Jump to content
Sign in to follow this  
dustoff

Arma server quad core

Recommended Posts

Does the arma dedicted server use mulitcores the right way ?

Strange question but running on a quad core kentsfield 3210 . 2.15 ghz quad core machine.

Noticed that running the server is using 25% cpu and does not matter if I set afinty to 1 core or all four. it does not apear that Arma dedicated server is taking advantage of all four cores.

Server itself seems to run smooth as silk just wish the gaming computer was as fast.

Share this post


Link to post
Share on other sites

From what I've experience with my game server (arma dedicated linux server) I have to conclude that arma does only partially utilize a systems smp-capability.

Although libpthread is compiled into the binary the fact that the cpu load arma causes raises very fast (8 players did suffice) up to 100% (that means that one core is at full capacity), but then it grows very, very slowly up to 125% (even with 32 players on it).

in case you use top -H in order to obtain detailed information on arma's threads you will end up with sth like that (I'm afraid i could not get any detailed values for a cpu load as high as 125%):

<table border="0" align="center" width="95%" cellpadding="0" cellspacing="0"><tr><td>Code Sample </td></tr><tr><td id="CODE">PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND

2099 user      20   0  568m 536m  10m R   62 26.7   1464:10 armad_1.14

2100 user      20   0  568m 536m  10m S    0 26.7   0:00.13 armad_1.14

2101 user      20   0  568m 536m  10m S    0 26.7   4:40.70 armad_1.14

2102 user      20   0  568m 536m  10m S    0 26.7   4:10.14 armad_1.14

2103 user      20   0  568m 536m  10m S    0 26.7   0:51.01 armad_1.14

Main issue of multithreading is that it is very difficult to exchange information between threads which are running on two different cores. Additionally some calculations just are not threadable (due to logic reasons). Try to imagine how to thread 1+1.

Share this post


Link to post
Share on other sites
I don't think ARMA uses mulitple cores at all.

No it doesn't use multicore. Our OGN server is instead configured to run several instances of ArmA server, sort of 1 to a core.

Multicore then = multi servers

Share this post


Link to post
Share on other sites
Gnat @ June 24 2008,13:01)]No it doesn't use multicore.

I have to disapprove that. The linux dedicated arma server is smp capable (it caused 125% cpu load on my machine before), however due to smp's drawbacks concerning data exchange and the general issue of threadability (cheap example: how dou you thread 1+1?) it just cannot utilize an smp system fully.

Share this post


Link to post
Share on other sites

I'm a bit perplexed by this as well.  I have been testing a new server (Dual Core E8400 @ 3.8GHz running Windows 2003 Server R2 SP2).  Last night while playing a Warfare map I was monitoring the CPU utilization and the ArmA process utilization.  I would have expected to see a 50% overall utilization of the server under heavy load, but instead recorded an average of 50-55% with peaks at 65%.  This would mean it was utilizing more than one core, though minimally. I never seen it exceed 65% however.

Share this post


Link to post
Share on other sites

Something is offloaded to another thread.

Most likely it's the network communication.

It's been this way since OFP days..

Share this post


Link to post
Share on other sites

I have a Linuxserver with 2 x Xeon 2.6Ghz CPUs and it says 100% CPU usage but it's clearly using only one CPU.

However, the sad thing is that I have it running Evolution with persistent world on and even now, when there's noone on the server, it uses 100% CPU (and 62% mem).

We're usually just two people on there and when attacking a town I get a server FPS of 3-5...

Share this post


Link to post
Share on other sites
Quote[/b] ]We're usually just two people on there and when attacking a town I get a server FPS of 3-5...

Somethings wrong then, That box should be able to run that map on much higher FPS easily

Share this post


Link to post
Share on other sites

I have the same issues running Domination persistent. In the beginning it will idle very low CPU usage but once you get near a town the CPU hits 100% and even with all players disconnected it sits around 95-100% cpu usage.

Its a Linux box w/ 3GB ddr2 memory / core 2 duo 2.6 ghz cpu / 250 GB SATA HD

Share this post


Link to post
Share on other sites
I have a Linuxserver with 2 x Xeon 2.6Ghz CPUs and it says 100% CPU usage but it's clearly using only one CPU.

However, the sad thing is that I have it running Evolution with persistent world on and even now, when there's noone on the server, it uses 100% CPU (and 62% mem).

We're usually just two people on there and when attacking a town I get a server FPS of 3-5...

Beware of the kernel options you're using.

SMP, RT and all that stuff.

Evo is heavy on the CPU because of AI. Don't expect getting anything high in server FPS when running it.

There's no comparison possible between an AI heavy mission and a PvP only mission. That's why you can get way higher in players count with a full PvP mission.

An empty Warfare can use 100% of 1 of my 4 CPUs at some point.

Also, it's very probable the Linux dedicated is less good than Windows dedi for ArmA. Don't forget it's a port from Windows

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  

×