Posts: 3
Threads: 1
Joined: Jul 2011
Reputation:
0
Hello people!
i've installed (uptodate version from hldsupdatetools) css to Centos 5.6, (2xXeon system)
and after running with
-ticrate 66 +fps_max 600
i've got in "rcon stats" 160 fps.
i've tried to change ticrate and fps and pingboost, but it always near 160 fps.
Please help!
======
Also, could you please tell me, srcds using all cores or like hlds only 1,
is it necessary to run srcds with affixment to 1 core or srcds will be using cores normal?
Thanks! 
Posts: 2,031
Threads: 27
Joined: Nov 2008
Reputation:
17
first, there is no option like -ticrate or -tickrate and also no pingboost.
srcs uses only one core, there is no necessity to force it on a specific core (often that makes it worse, but sometimes better).
did you do any optimization of Linux and the kernel? CentOS uses rather old kernels which are not well suited to host game servers. See my signature for a howto to update your kernel and optimize the system.
Posts: 3
Threads: 1
Joined: Jul 2011
Reputation:
0
(07-14-2011, 05:36 PM)BehaartesEtwas Wrote: first, there is no option like -ticrate or -tickrate and also no pingboost.
srcs uses only one core, there is no necessity to force it on a specific core (often that makes it worse, but sometimes better).
did you do any optimization of Linux and the kernel? CentOS uses rather old kernels which are not well suited to host game servers. See my signature for a howto to update your kernel and optimize the system.
1.) So i shouldnt set option -tickrate 66 or -tickrate 33 ?
Only fps_max ?
2) Thank you! Noted!
But though should i force it 1 spec core or just do nothing and run server at all cores by default ?
3) No. Thank you, im gonna do it )
Posts: 2,031
Threads: 27
Joined: Nov 2008
Reputation:
17
1) correct. ideally try to match the fps to the tickrate, i.e. run with 66.67fps. usually that can now be achieved with fps_max 66.7
2) try it out. I would start without forcing it to a core.
Posts: 3
Threads: 1
Joined: Jul 2011
Reputation:
0
I've update the centos core and now fps is near 450 not higher.
By the way
Why i should update core if you advise me to set fps 66.67 ?
May be i should launched with old core in that case?
Also if i run a 2-5 servers. Should i force to cores them or not?
I hope there will not be such a thing like with 1.6 that all servers runned by default on all cores using in fact only first core.
Posts: 2,031
Threads: 27
Joined: Nov 2008
Reputation:
17
you wont easily get above 500 fps, as valve made kind of a limit with the last update.
height is not everything. it is important to reach stable fps, especially if you run with 66.67fps. usually kernels which do not achieve fps close to the "limit" of currently 500fps will also have stronger variations.
make sure your *average* fps is not higher than the tickrate. 66.67fps is good, 66.7fps is already not good. fluctuations may lead to single values above that, that's ok. but it's important to keep the fluctuations down, the difference between 66.67fps and 62.5fps now is the same as the difference between 500fps and 1000fps before the orangebox update (both 1 millisecond)!
Posts: 25
Threads: 8
Joined: Jan 2010
Reputation:
0
I was having some fun with your tool bethaartesetwas,
Just trying it out with centos 5.6
Without it, and default kernel, on xeon i7, 24gb ram, i get around 450-500fps
with your tool i can get it up as high as I set the value, but if I set 30.000fps it gives 5.000.
If i set 5000fps, it gives around 1800fps with fps_max 0. BUT if I set fps_max 1000, I get pefect 999.6-9 fps straight line in your fpsmeter, but consumption is very high if I limit the FPS with fps_max, if I set fps_max 0 the consumption goes very low when empty.
If I set 5000fps and limit fps_max 66.67 I get exacly 66.67 CONSTANT, wich is soposed to be perfect and consumption is very low. <- ATM Im trying this out with people without letting them know, if they ask I don't tell them untill after the game has played. I will talk about results, playing with 66.67 and 999.9.
As a Gamer, I have noticed that 66.67 fps has kind of a better registry BUT when you look at models they don't run very smooth, sometimes when they notice that they say the server is crap but untill they notice they think its great.
So for now I believe FPS aren't that important in their value but what is really important is that they are constant, it doesnt matter if its 10k, 5k, 1k, 500 or 66.67, meanwhile they are PERMANENT.
I will post later my conclusion in a developed post for future references.
Posts: 882
Threads: 43
Joined: Dec 2010
Reputation:
13
If you're talking about the bepingbooster or something, dont use it for a production enviroment. its no good and its only made to test the things of orangebox engine i presume.
Posts: 25
Threads: 8
Joined: Jan 2010
Reputation:
0
I know that, I use it just for testing and so, also because I cant get my servers fps higher than 450 without it :s
In a few hours im hoping to get some testing done with bepingbooster with people playing and then without it, both with fps_max 66.67
the curious thing is that with bepingbooster if i limit it at 66.67, it sticks to that value PERMANENT, without bepingbooster it moves around a bit, stays 66 but moves 66.5-6..etc
Posts: 2,031
Threads: 27
Joined: Nov 2008
Reputation:
17
07-20-2011, 05:35 PM
(This post was last modified: 07-20-2011, 05:35 PM by BehaartesEtwas.)
valve introduced a "limit", so that under normal conditions the server runs only with max. 500 fps. that limit is a bit strange, as those preload libs can circumvent it, but they don't reach their nominal fps but e.g. only half of it. I personally would appreciate if valve would introduce a real hard limit, so that this kind of discussions finally stops.
you are incorrect in some points: important are not fps-values but the time between two ticks (i.e. updates sent to the clients). that has very little to do with hit registration (which could be affected by fps before the orangebox update) but with smooth game play and model movement. if you are running with e.g. 67 fps (i.e. something little higher than the tickrate of 66.67) you will have every then and now a gap between two ticks due to the aliasing effect. if you run with little lower fps (e.g. 66.0) that effect will be gone (and nobody will notice the reduced tickrate by only 1%). alternatively you can run with "much" higher fps like 200~300, than the real fps value does not play any role anymore. meaning it can fluctuate from 200 to 10000 fps if it wants to, gameplay will not be affected. it only must not drop below a certain point (like 200 fps).
the "pingbooster" does not help you in any way to get the fps more stable around 66.67fps. a certain degree of fluctuations is very normal, in fact a constant output of 66.67 (and not a single different value) would be even suspicious.
Posts: 882
Threads: 43
Joined: Dec 2010
Reputation:
13
If they'd just cap it at 66.67 exactly you wouldn't have to teach everyone about it either.
Posts: 2,031
Threads: 27
Joined: Nov 2008
Reputation:
17
in theory yes, but that would create problems for everyone who cannot not achieve that stable enough. for them a couple of 100 fps is better.
Posts: 882
Threads: 43
Joined: Dec 2010
Reputation:
13
Well now they are capping it. Gj.
Posts: 25
Threads: 8
Joined: Jan 2010
Reputation:
0
Posts: 882
Threads: 43
Joined: Dec 2010
Reputation:
13
07-22-2011, 06:22 AM
(This post was last modified: 07-22-2011, 06:25 AM by Mike.)
They are going to cap the fps at 66.67. Or whatever number VALVE decides they want to limit it to.
Edit: I was wrong. They are not capping it, they are restricting the fps_max command. E.g removing it.
|
