How to increase dedicated server FPS

[[ZSU]Preacher]

I have a Core2Quad 2.83GHZ server with 4GB of RAM running Windows 2008 x64 on a 100Mb internet connection. I can't seem to get more than about 25fps on the server for missions (even with only 1 player connected). Looking at the Windows performance monitor, I can see all 4 cores are utilised, but no core seems to go over 50-60%.

So given that there doesn't seem to be a CPU bottleneck, why can't I get the server's maximum framerate of 50fps (48/48fps in reality)? I've tried every arma2.cfg network setting known to man, but it doesn't make a difference.

Here is the screenshot of a mission with one player not doing anything, and getting 23fps.

[.kju [PvPscene]]

No need to get more FPS.
Only below 10 FPS you get slight more warping etc.
Only below 5 FPS you get real troubles and mission no longer working correctly.

[Rommel]

Even on an blank map with no one connected?

[Taxman]

Post your arma2.cfg buddy... some of the guys here may be able to help with that.

[Game__On]

No need to get more FPS.
Only below 10 FPS you get slight more warping etc.
Only below 5 FPS you get real troubles and mission no longer working correctly.

You need the server to be above 25 fps. Maybe not when all you use is the dmr , but with machineguns your rate of fire drops dramatically, and your groups (when you fire a burst of say 5 or 10 bullets) scatter all over the place, meaning you can't hit sh^t at larger distances (300 to 500 meters)

[Cross]

Many missions have scripts in them running when you start the mission and most of them are buggy/unoptimized so it is common to have 20-25 FPS when you load a mission.
Make a mission yourself with no AI/scripts in it and test with that.

http://cross.kellys-heroes.eu/public/mpsimple.utes.pbo

[Proton]

I am also trying to improve the dedicated server's performance with a complex MP mission. Just right now the mission runs with 554 units and 124 vehicles, usually hovering around 10 FPS. A bit higher would be desirable. The question is what is the most important influencing factor.

I can view the CPU usage of threads in ProcessExplorer. According to that, the server runs 12 threads, of which 4 is taking measurable CPU time. One is fully using a core (25% on my quad machine), and 3 others using half of a core together (7%, 4% and 1%). Probably the one consuming 25% is the bottleneck.

Can a developer please enlighten us that what actually the threads do? Especially the 25% one. I would split the server into 4 main threads: one for general game logic, one for scripting, one for network and one for AI pathfinding. Is this guess true?

The reason I am asking this is to decide how to try optimize performance. If the scripts actually take a separate thread and they are not executed in the 25% one, it makes no sense to reduce the amount of running scripts, as the advantage is going to be zero. The mission uses about the same number of scripts at start and has 40 FPS, so probably not the *amount* of scripts which matters.

If the number of AI units is the most important factor, does it help if I do not move them when not necessary (to reduce pathfinding)?

[rexehuk]

Post your bandwidth settings here.

[Proton]

language="English";
adapter=-1;
3D_Performance=1;
Resolution_W=800;
Resolution_H=600;
Resolution_Bpp=32;
Windowed=0;
MinBandwidth=800000;
MaxBandwidth=25000000;
MaxMsgSend=256;
MaxSizeGuaranteed=512;
MaxSizeNonguaranteed=256;
MinErrorToSend=0.003;
MaxCustomFileSize=100000;

Although I am testing the mission with just a single client at the moment.

[Terox]

what we really need is a server benchmark mission.

Heavy AI useage with some heavy cycling function calls

Bunch of continously firing mg guns
large groups of ai looping waypoints, continously changing formation
made say for 5 players
with benchmark times every say 3 minutes and mission running for 10
Outputting to the RPT
Mission has no player interference, just to be there so that it creates network traffic and therefore every mission replay isd the same, so possibly no enemy to interact with

A mission which basically stresses all elements of the .cfg settings using a set amount of players that a typical small community can gather for optimisation testing


With this we may at last be able to put together a guide for server settings. To date there hasnt really been such a thing and I have found the BIS WIKI entries to be somewhat misleading and most server's .cfg's seem to be nothing like one n other which just creates more confusion. Ideally this is a task for BIS, but that aint gonna happen anytime soon

---------- Post added at 04:01 PM ---------- Previous post was at 03:56 PM ----------

language="English";
adapter=-1;
3D_Performance=1;
Resolution_W=800;
Resolution_H=600;
Resolution_Bpp=32;
Windowed=0;
MinBandwidth=800000;
MaxBandwidth=25000000;
MaxMsgSend=256;
MaxSizeGuaranteed=512;
MaxSizeNonguaranteed=256;
MinErrorToSend=0.003;
MaxCustomFileSize=100000;

Although I am testing the mission with just a single client at the moment.

Thats bloody unusual, they are exactly the same settings we use and I mean exactly. I dont think its as optimised as it could be though