This is just for personal expierience and may vary with other people. Also, this is total cpu usage, not just on one virtual cpu.

OS - Windows 2003 Server

16 Player CoD - 4%
16 Player CoD: UO - 10%
32 Player BF1942 - 15%
32 Player BF:V - 20%
16 Player Mohaa - 4%
16 Player Mohaa Breakthrough - 4%
16 Player Mohaa Spearhead - 16%
16 Player Wolfenstein: Enemy Territory - 7%
16 Player UT2004 - 16-18%
16 Player C & C Renegade - 15%
16 player CS 1.6: 6%
16 player CS:CZ: 7%

If I'm wrong, or you have other server CPU usage, post it here.

Are these statistics for full servers or just your average usage?

full

16 player UT2K4 Onslaught: 18-25%
16 player CS 1.6: 5-8%
16 player CS:CZ: 6-9% (Slightly higher than CS 1.6)

Bonus stat: Empty 16 player CS:CZ with 6 bots: 0-1% background CPU usage.

Ok, I edited.

Our UT2004 seem to confilct hmmmm.... can other people post their cpu for ut2004?

You didn't state whether you benchmarked Onslaught or a different gametype for UT2004, most likely it was Onslaught though right?

Yeah, that's why I specified it. If I were you, I'd mesh ours together as 16-25%... UT seems to fluxulate widely, and the various game types use different amounts... I think you were posting averages, and I was posting the kinds of ranges I see when the server is full. On average I would say a full 16 player onslaught is about 16-18%, with some random spikes to a full 25%.

alright

Ours runs a constant 20-25% and spikes to 40%

UT2004 is a Hog!!

on what type of server?

You edited spearhead, not UT2k4 ;p

Now you forgot a % sign after Spearhead.

LOL, just kidding with you.

lol oops

anyone got any other numbers to share?

could anyone possibly give some statistics for CS 1.6 and W:ET on the super gamerz P4 3Ghz ? thx

*stickied*

Good info here! If you can post what OS you're using as well.

I think this would be an excellent thread that would solve alot of questions if it were more general than just xeons.

Something like:

Operating System:
CPU:
RAM:

Game: (% of cpu and mem usage w/ # of players)

Etc.

People need to post that stuff then, I only deal with dual xeons

Good work Allplay.

But.. I would advise using these figures as "take it to the bank" This would be more of a baseline.

You have to understand alot of this depends on how the machine is setup. And the skill level of the operator. Your machine is only as good as who is driving.


Two identical configurations will run differently. Just read a few threads here and you will see some swear they can run xxxx servers on their machines while others can only run xxxx amounts.. Why do you think this is?

does anyone have tips on optimizing Win 2003 server for game server hosting ?

I mean if there's some secret tricks I'd like to know what they are lol.

Besides running your game server's games with a decent server.cfg file that is.

On a full 16 player CS 1.6 running Linux 9.0 with the 2.8 server.

We run a constant 15-20% cpu load. Spikes to 30-40% on map change only for a few seconds that back down.

With that in mind, we are getting a total control Dual 2.8 for our CS customers.

Just wanna share my exspience with BF1942, DC mod.

I run a 64 player server with 56k cleint BW setting.
Server is 2.4 P4, runs about 90% with server full. About 35Mhz a player of used CPU. Server uses about 100MB of ram. And with 56k bandwith setting uses 3Mbit total.

Win 2003
TC2800 Dual 2.8Xeon with 1gig of ram Dual 80 gig drives.

16 player Battlefield Vietnam (Full) 0-3% cpu ~88mb of ram
(128k connection setting)

20 player CS (Full) 2-6% cpu ~90mb of ram
I have a 24 player but have only seen 16 at a time in it yet.But with my BFV and 20 player full plus the 24 player sitting at 14 players I'm running at 10-14% .This is on a TC p2800.My server has only been up a week so all my community has yet to have moved to the new servers.But I will tell anyone looking into buying a server to host clans servers or just to run a bunch of servers pay the extra cash and go with a Total Control or a Super Xeon you will not regret your decision....

BeNt
(BTW the TC servers do take longer than 24 hours to setup.)

Red Hat Enterprise
Dual xeon 2.4
1.5 GB Ram

6 CS Servers running

21/70 Players in all servers combined (+6 HLTV's == 27 Players)

15 Players are in ping-booster servers (NOT INCLUDING HLTV)

uses 37~40% of the CPU :shock:

just got the tc2800 for our clan and am running some stuff for another clan and bringin more on board

win2k
1gig

1 24 cs deathmatch 69.93.114.118
1 30 cs pub with cz and regular maps 69.93.114.117
1 24 cs warcraft3FT with saved xp on mysql 69.93.114.116
2 12 cs scrim servers
2 hltv
1 24 Natural Selection PUB 69.93.114.126
1 12 Natural Selection Priv

stats pages for each server
php, perl installed

gonna start hosting some clan sites
biggest cpu spike ive seen is 10%
but they arent all in use yet

I run a few different servers, though on one red hat enterprise we run, well, the number of games running on this one box may surprise you.

We have a dual Xeon, 2 gigs of Ram, and we're running 17 servers with very different player limits.

Call of Duty, RTCW, etc (no battlefield 1942, etc on these--those are resource hogs period). Some "servers" are pubs, some are private. 17 in total, with no real performance hits to the game. I think we've hit 70% across 2 xeons (we monitor this closely--70% is cummulative across two processors and adding up the 2 pseudo processors from hyperthreading).

Before anyone cries foul, I do play in clans and I have hopped from different game servers we run (i love it when these guys say, well, let's use our server and it's running on the same box cause I run it..lol). No problems. We did shut stuff down, no kernel recompiles (we're lazy) but that many running on one dual xeon running 2 gigs of ram.

Again, it depends on the game you are running. Also, and this is one point no one seems to mention, Server Matrix has some very, very good routing. That is very important. The number of game servers is one thing. Your connection is important as well. Keep that in mind.

I not only host those 17, I play on them as well during peak times.

Would it be possible to make calculations for each game where there would be a formula depending on the # of players?

Or maybe just have some more players (i.e. 32, 48, 64) something like that?

xeon 2.8ghz

ut2004
16 man onslaught ranges from 60%-90% depending on the map occassionaly spike to 95% or above, keep in mind this is one of the 4 virtual CPUs. and the unreal engine does not utilize dual processors

Stop all services and programs you dont need. I dont really do the optimizing of my server the co-owner does it but he spends about 5-6 hours when we get a new server just tweaking settings.

well I found a guide of a list of services that weren't actually needed for Win 2003 server on the net somewhere.
I'm not sure what other "tweaking" I could do though.

Try looking at blackvipers website at http://www.blackviper.com/WIN2K/servicecfg.htm it for windows 2000 but may share some light upon your 2k3 box.

BeNt

OS: Debian Stable, fully patched, upgraded by source to 2.6.4
CPU: Dual 2.4ghz Xeons, Hyperthreading enabled. (4 'cpus')
RAM: 1024megs

Servers:
UT2K4 Onslaught, 16 players. 200megs of RAM, 75+% CPU while full. (% of one processor)

6x HLDS servers (4x CS, NS, DoD), 16 players, 100-ish RAM, 25-30% CPU while full (% of one processor)

Everything full, the CPU load would hover around 60%

yeah that's a good guide. That is the one I looked at before.
It probably can be useful to others here as well.

Lol Err...

On my err Celeron 2.4 I have 2 ET game servers running... Please note this is temp until i have enough to get 1 or 2 dual xeons

With no players on i get around 0-3% for both servers

When the 16 slot is full it uses 38% cpu
When the 22 slot is full it uses 49% cpu

But odley enough combined it never really peaks over 65% with 512 ram
on a Windows 2003 server, Just run a couple of Memory and Processor Optimisers and it works quiet a bit better

PS. please do not reply with how mad i am running these on my Celeron 2.4 It holds out well

Celeron 2.4
512Mb
Redhat 9.0

Battlefield 1942 - 40 players - BBO - 80% CPU. Never paid attention to the memory usage, but I can tell you that Vanilla BF ran very, very stable, even at 80%

Battlefield Vietnam - 32 players - BBO - 90% CPU. Once again, never paid attention to memory usage, but 32 players ran stable as long as it wasn't custom maps.

We also run Apache, MySQL and TeamSpeak from this same server. Apache/MySQL get's very limited use. Only used it for stats, but TeamSpeak will have 10+ at any given time. Any other services have been disabled unless absolutely neccessary.

I have been very suprised by this server. We deffinately pushed it beyond the limits of what people said it could do.

Recently, we played BFV custom maps and I can tell you that anything over 20 players was lagging too bad to play. The CPU Usage would spike to almost 100%. Not many people play custom maps, but I thought it was worth noting.

Just got UT2004 32 player Onslaught on a 2.4 with 512 Ram.
Running win 2003.

It averages around 90% usage (alot, I know) and peaks at 100%. Probably not great for gaming, but it works. I think I may need to tone it down a bit and maybe think of breaking out for a dual Xeon.

I just set it to 16 player onslaught, and its running at about 22% (full) peaking to around 48%
I havent been able to test vietnam on full yet, or bf1942.

I got the superserver version.

We are looking to upgrade to the Dual Xeon 2.4 but have a few questions


Right now we have the Super Gamerz 3.0GHZ
Running win2k and 1.5gigs of ram

Games we currently run:
These are when servers are full all servers spike to 50% when loading a new map
16 Player RvS - 30-40%
28 Player CoD – HQ 15-25%
20 Player CoD – TDM 10-20%
24 Player CoD – CNQ 10-20%
30 Player BF:V 30-40%
16 Player UT2k4 – Onslaught 18 -25%

We also have TS and MYSQL running in the background

Now my question is how will a dual system run these games better then a single 3.0GHZ?

I understand that they will share the load, but doesn’t the OS only use about 60% of the 2nd CPU? And I know the Xeons have more cache..

We just don’t want to upgrade and get either the same performance or worse, it needs to be a nice increase..

AHH thank you all, EXCELLENT thread! VERY VERY Helpful!

what are "kernel recompiles"?

Also, what do you use to monitor performance? I find that when ever I login into our machine (I manage mine for my clan) every one cries that it causes lag so I avoid logging in, but I want to see the performance meters... Do you monitor your performance remotely? Do you log the performance somehow then check it after the fact or something?

check logs, shut down anything that doesn't absolutely have to be running (good security strat as well). Make sure people aren't uploading things to the server during peak usage, etc.

we check logs, play on the servers during peak usage, and monitor usage using different tools during peak hours.

Kernel recompiles refers to Linux. We have just shut down unnecessary services and anything not needed to run a game server. A prime example on a linux box functioning as a strict game server would be things like apache and sendmail. Shut them off if you don't use them.

It's not a Dual Xeon, but to throw into the pot for the heck of it:

RHEL3, SuperCeleron 2.4 (512MB ram), UT2K4, 10 players, Invasion: 50%, spikes to 75%

Sometimes it drops to 20% for a while, I'd guess that's between waves. Most of the time it seems to average 50%.

Tribes: Vengeance MP Demo is out for those who haven't been pressing F5 for the past day.

Anyways, just wanted to give some numbers:
[fta] Phoenity Unlagged (Central) 32-slots FULL
CPU Usage: 25-27% (Dual 2.8 xeon, win2k3 server)
Memory Usage: 350MB
Bandwidth Usage (approx): 950kbit/s out, 500kbit/s in, ~1.5Mbit/s total

The code has been cleaned up and optimized a bit over the beta as you can see.

Win 2003
TC2800 Dual 2.8Xeon with 1gig of ram Dual 80 gig drives

How many battlefield 32 player do you think can run on this server?

Are there any problems when the maps change?

Any tweaking needed to operate multiple game server's ?

I was reading your post concerning your CPU statistics on the dual Xeon server. I am running Windows 2003 server with 2gig of memory and when it comes to Spearhead servers they use CPU resources like nothing I have ever seen. I run quite a few game server as well as game types and the CPU usage information I am listing here is on a 24-player Spearhead server that is heavily moded. When the server is empty it trucks along fine at about 4% on average however when they are full each server hangs at about 18 and spikes as high as 24% as for the other server that you have listed I run at about the same average percentages BF1942 Desert Combat is a little higher. I run a total of 7- Spearhead, 1- Breakthrough, 1- Call of Duty, & 2- BF1942 server. 1 is a RTR, the other is a Desert Combat and believe it or not it runs fine the majority of the time and with good pings, I average about 45 to 55 on my pings from Cincinnati. Most of the time not all the servers a rocking with players but on weekends I can fill up any combination of 4 to 5 servers at once. I am going to be dropping several of the Spearhead servers and adding BF Vietnam, Pacific Assault & Call of Duty United Offensive. I am hoping to be down to only running 3 -of the Spearhead resource hogs by the 2nd week of November. I have a question about BF1942 Desert Combat and Fire Daemon but I will put that in a separate post.

I need more power Scotty

how did you tweak your win 2003 server?

When it comes to tweaking the server, you need to remember that I use this box exclusively for game servers. So with that said, I very carefully looked at every service that is running on the server by default and any service that I did not need I disabled. If you plan on doing this make sure you are extremely careful not to disable a service that will affect any service or program that is critical to another service or process or any service that is critical to Windows running correctly so that you’re OS does not become unstable. I also setup a 4gig swap file on my secondary hard drive to help out how Windows utilizes virtual memory. I am running all of this on a box with Dual 2.4 Xeon processors, 2048MB Ram, 2X200Gig (RAID1) hard drives, 4048MB page file on "C' & 4048MB page file on "D" drive and as I said earlier I have only the services running that need to be running so that the operating system will perform correctly and not become unstable.

How did you set up , the swap?

How often do you reboot?

I have always been told that servers running Windows operating systems should be rebooted once weekly. I also recommend that my clients reboot their game servers every 72 hours.

could any one give me the cpu usage of a UT99 server

and would i be able to run

4- 16 player UT99 servers
2 -16 player UT2k4 servers ons

on a dual xeon server with 1 Gig mem

or what could i run on this type of server

I don't have the exact usage yet, but I was able to run a 32-Player BF1942 at almost full capacity (Trying to get it there ) on my P4 2.8ghz/HT with 1Gig of ram with pings <80MS and no server lag.

I'm really suprised how well it can run the 32-Players, and once I fill to capacity going to boost up from 32 >> 46 and see how that goes.