Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Performance of LAN games
#1
Question 
Hello!

I know this issue hasn't got something to with Ares, but the Ares guys can probably answer this question: why do LAN games heavily slow down if there are a lot of units in the game?

When playing a skimirsh game, I can build thousands of units without slowing the game down. But in LAN games, with 6 to 8 players on a big map with about 300 units overall, it's getting very slow. Not yet unplayable, but about half as fast as normal speed.

I checked the network load in the network status of windows and it seems that only a few KB of data are received. Forcing my router to use 10mbps mode does not slow down the game. So network speed isn't the problem.

10 years ago, when using IPX, i think the games were not slower altough the computers were. My old P3 1.4Ghz rocked ass 10 years ago. Today, I cannot use this computer for playing YR over LAN, because it slows down the game too much. All computers must have at least 2 Ghz to reach a proper speed. Might cause the LAN-patch (wsock32.dll) the slowdowns?

This problem has been discussed in hundreds of threads (see google +slow +LAN +RA2) before, so there is probably no solution... but there is a reason and I would like to know it.

Cheers!


Reply
#2
If you provide a log from Very Sleepy (http://forums.renegadeprojects.com/showt...p?tid=1476 ), we might get a better idea.

Worth playing: 1 | 2 | 3
Reply
#3
Its becuse the net code is old and crap. Even if the ping is under 50ms, every command will be delayed.
In fact, the net code is so bad, I only play 1vs1 games. Even trees create lag >_>
The problem have never been the connection. I can confirm this with years of testing(with a very good conenction).
I dont know whether or not it is possible to improve the net code with Ares/Syringe(I think a complete rewrite is the only thing that can be done), but if it is, then I suggest(and probably many more) that you guys should try to improve it. But that may be more work than all the features combined.
Java student.
Reply
#4
Weapons system is crap? Need to rewrite that shit.
Graphics code is slow? Need to rewrite that shit.
Net code is even slower? Need to rewrite that shit.
AI is braindead? Need to rewrite that shit.

I can almost smell a trend here... The more of these problems are brought up, the less motivation there is to keep piling lipstick on this pig of an engine.

Nobody should take this personally and stop bringing up actual problems. This is just an observation about the general state of the engine.

Worth playing: 1 | 2 | 3
Reply
#5
(23.04.2011, 12:47:46)DCoder Wrote: If you provide a log from Very Sleepy (http://forums.renegadeprojects.com/showt...p?tid=1476 ), we might get a better idea.

Here is it. I used four computers, one of them much slower compared to the three others. On that one I ran VerySleepy during a multiplayer game. I used a huge map, which causes the game to be slow from the beginning on.

I exspected the CPU load to be close to 100% on the slowest computer and much less at the others, but the CPU was about 75% on all four. It indicates that the game is waiting one quarter of the time for synchronizing messages/events to arrive. No doubt - that's crap. :-)

But there is also an Ares function, which seems to eat up a lot of cpu power: BulletTypeClass_Load_Suffix. Ares version used is 0.1.1036.

BTW: Is there a way I can find the reason for an IE? Maybe some crash dump analyzing tool? I got two IEs during this test and I am curious whats causing them. I don't want to flood the bugtracker with my IEs, because I am sure the reason is my mod and not Ares...



Attached Files
.zip   m4.zip (Size: 126.31 KB / Downloads: 447)
Reply
#6
(25.04.2011, 01:16:43)GMBigB Wrote: BTW: Is there a way I can find the reason for an IE? Maybe some crash dump analyzing tool? I got two IEs during this test and I am curious whats causing them. I don't want to flood the bugtracker with my IEs, because I am sure the reason is my mod and not Ares...

Besides the list on ModEnc (http://modenc.renegadeprojects.com/Internal_Error), only reading through debug.log can tell us, ordinary modders the cause. Extcrashdump and except needs an actual YR knowledge to read from.
Reply
#7
(25.04.2011, 01:45:34)Graion Dilach Wrote: Besides the list on ModEnc (http://modenc.renegadeprojects.com/Internal_Error), only reading through debug.log can tell us, ordinary modders the cause. Extcrashdump and except needs an actual YR knowledge to read from.

007569FA is not in the list. At least all IEs seem to have the same cause. A file debug.log was not written. Most IEs happen instantly after doing something special, but this one occures without being triggered directly by a user action.

Do the Ares developers know what most of the function do? It might be useful to get a roughly idea of what YR is doing before the IE happens.
Reply
#8
007569FA means the game was trying to draw a voxel and something went wrong. To get more data, a full crash dump is needed.

BulletTypeClass_Load_Suffix is not the problem, it's simply the last exported function in Ares.dll, so all the calls to the part of Ares after it are blamed on it. Most of the callers for this record seem to be related to TechnoClass::Update, which seems reasonable - that function is called for each unit/infantry/aircraft/building on the map, every single frame.
Other than 7BC806, which is responsible for drawing a lot of crap and might be sped up by setting all surfaces to system memory ( http://ares.strategy-x.com/Documentation...aceDrawing ), I can't see any unusual issues.

Worth playing: 1 | 2 | 3
Reply
#9
Oh no... not again. I had a similar problem earlier: http://bugs.renegadeprojects.com/view.php?id=1527
Is it HIND2 again? Or another unit this time?
http://www.game64.de/1stcrash.7z

The IE does no longer occur in skirmish games. In have build HIND2 in this particular game, but the error did not happen when it was first drawn. And I did not move the unit. Only one player got the IE and that player was not the one who build the unit. That's what I remember for sure.

I will try to make a performance test with surfaces in system memory. Maybe it gains a few percent in speed.
Reply




Users browsing this thread: 1 Guest(s)