Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
Hello everyone! I've been pulling my hair out, followed the guide here to the letter: http://wiki.fragaholics.de/index.php/EN:Linux_Kernel_Optimization
Have idlers running on all 4 cores (intel q6600), taskset them and the hlds instance, tried HPET and TSC. No matter what with just 2 players on the server the fps will dip to 500 or less. Is there some voodoo magic i'm missing that isn't listed on the page above? Thanks for any help!
Posts: 1,033
Threads: 13
Joined: Oct 2008
Reputation:
11
What is the startup command?
Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
pengy Wrote:What is the startup command?
Tried all sorts of things, but the gist of it is: ./hlds_run -game cstrike +map de_dust2 +sys_ticrate 1500 -tickrate 100 -pingboost 3
Tried raising/lowering sys_ticrate as well. Also, i forgot to mention this is on debian lenny, 64bit. Thank you for the reply.
Posts: 1,033
Threads: 13
Joined: Oct 2008
Reputation:
11
Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
pengy Wrote:Did you chrt it?
Sure did, makes no difference. I have tried EVERYTHING on that page, along with everything i could find digging around on these forums/google. I've tried different hardware as well, been banging my head for weeks on this one :/ Thanks again for the reply!
Posts: 1,033
Threads: 13
Joined: Oct 2008
Reputation:
11
OK, try the 2.6.28 kernel, with the same settings, except the preemption has to be low/server.
Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
pengy Wrote:OK, try the 2.6.28 kernel, with the same settings, except the preemption has to be low/server.
Alright i'll give it a shot, thanks.
Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
bleomycin Wrote:pengy Wrote:OK, try the 2.6.28 kernel, with the same settings, except the preemption has to be low/server.
Alright i'll give it a shot, thanks.
Just to confirm, does it matter which kernel in the 2.6.28 series? Is 2.6.28.1 alright?
Posts: 1,033
Threads: 13
Joined: Oct 2008
Reputation:
11
Don't know, it might work
Posts: 226
Threads: 2
Joined: Aug 2009
Reputation:
1
If I remember correctly pingboost 1/2 use timers that are based on jiffies, you should try using pingboost 3
Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
08-19-2009, 03:42 PM
(This post was last modified: 08-19-2009, 03:42 PM by bleomycin.)
pengy Wrote:Don't know, it might work
Alright, with 2.6.28.1 running with no preemption (server), and 4 people on the server the fps were mostly stable, but if you hammer rcon there are dips into the 780's at times. Should i try setting the preemption to low latency desktop? These are much better results than with the 2.6.26.8-rt16 kernel though, i'm excited!
Posts: 1,033
Threads: 13
Joined: Oct 2008
Reputation:
11
make sure that the sys_ticrate is 1001 or something like that, also chrt the server process, hmm, it might be low latency desktop
Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
08-19-2009, 05:20 PM
(This post was last modified: 08-19-2009, 05:24 PM by bleomycin.)
pengy Wrote:make sure that the sys_ticrate is 1001 or something like that, also chrt the server process, hmm, it might be low latency desktop
Alright, set sys_ticrate to 1001, i have been chrt'ing it all along. With low latency desktop fps are completely stable +-4fps or so which is fantastic! However, with only 4 players cpu usage is hovering at around 55%!?
*edit*
Monk: Sorry, i missed your reply. I have been using pingboost 3 all along as well.
Posts: 1,033
Threads: 13
Joined: Oct 2008
Reputation:
11
Yes, thats the only problem with it, if you dont chrt it , it drops down to like 10% usage, and fps goes down to about 800 or something
With chrt http://www.fpsmeter.org/p,view;1260.html
w/o chrt http://www.fpsmeter.org/p,view;663.html
Posts: 40
Threads: 7
Joined: Dec 2008
Reputation:
0
08-19-2009, 05:32 PM
(This post was last modified: 08-19-2009, 06:04 PM by bleomycin.)
pengy Wrote:Yes, thats the only problem with it, if you dont chrt it , it drops down to like 10% usage, and fps goes down to about 800 or something
With chrt http://www.fpsmeter.org/p,view;1260.html
w/o chrt http://www.fpsmeter.org/p,view;663.html
Hrrrmmm, in some quick testing it only seems to happen when i run it in a screen, chrt by itself doesn't cause any increase in cpu usage. Unfortunately i have to run it in a screen...crap. Has anyone had any success in decreasing the cpu usage while chrt'ing the process running in a screen?
|
![[Image: b_350x20_C002748-004880-FFFFFF-000000.png]](http://cache.www.gametracker.com/server_info/173.58.194.29:27015/b_350x20_C002748-004880-FFFFFF-000000.png)