Try increasing net_splitpacket_maxrate to something usable like 50k (We're using 75k, but we have 64 players). It looks like we also have net_splitrate set as well (to 5), but I don't remember that having a huge effect.
Thanks, Kyle. On Sat, Jan 5, 2013 at 10:51 AM, Essay Tew Phaun <sc2p...@gmail.com> wrote: > We used to, but haven't for several months now. > > On Sat, Jan 5, 2013 at 1:32 PM, ics <i...@ics-base.net> wrote: > > > If you started seeing spikes recently, after the xmas update, there's > that > > spam in the console that makes my own game lag like hell when it happens. > > Screen even jams occasionally. > > > > Before that, i saw mostly players going fast forward with leaping place > to > > another when issues like this occurs. It's just some players, just as > some > > packets would be "lost" (or not given to you) at that time and only part > of > > his actual movement is shown (like packet 3, 35 64 and not > > 3,12,23,35,43,51,64,76,89 etc). As i recall, this started long ago, at > the > > time that valve optimized the code and bandwidth usage for servers > dropped > > at the same time. > > > > Also one other thing happened at that t ime, which is that now the game > > draws only players that are around the next corner and not the ones at > next > > corner and after the next one. This of course can be coped with some > > optimization on maps but this is now default setting. If you want to > > experience this as a test, go in map harvest and if you spec a sniper > that > > is in the roof but is not peaking over the rooftop to middle. You cannot > > see the enemies that are in area on front of the rooftop. However, as > soon > > as he peaks over, the enemies magically appear for your screen, which > means > > you see the same thing as he does. Only he cannot see what you see from > > above his head, sooner than he. So if you see the enemies and he is > peeking > > over and then goes down crouching, those enemies will disappear. > > > > Before Valve did the optimization over a year ago, as a 3rd person, you > > could see all that is happening on the battlefield while spectating a > > person. Now, this effect happens. In some rare and not so rare cases, it > > can lead to sudden lag as your hardware tries to catch up on whats going > on > > at the screen. Naturally this part has nothing to do with servers > > themselves, it's just ingame thing on players. That recent dynamic model > > loading might have done this even more visible but i think they are > loading > > the models that should be coming up in advance but i have no way of > knowing > > how advanced that system is at guessing what comes next. > > > > But also one question, do you force your servers to run on specific cpu > > core? I've also seen that causing similiar lag effect. > > > > -ics > > > > 5.1.2013 19:53, Jake Forrester kirjoitti: > > > > I have the exact same issue. When turning the corner, getting team > >> changed, or just running into a bunch of other players, I get huge > >> spikes in my graph (with gaps). It's gotten so bad that it can become > >> impossible to aim during large enemy encounters, so I've just given up > >> on TF2 for the time being. I guess on to DOTA? Kind of sucks being in > >> charge of a server that you can't even play on. > >> > >> On 1/5/2013 11:49 AM, Essay Tew Phaun wrote: > >> > >>> I know this isn't exactly scientific but we weren't having these > reports > >>> of > >>> lag before. Of course, you're always going to have a few people > >>> complaining > >>> about lag and blaming their performance on it, but recently, probably > as > >>> of > >>> the big update (Halloween or Mecha, not sure which) We've seen a > >>> humongous > >>> rise in the complaints about lag. Whether that's the cause of the > server > >>> or > >>> client, I guess that remains to be seen, but we're having a lot of > >>> complaints and have done everything we can on our end to try and > address > >>> it > >>> including moving down to 24 player servers which we did *NOT *want to > do. > >>> The problem is still there. An hour or so ago I loaded up a completely > >>> vanilla server with no MM:SM, the same issue appears. > >>> > >>> We get the gaps and spikes in the net_graph a lot. I get HUGE gaps when > >>> going in to a new area of a map a lot of the time. Here's a screenshot > >>> from > >>> just now when I tested. > >>> > >>> http://i.imgur.com/4V1TZ.jpg > >>> > >>> > >>> > > > > ______________________________**_________________ > > To unsubscribe, edit your list preferences, or view the list archives, > > please visit: > > https://list.valvesoftware.**com/cgi-bin/mailman/listinfo/**hlds_linux< > https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux> > > > _______________________________________________ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux > _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
