Wirehead Studios

Uh oh, here goes Tab, ranting about a problem again..  ;]

What problem? The problem of choppy connections to servers. While the game may seem to run smoothly on the client side (it is actually transparantly choppy, so to speak), there can be some real scary shit going on for the rest of the players.

http://www.tabun.nl/tmp/gen/laggy_doom_ohdear.zip (http://www.tabun.nl/tmp/gen/laggy_doom_ohdear.zip)

While I by no means am aiming well in this demo, I guess anyone would have a hard time here ;]

For more info on helping prevent this effect (provided this is possible and/or not ISP-related, ofcourse), read the FAQ on our website: http://www.wireheadstudios.org/generations/faq.php (http://www.wireheadstudios.org/generations/faq.php)

Something not noted in the FAQ (unless I added and forgot about it) is that very low FPS (say, 40 and below) generate choppy play similar to this too. So if you're running high quality and your machine can't quite handle it, tune things down a bit so you won't chop so much.

Disclaimer: Not directed at you Washu, but I used tonights demo simply because it shows the problem so clearly. Normally you don't chop this much, of that I am sure.

Washu I didn't have trouble hitting, except from just plain missing and being out of practice, but I didn't run into anything ping-related outside of my own stupid lag.  I still have trouble with Kingu teleporting all over the place though, and poor scalliano was having system problems and kept disconnecting. :(

Oh that's no problem Tab, I'm flattered that I'm the center of a demonstration. :]
I seen the demo, and that is so odd, why would that happen? I was just fine from my end.
My graphics were really high, 1280X1024, 32bit textures,, highest quality textures, adv. shadows, but my machine ran perfectly, because (not to brag :] ) it is extremely powerful and fast. I don't know how that can effect netcode though.
Thats weird, watching myself, attack Tab from his perpective, weird, and fun too.

I think that on that machhine, celeron 1.3   256 sdram (100hz) Radeon 9250 256MB  i cant help getting more fps.
 I use this settings now:
r_mapoverbright 1
cg_dlightlevel 1
cg_dlighttype 2
cg_dlightlightning 1
cg_dlightflame 1
r_picmip 1
Cant help it but r_shadows 2
Changed now r_detailedtextures to 0 but i think it doezn't help mutch.
 1024x768 if fullscreen and 960x720 if in window
geometric detail high, 32 bits

 Any ideas ?

32 bit textures really does a bit on fps turn it to 16 bit, and I'd change the rez to 800X600 for good measure. And if you have those complex shadows, turn them to blob, I tried that complex shadow thing, and it cut my FPS in half!

Dynamic lighting and r_shadows 2 (1 doesn't hit very hard) are kind of tough on the system. You can always make a backup of your q3config.cfg, and disable ALL beautiful stuff, and see how much FPS you can max out at - then you know what tweaking can optimally do, and how beauty vs speed adjustments pay off. The quickest way to glory is ofcourse a reduction of the resolution..

Ofcourse, I wouldn't ever support turning Gen ugly - maybe someone should sponsor Kingu, and get him a brand new system ;]

Changing dlighttype to 0 or 1 usually helps with dynamic light saturation, but at dlightlevel 1 it shouldn't be a problem.  Are you running FSAA or Anisotropic Filtering?  Check your video drivers and see if either are enabled.  Those are framerate killers on all but the most powerful video cards.

I run shadows 2 but with other i was teleporting too, but i'll try it next game.
 And Phoenix,  Anisotropic Filtering is set to program selectable, and if i correctly (translating) understand FSAA  so that is set like that filtering.

There is teleporting, and there is teleporting. If you run at 20 fps on average, you'll be impossible to hit as oppossed to being simply moderately annoyingly lagged at 40+ fps (or thereabouts). Tweaking will only yield big results if done on a bigger scale and tested per setting (cg_drawfps 1 ;))

I usualy have 40 to 70 fps,  average
 Maby some times with close napalm it goez lower  but that is averege.
 I use com_maxfps 85 becose of monitor refresh. | :

Hmm, then apparantly the FPS levels for lag problems are different than I thought they were. It may be worth trying to see what happens if you restrict your FPS with a com_maxfps 60, and see what that does? I'm quite interested in the effects and causes here, since anything we find out can prove useful for our FAQ :]

A few additional questions::

1)  Do you have Compressed Textures enabled?  Turning that on is a tremendous help and is always recommended.
2)  What is your network connection?  Dial-up, Cable, DSL?
3)  What are your /rate and /cl_maxpackets settings?
4)  What does your netgraph typically look like in-game?
5)  Do you leave other programs running in the background that might communicate over the net?

Those might help give us a picture of your setup.  If your FPS is holding steady and not dipping down when you get lag spikes, then it's not an FPS related issue, but something network-related.  Maybe we can help iron it out.

My monitor refresh rate is 100hz and my FPS is right on 100. May it be that?

That's pretty good Washu - it probably means your system can handle more, but is limited at 100 (either through com_maxfps or r_swapinterval?) - 100. 125 is considered ideal, 80 is considered acceptable, and I don't think anything stable above the latter will give you serious problems.

I haven't tried any higher yet.

Okay I've did some adjustments FPS. Right now, beleive it or not the game's stable FPS online is 200!  :wtf:
Single player is over 300! It seems like what ever I set it to, it just wants to keep increasing. I'm pleased with that.

Remember that you should probably never use anything above com_maxfps 125 for online play. Higher will clog the connection thanks to id's wonderful code ;]

1 I have it on
2 Cable 786 kb/s dl and 128 kb/s up
3 Rate 20000 Cl_maxpackets 125
4 Like in atachment, that  yellow and red spice are showing wiht 10 sec frequency.
5 Actualy i have communicator runing in background, but when i had if off, Tab didn't sait that i teleport less.
 That communicator once in a wile writes some things of h.drive, so than i have drops in framerate, but that is not helping me while playing either.

 I have to check that  averege fps again, becose i'm not sure them now.
 Does q3 is more optimised for Geforce cards than Radeons ?
 Earlier i had GForce4 mx and it looks like new card didn't increased performance wery much.

Your rate and cl_maxpackets look like Tab's settings.  They're great if you have a LAN connect or one very clean connect to the server, but you might try lowering the rate to 15,000 or 10,000, and try cutting the cl_maxpackets in half.  I don't see anything that indicates your FPS is to blame.  It sounds like your network is overloaded with packets.  Both rate and cl_maxpackets load your upstream.  If set too low (like 3000 rate and cl_maxpackets 20) you'll have constant chop, if set too high, you can get very choppy or even freeze up for a moment when there's a lot of activity, especially if you're issuing a lot of commands to the server, like firing a chaingun while jumping and dodging.

I've tried higher rate and cl_maxpacket values without success.  Typically I use a rate of 10,000 and I use the default cl_maxpackets of 30 because my upstream is limited to around 35 - 45KB/s.  Try experimenting with those values, start at half of what you're using now, and see if it improves any.  Once you get a stable game, then see if you can run with higher settings.

What would be appropiate settings for maximizing a 56k connection, I wonder. >_>

That information I have in the FAQ.  I used to have a 26.6 connection for a very long time on which Quake 2 was completely unplayable, so I learned what the "worst case settings" are.

Here's where Pho and I are totally in disagreement I guess.
As long as you ping lower than 80 to a server, and your bandwidth exceeds that of 56k dialup cl_maxpackets 125 is nearly always optimal. Rate 10000 is rather on the low end of the spectrum, 25000 is recommended by pretty much all the professional quake players (and tweakers) for (Dutch) ADSL and upwards. Thus, a stable connection with 256+ kbps upload and a half-decent link to the server would be underused with anything below maxpackets 100+ / rate 20000+.

You can easily use the lagometer to test if your settings are killing the connection. A lot of yellow triangles (mind, something is wrong with the server or link to the Euro server causing hitches, so don't confuse that with a client problem!) indicate that something is amiss, and the easiest way to test this is to change your settings to the bare mininum and see if that makes them disappear. Then gradually increase the values (synchronizing them), until you see yellow or red spikes in your lagometer. If the settings don't show any such effects, you can safely use them.
Using settings on the lower end of the spectrum will always have the effect of making the game more choppy, decreasing the window in which you can hit as well as in which you can be hit, generally turning the game into more of a gamble-run. It may be my personal opinion, but I think one should generally try to avoid that if at all possible.

Futhermore, low FPS (or a strongly, rapidly varying amount of FPS) causes hitches and chops. I know this, but not the specifics - I have to rely on more fervent tweakers for details and hints about how the net system works as disclosed by those in the know.
It also makes sense when you keep in mind that the net load is affected by your FPS, and that local prediction affects the game aswell - a higher FPS rate demands more data, and then receives the data during a time when less FPS can be drawn, causing the system to get clogged. Next, a lower rate of data is 'negotiated', after which the local renderer suddenly has a lack of network information to display its frames correctly. At such times you will see great shards of yellow (or even red) spikes in your lagometer.

Concerning Kingu's connection, I know that his choppy apearance on the server has remained unchanged. He teleported all over the place during the times he used the bare mininum default settings in Quake 3, aswell as after he tried the settings I recommended, and possibly values in between - if he tried them. Since his latency is low and doesn't spike badly, the only non internet-routing related cause I can think of is FPS-problems, especially since it is known that he does 'suffer' from a low and strongly varying framerate.


I think I'm going to run some tests with Alucard later, loading CPU-limiters to drop my FPS and make him record demos me doing of a repeated run through a long hallway. That could probably clear some things up as to what might cause gappy appearance in-game. And, hopefully, what best to do about it.

Disagreement indeed, on some points.  Yes, FPS affects outgoing network traffic because it affects the client's command rate to the server.  STABILITY is what is most important, not just raw speed.  An FPS of 60 that is STEADY is preferrable to a maxFPS setting of 125 that dips down to 40.  Also, setting a com_maxfps setting absurdely high, like 300 (even if your system can handle it) means you're going to be dumping more data down the pipe than is sane.  Remember that each packet sent to the server has to handle all the client commands dumped into it.

Cl_maxpackets affects outgoing network traffic.  Cl_maxpackets is how many times per second the client sends commands to the server.  If set to 125 (the max), it will send commands every 8 milliseconds, provided the framerate keeps up.  A lot of commands sent to the server are already duplicate commands owing to how the Pmove module works on the client.  Unless you have a PERFECT 125 FPS, you're just going to be dumping duplicate commands down the pipe, which is wasting network traffic, so optimally, cl_maxpackets should match your highest STABLE FPS provided your upstream handles it, but should not exceed it.

Rate is outgoing from SERVER, not client, so it affects incoming network traffic on the client.  Sorry, I'm used to people lagging idlebot and I mixed it up earlier.  This affects downstream, not upstream.  Keep in mind it makes no difference whatsoever if the server's sv_maxpackets is set lower than your rate setting.

This is the most important part, something everyone seems to forget.  Outside of having a crappy ISP, network conditions between you and the server are the biggest bottleneck factor in gaming.  If you have 15 hops to get to a server, that means 15 different computers that handle your data.  That means 15 chances for packets to get lost, damaged, or piled up.  I will never, ever recommend running extremely aggressive network settings unless you're sitting right next door to the server for this reason.  Not everyone has the privelege of Dutch ADSL or upstreams of 256KB/s or higher.  If I make network recommendations, it's on the following assumptions:

1)  The upstream is limited by the ISP
2)  The downstream is limited by hops between the server and client
3)  There are bottlenecks and problems in the network SOMEWHERE
4)  There's software running in the background that utilize network traffic to some degree

This is why I suggest people start with conservative settings and work up to what is stable.  Running a rate of 20,000 and cl_maxpackets of 125 may be great if the network is clean, but more often than not, it's not clean.  I don't care what "pros" and "tweakers" say if it's not in line with realistic network conditions.  What's good for a LAN is NOT necessarily good for the net, and I'd like to think Id kind of knew what they were doing when they developed certain network settings for Q3 based on connection type.   Cl_maxpackets 125 wasn't even an option until 2002 according to the source log.  

Now as for personal experience, I tried cl_maxpackets 125 to Central, and even at a steady 125FPS my shots felt sluggish, I constantly missed, and my movements felt delayed.  Compare that to cl_maxpackets of 30, and I had a much cleaner game.  I tried 125 on Euro, and I experienced actual hitches during play.  This last game on Euro (thursday), I did not experience hitches or freezes, just the usual delay associated with ping.  This is because I DO have a very limited upstream.  It's not just me, either, I don't know how many times I've had people complain of lag or spikes and I ask them "what's your rate?" and they say "25,000".  I say "try lowering that to 10,000" and viola, problem fixed, no more lag, spikes, or chop.

Tab, I realize you want everyone to have a "perfect clean game" in which everyone never skips, chops, or lags.  Believe me, I'd like it too, but that's just not a reality outside of a LAN with all identical machines.  I'm not saying everyone must use certain network settings because Pho says so, all I'm saying is that if you have problems, ease back and test it out, and work with it until you find out what works best for you.  Everyone's connect and hardware is different.  There is no "one-size-fits-all" connection solution.  I just like to give a "here's something I know works for everyone" value as a reference point.

That spikes yelow and red, are showing no mater what rate and maxpockets are.
 And from my side game runs normaly i don't see any choppy view, other than the others see.
 
On todays game i'll limit to 60 fps and rate, max_packets that phoenix sugested.
 Propably some help would be more ram but last time i tried to add they didn't worked with my mainboard.

I fully agree that my personal preferred settings will only work for relatively clean and fast connections - I wouldn't recommend someone with, say, 64 kbps upstream to use high settings, but neither would I recommend settings that are well below what is better all-round, shying away from nerdish tweakage :]

The opinions of "experts" (as you scare quote it) does reflect on ordinary net connections aswell, and I do try to keep that in account. Of course there's always the handful of dorks that can't resist making bold claims like "playing with(out) mouse acceleration is the only way to win!" etc. - those aren't the people whose word I take seriously.

I simply have the idea that a good deal of the players (Euro, at least) have a fairly decent connection, and are generally underusing it in an almost criminal manner. They can definitely make some improvements, or at least learn a bit and play around with the settings discussed. It seems like a waste to me to say: "just stick to worst case scenario values" instead of "try and see what can be improved".
I'm also playing the role of devil's advocate, as usual, so the optimum will undoubtedly lie in the middle. If you're running at 125, try 60 and see if it works for you, if you're running at 30, try 60 too. After that try the other extreme and do some educated guesstesting ;]

The problem I encounter most often is when someone finds something that works for them they often assume it automatically works for everyone else.  Not saying that's what you're doing, but there's also the flip-side, which I do worry about people doing; that is, when someone reads what works great for someone else, they assume it will work for them just as well - especially when the person who recommends it has a reputation of pounding the snot out of everyone else on the server on a regular basis.  I just try to take that human tendency of "I want to be like [insert name here] so I'll do what he/she does" into consideration.

Ah ok, hopefully I didnt reignite a strife between you 2 developers. I did alot of guessing on some settings and finally got my ping to about 200-220. Thanks for the help, BOTH of you. XD

Oh, no worries there.  Tab and I can be very opinionated and still work together just fine.