1000 FPS + on Debian 5 please

[WolwX]

(11-10-2010, 08:20 PM)lol554 Wrote:  FPS isn't really bad when it doesn't drop below the tickrate.

Yes I know, but I would try to reach a stable 1000 FPS, after I will choose between if I keep my gamerservers at 1000 FPS settings or no

(11-10-2010, 10:05 PM)BehaartesEtwas Wrote:  

(11-10-2010, 02:19 AM)WolwX Wrote:  

Code:

n1-SerVerS:~# chrt -p 11797
pid 11797's current scheduling policy: SCHED_OTHER
pid 11797's current scheduling priority: 0

that's bad. resched.sh does not work. if you run it manually, does it print any error messages?

Quote:When I did that, my processus change to 0 at IO priority, but still 0 at nice level and defaut at IO scheduling class (this one have Defaut, Real Time, Best effort and Idle options)

when you did what? neither IO scheduling nor nice level play any (significant) role. the scheduling policy on the other hand is very important.

And here it's a test result :[/quote]
which is basically the same as before, because nothing changed? the scheduling policy is still SCHED_OTHER, which is simply wrong.

Quote:How I can check, and where I can see the "real time" effect ?
I mean, what command I can use please ?

that was the chrt -p <PID> stuff...
[/quote]

Ok thanks again BehaartesEtwas

I redo all things since the end of the kernel compilation and so here it's a good change à think :

Code:

n1-SerVerS:/usr/local/sbin# ps ax | grep hrtimer
   12 ?        S      0:00 [sirq-hrtimer/0]
   26 ?        S      0:00 [sirq-hrtimer/1]
   39 ?        S      0:00 [sirq-hrtimer/2]
   52 ?        S      0:00 [sirq-hrtimer/3]
   65 ?        S      0:00 [sirq-hrtimer/4]
   78 ?        S      0:00 [sirq-hrtimer/5]
   91 ?        S      0:00 [sirq-hrtimer/6]
  104 ?        S      0:00 [sirq-hrtimer/7]
10054 pts/0    S+     0:00 grep hrtimer
n1-SerVerS:/usr/local/sbin# nano resched.sh
n1-SerVerS:/usr/local/sbin# chmod 755 /usr/local/sbin/resched.sh
n1-SerVerS:/usr/local/sbin# nano /etc/crontab
n1-SerVerS:/usr/local/sbin# /etc/init.d/cron restart
Restarting periodic command scheduler: crond.
n1-SerVerS:/usr/local/sbin# chrt -p 12885
pid 12885's current scheduling policy: SCHED_FIFO
pid 12885's current scheduling priority: 98
n1-SerVerS:/usr/local/sbin#

So time to do some other test ^^

[WolwX]

result here :



so perhaps the last step is tu use the "idler" ?

[Monk]

You could try it. I guess. I still think it's a futile effort, though.

[BehaartesEtwas]

I don't think the idler will help here. but it doesn't take long to try...

what are your server specifications? this looks a bit like your cpu cannot handle 10 players... (post the output of "cat /proc/cpuinfo")

also, are you using any kind of ld_preload lib? because your fps are higher than 1000...

[WolwX]

(11-11-2010, 06:44 PM)BehaartesEtwas Wrote:  I don't think the idler will help here. but it doesn't take long to try...

what are your server specifications? this looks a bit like your cpu cannot handle 10 players... (post the output of "cat /proc/cpuinfo")

here it's the output :

Code:

n1-SerVerS:~# cat /proc/cpuinfo
processor    : 0
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 0
cpu cores    : 4
apicid        : 0
initial apicid    : 0
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3723.43
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

processor    : 1
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 1
cpu cores    : 4
apicid        : 2
initial apicid    : 2
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3722.91
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

processor    : 2
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 2
cpu cores    : 4
apicid        : 4
initial apicid    : 4
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3722.91
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

processor    : 3
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 3
cpu cores    : 4
apicid        : 6
initial apicid    : 6
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3722.91
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

processor    : 4
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 0
cpu cores    : 4
apicid        : 1
initial apicid    : 1
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3722.90
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

processor    : 5
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 1
cpu cores    : 4
apicid        : 3
initial apicid    : 3
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3722.91
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

processor    : 6
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 2
cpu cores    : 4
apicid        : 5
initial apicid    : 5
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3722.91
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

processor    : 7
vendor_id    : GenuineIntel
cpu family    : 6
model        : 30
model name    : Intel(R) Xeon(R) CPU           L3426  @ 1.87GHz
stepping    : 5
cpu MHz        : 1861.717
cache size    : 8192 KB
physical id    : 0
siblings    : 8
core id        : 3
cpu cores    : 4
apicid        : 7
initial apicid    : 7
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips    : 3722.91
clflush size    : 64
cache_alignment    : 64
address sizes    : 36 bits physical, 48 bits virtual
power management:

(11-11-2010, 06:44 PM)BehaartesEtwas Wrote:  also, are you using any kind of ld_preload lib? because your fps are higher than 1000...

yes I have tried the ld_preload lib for fps but wasn't sure it worked before

Code:

[...]
SourceTV broadcast active.
Installing breakpad exception handler for appid(srcds_linux)/version(1.0)
Connection to Steam servers successful.
ERROR: ld.so: object '/home/n1-SerVerS/scripts/libBEpingboost.so.1.0.1' from LD_PRELOAD cannot be preloaded: ignored.
ERROR: ld.so: object '/home/n1-SerVerS/scripts/libBEpingboost.so.1.0.1' from LD_PRELOAD cannot be preloaded: ignored.
ERROR: ld.so: object '/home/n1-SerVerS/scripts/libBEpingboost.so.1.0.1' from LD_PRELOAD cannot be preloaded: ignored.
ERROR: ld.so: object '/home/n1-SerVerS/scripts/libBEpingboost.so.1.0.1' from LD_PRELOAD cannot be preloaded: ignored.
VAC secure mode is activated.
**********************************************************************
BEpingboost Copyright (C) 2009 BehaartesEtwas@fragaholics.de
This program comes with ABSOLUTELY NO WARRANTY.
This is free software, and you are welcome to redistribute it
under certain conditions.
Released under GPLv3 license.

Version: 2009-12-22/1
Initialized with 1500 fps nominal
**********************************************************************
L 11/12/2010 - 07:59:21: World triggered "Round_Start"
L 11/12/2010 - 07:59:21: server_cvar: "sv_pausable" "0"

I will try this afternoon with the idler and without your BEpingboost library.

[BehaartesEtwas]

then it works. but I recommend not using it. as clearly stated on the download site, this is for playing around, not for productive environments.

[WolwX]

(11-12-2010, 08:30 PM)BehaartesEtwas Wrote:  the it works. but I recommend not using it. as clearly stated on the download site, this is for playing around, not for productive environments.

I agree, that's just for testing purpose ^^

I just want at max 1000 FPS stable, any higher value doesn't matter for me, but every test are always good for the knowledge

About my processor, there is no problem no ? ^^

Perhaps the SourceTV provoque thoses fps drops/instabilities ?

[BehaartesEtwas]

sourcetv quite often makes trouble. maybe you should first test without it.

the cpu should be fine for smaller servers, but I am not entirely sure. especially I don't know what implications that turbo boost technology has on game servers. I could imagine instable fps if the cpu changes its clock speed. try switching it off. also try switching hyper threading off.

in any case - sorry for repeating myself, but this cannot be said often enough - fps > 66 will only increase cpu usage and not improve the quality. especially if you have low cpu resources higher fps will only *reduce* the server quality. so my final recommendation is: run with 66 fps, without source-tv, and get the fps stable. look at the "TIME RMS" value in the graph, it is the variance in microseconds. it should be way below 1000. something around 20 is a very good and realistic value.

[WolwX]

(11-12-2010, 11:15 PM)BehaartesEtwas Wrote:  sourcetv quite often makes trouble. maybe you should first test without it.

the cpu should be fine for smaller servers, but I am not entirely sure. especially I don't know what implications that turbo boost technology has on game servers. I could imagine instable fps if the cpu changes its clock speed. try switching it off. also try switching hyper threading off.

in any case - sorry for repeating myself, but this cannot be said often enough - fps > 66 will only increase cpu usage and not improve the quality. especially if you have low cpu resources higher fps will only *reduce* the server quality. so my final recommendation is: run with 66 fps, without source-tv, and get the fps stable. look at the "TIME RMS" value in the graph, it is the variance in microseconds. it should be way below 1000. something around 20 is a very good and realistic value.

Tried without the boost fps and with the idler =>



That's not better :s

So I will try to find another way to improve now

[Terrorkarotte]

simple question. Why do you ignore the advice and do not lower the useless high fps?

[WolwX]

(11-14-2010, 03:44 AM)Terrorkarotte Wrote:  simple question. Why do you ignore the advice and do not lower the useless high fps?

I'm not ignoring that, I know that's not necessary since the server must have at less the same number of fps than the tickrate.

But since many people look like to reach this goal of stable higher FPS, I want to reach the same goal, and after that I will balance the two settings to choose

[Terrorkarotte]

Or in other words you want to sell 1000fps and want us to do the work for you so that you can get money out of the people for products they do not need...

[WolwX]

(11-14-2010, 06:41 AM)Terrorkarotte Wrote:  Or in other words you want to sell 1000fps and want us to do the work for you so that you can get money out of the people for products they do not need...

Lollllllllllll

You are joking ?

Where do you see I'm searching to reach 1000 FPS to sell 1000 FPS server ?

Tell me what things you told me to help me ?

If you are here to flame, go reply in another topic please, if not, be fair and don't start to tell bad things.

Another test, with another dedicated server using different processor :



and resched.sh ON, so perhaps I must use some optimisations

Anyone having a 1000 fps stable can help me ?

[Terrorkarotte]

I have writen enough things that would have helped any person that would listen to reasons.

Read the second post. There I have written what need to be known. A Server at +100 fps will have the same ingame performance as a stable 1000 and probably no drops.
Also I told you already do not use RT kernels, still you use them.

We are tried helping you to run good gamesrevers. FPS are not really that relevant anymore.

If you do not listen and still aim for high fps there are only two conclusion. First you do can not be brought to reason second is that you might have the reason in selling those gameserver. Reading the servername n1-servers.fr tells me that a company is behind this.

[lol554]

Hi. I can suggest ubuntu. Im getting this on a standard kernel with stv on. http://www.fpsmeter.org/p,view;99595.html