Is my server underpowered?

[DarkShade]

Hey all, I figured this would be the place to ask my question as it directly pertains to hardware management and Source-based games.

First off, the specifications:
2x Opteron HE8346 Quad Core (8 total cores) @ 1.8GHz
8GB ECC Registered DDR2 Kingston
ASUS KFN4-DRE Motherboard
Linux Ubuntu 8.04 Server

It runs my webserver, low-end games like Quake 2/3 and multiple (tested up to 7 at once) instances of Left4Dead beautifully. Never had a problem.

Recently I've had a growing population on my TF2 server, and it seems TF2 is extremely taxing on my system. Just by having a map loaded and one person in the server, the CPU% is up around 50%, and tonight I saw around 99% consumption of one core when there were nearly 20 people in the server. What gives? Is TF2 really that much more demanding than Left4Dead? What can I do (if anything?) to mitigate how much power TF2 takes up?

[loopyman]

No, this doesn't make sense... a TF2 Server with one player should use no more than 5% of one core of that processor... Make sure something else isn't going on. srcds.exe is not multi threaded so it will only use one core of the 8 cores available. The processor that you're using still isnt very impressive, but if you can run 7 l4d servers with no problem, then you should be able to run 1 tf2 server just fine.

Also, how are you checking CPU usage? IF there is 20 people in the server and there is no lag you should be fine, if you're using the rcon stats to see CPU usage, it isnt very accurate at all.

[BehaartesEtwas]

cpu usage isn't very accurate for game servers in any case. depending on your hardware and your kernel settings you might observe like 4 different values for the cpu usage at the same time depending on how you measure it ;-) don't rely on it, look at the server fps. if they are stable there should not be any problems. if not, look at the server configuration. a TF2 server will take more than a l4d server (more players, higher tick fps), but not that much more...
on the other hand, AMD cpus are known not to be optimal for srcds-based game servers...

[DarkShade]

I'm using a combination of 'stats' from rcon and htop in Linux. 'Stats' reports similar usages as htop does, and I've played in the server while it's been packed and there are noticeable jitters.

Just opened up another TF2 server on another core just to see what's going on. The new server gets about 250FPS with one player in it

Quote:CPU In Out Uptime Users FPS Players
15.50 3837.14 6714.91 11 0 283.05 2

Meanwhile, the other server is chugging along:

Quote:CPU In Out Uptime Users FPS Players
99.50 74213.26 92201.76 169 3 11.71 21

BIG difference