late complaining about it.

i can sign galzohar's post 100%. it's pretty much how i perceive it right now.

thats weird, have played 30 ppl before, had 40-50 at base, and 30 at AO, maybe you entered shortly before restart no idea. And since i'm in your IL, ill ignore you back, so don't bother quoting my posts anymore.

[terox 316]

Can you please stop turning this thread into a battlefield

[bad benson 1733]

yea this whole discussion got me curious and i tried some more servers.

my conclusion is: still not playable on most servers. i even asked people about their max FPS on the hostile takeover server and added"...be honest!" to the question. the answers ranged from 16-32 where i assume that 32 means that person has a really good CPU. i did the same thing some months ago and the answers were the same generally. and i'm talking about PvP here. i then went on trying to have some fun. what can i say. aiming with FPS jumping between 15 and 18 gives me headaches and needless to say you can't hit anything. pretty much unplayable (yes indeed i said it) especially for PvP (we'Re talking max FPS here. not steady).

thats weird, have played 30 ppl before, had 40-50 at base, and 30 at AO, maybe you entered shortly before restart no idea.

i don't care what the reason is. the server you play coop on has pretty much the same FPS as any other server i tested. i can post a screenshot if you want.

sorry about derailing the thread a bit. i'm done here. will check back in a few months. what can i say. i just want to play the game...

[whiztler 135]

We (Nopryl Community, nopryl.no) played a 6-day campaign on Altis. That is 6 Tue nights of gameplay with 23-30 players. We started off using v1.04 and finished using v1.08. BIS supplied us with a dev server version half-way through.

Since this was our first campaign on A3, it was a learning project for the mission maker team as well. We have done many missions and campaigns on OPF, A1 and A2. Thankfully when we started we already knew of server performance issues. To support our A3 game nights, a new server with the latest hardware was bought . Bob was our uncle ... we were good to go ...

Day 01: 135 AI's, 30 Clients, few triggers and 40+ waypoints. (A3 v1.04)

CPS beginning of the mission: 5-7

CPS halfway through the mission: 12-15

AI's were sluggishly hopping around. Their reaction to Bluefor was slightly delayed.

Day 02: 195 AI's, 28 Clients, 15+ Triggers and 40+ waypoints (A3 v1.06)

CPS beginning of the mission: 3-6

CPS halfway through the mission: 9-14

AI's were sluggishly hopping around. Although things improved near the end when most AI's were killed.

Day 03: pre-placed 120 AI's, 27 Clients, few Triggers and 8 waypoints (A3 v1.06)

Spawn on demand: 60 AI's (mostly patrols and reinforcements)

Garbage collector script running. VD: 1500m

CPS beginning of the mission: 9-16

CPS halfway through the mission: 15-19

Quite playable. AI's not moving around sluggishly and responding appropriately

Day 04: pre-placed 47 AI's, 26 Clients, few Triggers and 8 waypoints (A3 v1.06)

Spawn on demand: 130+ AI's (mostly patrols and reinforcements)

Garbage collector script running. VD: 1500m

CPS beginning of the mission: 20+

CPS halfway through the mission: 30+

AI's count at any moment was never higher than 60. Very playable.

Day 05: No data

Day 06: pre-placed 79 AI's, 23 Clients, 10+ Triggers and few waypoints (A3 v1.08)

Spawn on demand: 70+ AI's (mostly patrols and reinforcements)

Garbage collector script running. VD: 2000m

Dynamic Civilian Life script running (spawn on demand civilians)

CPS beginning of the mission: 12+

CPS halfway through the mission: 20+

AI's count at any moment was never higher than 50.

Very playable.

Lessons learned:

- ATM spawn on demand (dynamic spawn) is the way to go.

- Use a garbage collector script.

- Make sure that air units don't trigger the 'spawn on demand' groups all at once.

- Scripted triggers, waypoints, etc seem to execute faster. This is a hunch, not proven/tested.

Future missions:

- HC (is this fully functional?)

- (or) Caching

- All AI's spawn on demand

- No sites modules

[Wolfenswan 1]

We ran our first successful 35+ players session yesterday without any hick-ups and constantly high server (40+) FPS. 1.10 stable seems to be a vast improvement.

Links lead to screenshots of bandwith/FPS graphs.

36 Player adversarial, no AI.

37 Player coop, 186 AI (all present at start), using global caching (hideObject and disabling Simulation).

Same Coop further in (players dropped)

The minimal FPS (grey) are somewhat weird, but might be due to the (un)caching of groups. It's safe to assume that even without caching missions with <100 AI should be playable with those player counts. A headless client does not appear a necessity but might be worth investigating for higher player counts.

Client FPS was not massively affected by caching as there was no Fog present and thus GPU provided the bottleneck for most ppl. However, in single testings before it became clear that even with light fog caching can provide a significant FPS boost (up to 10) as the GPU stops bottlenecking the system and the CPU is relieved.

This is the caching solution I used in the missions. This was the COOP played.

Server CPU:

xeon e3-1245v2 @ 3.4ghz

Server.cfg:

language="English";

Windowed=0;

adapter=-1;

3D_Performance=1;

Resolution_Bpp=32;

MaxMsgSend=256;

MaxSizeGuaranteed=256;

MaxSizeNonguaranteed=256;

MinBandwidth=10485760;

MaxBandwidth=110100480;

MaxCustomFileSize=0;

class sockets

{

maxPacketSize=1400;

};

Edited February 3, 2014 by Wolfenswan

[fred41 42]

@Wolfenswan, very interesting ...

If we compare the FPS graph before and past this low FPS event (second screenshot),

it seems that this event has caused an irreversible state change in the engine/mission.

Do you have an idea, what the massive FPS loss in your second ASM screenshot has caused?

Maybe a firefight or something like this?

[Wolfenswan 1]

I might be mistaken, but I think that is the Coop loading (and everything to the left is idle state, then adversarial).

Bit hard to judge, as ASM doesn't have any timestamps or a scale ( ;) ).

[fred41 42]

I might be mistaken, but I think that is the Coop loading (and everything to the left is idle state, then adversarial).

... hmmm, players, vehicles, etc. seems to be already fully present in the "idle" phase ...

Bit hard to judge, as ASM doesn't have any timestamps or a scale ( ;) ).

... 1 pixel/second, except you changed timedivisor to 2, 5 or 10 sec ...

For more precision, you could use the logfile feature (timestamps per sample).

[big_t 1]

I'm going to chime in with my experience. Overall the FPS seems to be fine in general operations but it really really drops a substantial amount whenever some serious fighting action occurs. (usually while looking down the scope at an enemy). It's frustrating because this is exact time when you really need stable fps. Seems like a cpu bottle neck but a 4 core intel at 4.5ghz, is this still not enough?

Is it ballistic calculations that are slowing everything down? It doesn't seem like it would be a graphical issue unless the graphics cards are handling the immediate action.

My specs below:

3770K @ 4.5

dual 6870 (crossfire)

16gig 7-7-7-21 ram @ 1333