Hoi!
I have a big problem with the new spring version.
If i host more then 8 people the game will always go out of sync,
resulting in ping times like 20000 ms for all players.
Even with 8 players this will happen often.
In the previous Spring version i could host 10 with 6 specs without problems.
I have disabled all new graphics options but i wont help.
I think this is a really BIG BUG in the new spring version.
Maybe it is caused by the new ping calculation.
I hope you have a solution for me...
jj
High ping times :(
Moderator: Moderators
I also noticed considerably slower gameplay due to network load, i think. I was playing 2v2 on core prime , total of about 1200 units (all 4 teams) and ive never seen a slower battle than that in my whole spring experience. (slow as in 1 second taking 10), when opponent ctrl+d'ed it even killed my connection
Re: High ping times :(
I think that the new ping calculation is just to old one, except the number is ten times bigger and a nifty ms is added on the end.jj wrote:Maybe it is caused by the new ping calculation.
jj
Last edited by patmo98 on 14 May 2006, 02:37, edited 1 time in total.
ping problems can come when HOST has slow upload. I tested spring and wih like 30 units it was consuming 1kB/s per player. This could grow up to 3kB/s (not tested).
So when u host big game with 16 players host needs 16*3 = 48kB/s = 384kbit/s minimum.
When u count in overhead 512 kbits line would do it if it is good and clean.
So good host must have altest 512/512 line and decent cpu.
So when u host big game with 16 players host needs 16*3 = 48kB/s = 384kbit/s minimum.
When u count in overhead 512 kbits line would do it if it is good and clean.
So good host must have altest 512/512 line and decent cpu.
I have 1536 kbps down and 512 kbps up.
So this is not the problem.
It's strange that i could host 16 with the old version and 8 with the new version, with the same connection.
by the way, I dont believe TA will need 512 kbps
So this is not the problem.
It's strange that i could host 16 with the old version and 8 with the new version, with the same connection.
by the way, I dont believe TA will need 512 kbps
Last edited by jj on 12 May 2006, 18:07, edited 1 time in total.
HIGH SPEED internet... :)
Ok, i now have faster internet:
5120 kbps down
1024 kpps up
I hope this will help.
I will try to host 10 players with a lot of specs.
If this not help the ping times then i'm doomed,
because i have to pay a lot to downgrade my internet connection
5120 kbps down
1024 kpps up
I hope this will help.
I will try to host 10 players with a lot of specs.
If this not help the ping times then i'm doomed,
because i have to pay a lot to downgrade my internet connection
A report I was going to make a while ago, but retracted because I didn't have enough details:
For reasons unknown to me, spring floods larger UDP packets to resync (?), usually over 500 bytes in size at a very fast rate.. sometimes to various hosts, taking well over 600kbit in upstream bandwidth.
This prevents me from receiving any data, since upstream is required for downstream, causing a massive nosync error.
I'm sure this isn't supposed to happen, but I think it'd help if Spring could cap the amount of data it sends to each user to a certain rate, so that way if there's a problem everyone doesn't fall out.
It appears to happen more often when I host larger games (AA). However the same error occured on NanoBlobs Mod with xenoargh, which was a 1vs1.
I've included a PCAP file that can be opened in Ethereal if you wish to look at the raw udp packets.
--
Ignore the PCAP line.
For reasons unknown to me, spring floods larger UDP packets to resync (?), usually over 500 bytes in size at a very fast rate.. sometimes to various hosts, taking well over 600kbit in upstream bandwidth.
This prevents me from receiving any data, since upstream is required for downstream, causing a massive nosync error.
I'm sure this isn't supposed to happen, but I think it'd help if Spring could cap the amount of data it sends to each user to a certain rate, so that way if there's a problem everyone doesn't fall out.
It appears to happen more often when I host larger games (AA). However the same error occured on NanoBlobs Mod with xenoargh, which was a 1vs1.
I've included a PCAP file that can be opened in Ethereal if you wish to look at the raw udp packets.
--
Ignore the PCAP line.
Yes, i have also noticed this problem with the new version. Could this have something to do with the new water shader not working properly ? In some thread i read that the reason the shader is lagging is that it maybe tries to sync the water deformation with other players and that causes so massive net traffic it goes out of sync ?
No i'm just guessing. But what can you do? I wished the devs would be a bit more willing to help the community, inform where we are and whats going on.
No i'm just guessing. But what can you do? I wished the devs would be a bit more willing to help the community, inform where we are and whats going on.
Conclusion
With the faster internet connection i dont have problems anymore.
I now have 1024 kbps up.
My conclusion is: The new spring version needs more then 512kbps with 10 players.
The previous version didnt.
I think this is bad and need to be fixed.
I now have 1024 kbps up.
My conclusion is: The new spring version needs more then 512kbps with 10 players.
The previous version didnt.
I think this is bad and need to be fixed.
Re: Conclusion
What, the fact you need at least a half mbit connection to play a 10 player game?jj wrote:I think this is bad and need to be fixed.
I don't think anything changed in the netcode for a long time, except the sync check messages are a little bigger since a while, but that's at most 20 bytes for every player per 2 seconds (100 bytes/second in a 10 player game). Note that this is downstream for the host tho... upstream nothing changed as far as I can see.
E: maybe something in giving commands etc. changed (which are sent over the net), I don't know that
E: maybe something in giving commands etc. changed (which are sent over the net), I don't know that
Spring definately does use a lot more than ~100bytes/second. It sends ~500 bytes udp packets to various users last I checked, several times in a second, easily eating up a lot of upload bandwidth. I'm pretty sure this is a bug, one that didn't seem to be present back in 0.67.
This is a problem that occurs not only during a 10 player game, but seems to occur more often and a lot less likely to recover from it in a large game.
Several times I've tried to play a nanoblobs game with xenoargh (and no one else) and this problem has occured, regardless of who was host. Xenoargh's conclusion was that it was related to the fact we both loved to 'box build'.
I've included a link to fileuniverse for a Ethereal compatible PCAP file for debugging purposes. (If you're just browsing this thread and don't know anything about Spring's net handling code, you probably don't want to download this)
http://www.fileuniverse.com/spring/springerror.pcap
Best Regards and Thanks for the amazing work,
altruist
This is a problem that occurs not only during a 10 player game, but seems to occur more often and a lot less likely to recover from it in a large game.
Several times I've tried to play a nanoblobs game with xenoargh (and no one else) and this problem has occured, regardless of who was host. Xenoargh's conclusion was that it was related to the fact we both loved to 'box build'.
I've included a link to fileuniverse for a Ethereal compatible PCAP file for debugging purposes. (If you're just browsing this thread and don't know anything about Spring's net handling code, you probably don't want to download this)
http://www.fileuniverse.com/spring/springerror.pcap
Best Regards and Thanks for the amazing work,
altruist
Read my post. I didn't say spring uses that, I said that is the maximum it may have increased since versions before 0.70...altruist wrote:Spring definately does use a lot more than ~100bytes/second.
This doesn't sound like a bug to me... 3 kB/s/player (some people's bw usage estimate of spring) makes 6 times a 500 byte packet per second per player... Unless you mean it should be more spread out through the second...It sends ~500 bytes udp packets to various users last I checked, several times in a second, easily eating up a lot of upload bandwidth. I'm pretty sure this is a bug, one that didn't seem to be present back in 0.67.
All info sent over the net is saved in the demo file too, so I think that one can more easily be used for debugging. I might even make some binary replay file to text converter sometime...I've included a link to fileuniverse for a Ethereal compatible PCAP file for debugging purposes. (If you're just browsing this thread and don't know anything about Spring's net handling code, you probably don't want to download this)
http://www.fileuniverse.com/spring/springerror.pcap
Best Regards and Thanks for the amazing work,
altruist
Also, by dividing replay size by total game time you obtain the bandwidth spring uses (I think per player) on average.
Sorry I think I misunderstood you somewhere. I thought you said that it would take 100 bytes for a 10 player game, and that nothing has changed in the upstream code since from what you've noticed in a long time.Tobi wrote: Read my post. I didn't say spring uses that, I said that is the maximum it may have increased since versions before 0.70...
It definately takes more than 3 KB/s when everyone falls out of sync. I think I made a mistake here in saying several, because for a 10 player game, it can take well over 70KBytes/s. And I'm pretty sure it's a bug, because it has happened with only 2 players (me and xenoargh) a while ago.. we couldn't figure out why exactly. Someone told xenoargh it was related to the fact we both liked to 'box build' units.Tobi wrote: This doesn't sound like a bug to me... 3 kB/s/player (some people's bw usage estimate of spring) makes 6 times a 500 byte packet per second per player... Unless you mean it should be more spread out through the second...
That explains a lot. Another player told me he noticed lagged games created larger demos. Thanks for letting me know. Now I don't need to go through the trouble of packet sniffing :) Probably a good idea to get this bug to occur with the latest version. Next game I host with this problem, I'll remember to send the demo.All info sent over the net is saved in the demo file too, so I think that one can more easily be used for debugging. I might even make some binary replay file to text converter sometime...
I am certain this is a bug. It has happened during several games and does seem to be related to unit selection. I have added some more information to your original bug report in Mantis: http://taspring.clan-sy.com/mantis/view.php?id=201altruist wrote: It definately takes more than 3 KB/s when everyone falls out of sync. I think I made a mistake here in saying several, because for a 10 player game, it can take well over 70KBytes/s. And I'm pretty sure it's a bug, because it has happened with only 2 players (me and xenoargh) a while ago.. we couldn't figure out why exactly. Someone told xenoargh it was related to the fact we both liked to 'box build' units.
-
el_matarife
- Posts: 933
- Joined: 27 Feb 2006, 02:04
I just bumped this thread and the Mantis entry because this bug is still occuring. I just had a game fall victim to it yesterday. The steps to repeat it are simple, just drag out a large build queue. Suddenly, every player will report as sending no sync response, the units in game will not take orders and the game will act as if its just been paused.