SRCDS Steam group


Multiple servers, unstable fps [HLDS]
#31
Question; If i dont want 1000fps, could i select like 300hz with HRtimer instead of 1000hz?
Reply
#32
fps_max 300

you don't need to select your target fps somehow in the kernel or system. hrtimers are timers that can fire at arbitrary frequencies and are not bound to any ticks in the kernel. unlike with the old "timer wheel" in the kernel, hrtimers need no intrinsic overhead to allow high timer frequencies, meaning if you are able to produce 1000fps with your system but limit the server to 300, you will only have the overhead of 300fps.

but beware that hlds/srcds always waits a multiple of 1ms after each frame, so you can only achieve certain fps values like 1000, 500, 333, 250, 200, ... (or values slightly below those "magic numbers").
http://www.fpsmeter.org
http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide (Linux Kernel HOWTO!)
Do not ask technical questions via PM!
Reply
#33
Do you need dyn. ticks for HRtimer to work? or can i just go with 100hz / hrtimer without dyn. ticks?

BehaartesEtwas: Why 1000hz and HRtimer support in your guide? Smile And i know; "Dont look at the cpu usage", but why higher "usage" with HRtimer?
Reply
#34
i tried 100hz with hrtimer. i got with more then 4 processes on a quad core higher CPU usage than normal.
Reply
#35
matte563 Wrote:Do you need dyn. ticks for HRtimer to work? or can i just go with 100hz / hrtimer without dyn. ticks?
hrtimers work w/o dynamic ticks. dynamic ticks only lower the kernel HZ dynamically if the additional kernel ticks are not needed.

Quote:BehaartesEtwas: Why 1000hz and HRtimer support in your guide? Smile
Because that works for most people. I think it is better to provide a "safe" solution for a fast result and then give hints how to optimize even further. Else ppl might get frustrated even sooner ;-)

Quote:And i know; "Dont look at the cpu usage", but why higher "usage" with HRtimer?
Maybe the usage is not higher, only top displays a higher value... I don't know why. But I also don't know how the cpu usage is measured. It could be some correlation effect, e.g. if the cpu usage is always (or preferably) measured when the server renders a frame, it seems like the cpu is at 100% (or close) load. But this is only speculation, I really don't know. In theory the load should be lower than without hrtimers, because it should have less overhead.
http://www.fpsmeter.org
http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide (Linux Kernel HOWTO!)
Do not ask technical questions via PM!
Reply
#36
Quote:Because that works for most people. I think it is better to provide a "safe" solution for a fast result and then give hints how to optimize even further. Else ppl might get frustrated even sooner ;-)

True..
Reply
#37
thats true .... but it seems very bad ob top.
Reply
#38
Q: Why do you need pingboost if the server already is "boosted"?

EDIT: Im trying to figure why the FPS is unsteady with multiple servers :/. And yeah... the servers are empty.
Tried pretty much everything; 100hz/1000hz with/without HRtimer, Low/High ticrate, kernelversions etc :/.
Reply
#39
how to get 1000 fps in win xp?
Reply
#40
? what ? win xp server ? :-)
Reply
#41
zalvador Wrote:how to get 1000 fps in win xp?

this is a linux forum Toungue
http://www.fpsmeter.org
http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide (Linux Kernel HOWTO!)
Do not ask technical questions via PM!
Reply
#42
right.
Reply
#43
Is this really right: "Linux debian 2.6.28.10 #10 PREEMPT Fri May 29 14:50:58 CEST 2009 i686 GNU/Linux" ? (AMD Athlon XP 1800+)
Reply
#44
Bump: Actually works better in Archlinux with 2.6.26.8 kernel.

EDIT: I guess the "rest" depends on my hardware, old singlecore CPU.
Reply


Forum Jump:


Users browsing this thread: 2 Guest(s)