huge lag and sync problems with 0.2.8.1

iDiot / sPuNkY · Post by **iDiot / sPuNkY** » Fri Apr 21, 2006 6:15 pm

Hi

I've got huge lag and sync problems since i use 0.2.8.1-1...
When i'm playing in Singleplayer mode i've got no Problems..
but when i want to play online my bike just jumps from place to place...
also the speed jumps sometimes in 25 steps...
my walls went from visible to invisible by every turn...

I'm playing since 1 1/2 years now...
I'd played a long time with 0.2.7.1 and i had no problems but after most server switched to 0.2.8.1 i decided to use this version, too.

I downloaded armagetronad-0.2.8.1-1.gcc.win32.exe
and installed it over 0.2.7.1... this didn't woked at all...
So i uninstalled 0.2.7.1 and 0.2.8.1 to perform a clean installation..

This worked better but to my suprise i had to note the lag and sync problems...

I have skills in computer and programming, so i tried to check and modify the settings.cfg...
but there were no big changes...

i thought.. "well let's check the vc6 compiled version"... but it's the same behavior...

so i made big changes at the settings again... well now my bike isn't jumping anymore but i still have huge lag problems...

well my ping at the servers is between 70 and 120 but it acts like 1000...
it's unplayable...

I made a record with Armagetron Advanced Record, i don't know if you can use it.. but i uploaded it with my (now extremly modified) settings.cfg..

you can download it @ http://www.wirsch.com/arma.zip

I use:

- Strato DSL 2000

PC:
WinXP SP1
P4 HT 3 GHz @ 3,6 GHz
1GB DDR RAM
GeForce 4 Ti 4200

i hope my problems can be fixed...
if there are any questions.. please ask..
thanks ahead for the support...

greez sPuNkY

Website · Post by **Z-Man** » Fri Apr 21, 2006 10:50 pm

Thanks, a recording is just the right thing. I just glanced into it; your turns seem to suffer from packet loss (we have new compensation code and it's hard to test under real world conditions), the speed syncing problems may be a result from server and client disagreeing what the acceleration should be. I'll actually look at that in the debugger later.

You don't have to send the settings, they're included in the recording.

iDiot / sPuNkY · Post by **iDiot / sPuNkY** » Sat Apr 22, 2006 1:47 pm

thx i hope we can fix that problem...

well i asked while "playing" if anybody else have those problems and it seems that i am the only one...

so i volunteer as a test person..
i've got eclipse, cygwin and the gcc compiler installed, so if you want, you can just send me the test source code with some values i should modify, so that i can immediately see and report the changes...

Website · Post by **Z-Man** » Sat Apr 22, 2006 5:51 pm

Gee, I had only watched the first minute up to now, things get really messy later. It looks like the timer syncing bugs we sometimes had with early betas, so I checked the timer syncing and made this interesting discovery:
your system's real time clock is running about 5% faster than the server's clock. That either means the server sits in a low orbit around a neutron star, or something is wrong with your setup. I suggest you check for the latter given that nobody else has your problems. On Unix, I'd sync my timer with the standard internet time servers repeatedly and watch the reported correction, I don't know how that can be done in Windows.

We can in principle work around this; the 0.2.7.1 code indeed was more tolerant about timer speed differences, but on the other hand, it was more susceptible to fluctuating ping times which affects more people.

iDiot / sPuNkY · Post by **iDiot / sPuNkY** » Sat Apr 22, 2006 8:12 pm

wow.. that blows my mind...

i think this could have be something to to with my system..

I use a MSI Neo2 PFSIR motherboard with core cell and DOT...
It's a dynamic overclocking technology which modulates the FSB...
The FSB clock speed ranges between 100 and 120% against the system usage...

to synchonize my sytem clock i tried:

Code: Select all

net time /SETSNTP:ptbtime1.ptb.de  // in the windows cmd

but it didn't worked at all...well my time is set correctly now.. but i think it was just a one time synchonisation...

and well.. i don't know how to adjust my system clock speed.. but i noticed, that it is really a bit too fast...

i just uploaded my system info...
available @ http://www.wirsch.com/system.zip EDIT: i removed it because the problem is solved

perhaps u can use it...

Website · Post by **Z-Man** » Sun Apr 23, 2006 12:36 am

Sorry, I'm not a hardware zoologist, I can't help you there with the specifics.

If you still want to give hacking a try, there is code for another way to fetch the time in Windows; it only gets activated when the high performance counter is unavailable. Search tSysTime.cpp for the two calls to QueryPerformanceFrequency and replace them by false.

You can download the code where you got the executable, choose the zip. Also fetch the winlibs.zip from the sourceforge file releases. I'm afraid Eclipse is not supported right now as an IDE in Windows, you need either Code::Blocks or Visual Studio 6 or 8.

iDiot / sPuNkY · Post by **iDiot / sPuNkY** » Sun Apr 23, 2006 1:02 am

i've got it now... it was the MSI D.O.T...

i just disabled the D.O.T and it's working now...

well... i'm pissed of MSI.. they really just boost the FSB...(it was @ 210MHz instead of 200MHz)
so all of my components were too fast, even my clock...

but i think there also has to be a bug in the code of Arma, cuase it doesn't use the actual FSB speed for it's sync calculations...

when you programm the 0.2.9.X please include a system clock check routine which controls the values of the sync routine by using the actual FSB/system clock speed, cause i think there are other player who are using D.O.T ...

z-man wrote: I'm afraid Eclipse is not supported right now as an IDE in Windows, you need either Code::Blocks or Visual Studio 6 or 8.

with Cygwin (kind of UNIX emulation) it's possible to use eclipse as an IDE in Windows... i work with it in the advanced technical college and it runs great... i like it because of the CVS repository function... multiple developers can write at one program at the same time and compare it almost in realtime... perhaps i'll try to modify the tSysTime.cpp as you said.. then i'll report if it works...

thx for your great support.. i would never had figured it out by myselv..

Website · Post by **Z-Man** » Sun Apr 23, 2006 2:34 pm

Umm, how would we be supposed to check the clock? What should we compare it with?

And more importantly, what should we do if it goes wrong? If the DOT changes the clock speed while the game runs, what would we do about that?

iDiot / sPuNkY · Post by **iDiot / sPuNkY** » Sun Apr 23, 2006 2:45 pm

Well... i don't know.. i thought you use the FSB for the sync calculations..
but that is way out of my knowledge

...

(sarcasm: perhaps a realtime calculation based on everything thats going on in the CPU)
well... never mind.. it's almost working now... sometimes i have short sync trouble, but it's playable now...

as i said i'll try to modify the tSysTime.cpp... and change the QueryPerformanceFrequency to false... perhaps this will work better on my system...