Source Stutter revisited

[Groo]

Hey all (and I mean you Lafinass =)

I have been experiencing the "Source Stutter" issue off but mostly on for the last month now. I have already gone through the Source Stutter thread and have an idea what is causing it, but not a clue to why. So here are the specifics.

workstation: QuadCore2.4, 2 gigs ram, 8600GT, 400gig 7200 SATA HD, gig nic on board (and I've tried a separate pci gig nic card).

I notice the slowdown specifically when there are over 16 people in any of the eGO servers. I have tried disabling all but one of the cores, i have tried turning all video settings to low/disabled. Video drivers have been recently updated.

I thought i solved the problem when I did a system restore, everything was fixed. However it came back a week later and a system restore did not fix the problem that time =(.

What's happening in my stuttering is I appear to be moving around, but everything and everyone else around me pauses for about 1 second. So the rockets and players, etc are paused in midair while I continue to move. It will unfreeze for a few seconds and then freeze again for a second. When the stutter is really bad, i freeze too. This continues for the rest of the game or if the number of people drop down to less than 12 or so.

I have run pingplotter and I have a particular router (144.223.244.162) which has 20-30% packetloss (all pings are very acceptable). The part that is very odd is that my co-worker here shows 0% packetloss on that same router. Our machines are on the same LAN and are both going through the same cable modem.

Anybody have any clues why my computer and not my co-worker's would show 30% packetloss for one router?
Any help is appreciated.

Oh yeah, and I have tested other servers and they do not have the stutter. Only eGO's =(

 

[Lafinass]

Are you both using the same settings on pingplotter? Make sure you're both running with 'Trace delay' at 1 second and 'samples to include' at a very large number. The PL% becomes more relevant the more frequently the samples are taken and the more samples are included. It's possible that if he's running the default the router is spiking and just happens to be spiking between the default 15 seconds.

Run a sample at 1 second tracing with 500 samples total and put it up here (either in text or with a screenshot). I'd like to see your whole path. I'm not getting a significant packet loss to that IP, but pulling a pretty high average ping.

 

[Groo]

Well,

Things are a little goofy right now with the update and nobody presently on the eGO servers, but her is my screenshot.

With 1 second rate, it is hovering at 10% right now for about 5mins.

I had my co-worker run it too, and with the 1 second, 500 sample, he was just about the same as well. Originally we were both on the 15 seconds, 10 sample default settings and both ran it for 10 minutes with him showing 0 packetloss and me at 30%. That might be what you were describing.

I might have to wait until tomorrow to do more testing since the update was just released.

 

[Lafinass]

Yeah, with a slower sample rate and a lower total sample it makes a percentage calculation like that a crap shoot. Essentially, since it's not 100% packet loss rate it must be letting some packets through and with a poor sample segment he basically pinged it every time it was working fine and you didn't.

A 9-10% packet loss rate isn't really high enough to be concerned with and since it is likely only a small fraction of the traffic passing through that router is gaming load, the state of our server shouldn't affect your result. Also, your ping looks solid across the board.

You said you were testing this with a co-worker? You should try running pingplotter on the machine you use for gaming. I'd like to rule out any connection issue between the gaming rig and the rest of the hops before we start looking for other problems. Unfortunately, Source responds to almost all problems with a similar stutter- partly because of the lag compensation mechanisms it uses to try and provide smooth gameplay for everyone.

Lets assume that on your game rig you get a similarly clean path...

You say the stutter seems to occur when a certain number of people are on the server. Does this stutter occur if you're away from combat in an unpopulated section of the map?

 

[Lafinass]

Also, can you run net_graph in game? Let me know what happens when these spikes happen? What happens to the FPS, PING, IN, OUT, and the LERP?

Or just gimme a screenshot of net_graph 4 when this is occuring and one where it isn't.

 

[Groo]

pingplotter was done on two workstations: my co-worker's TF2 game machine and my game machine with the above listed discrepancy in results. the graph is from my game machine.

The stutter occurs worse with more people, but is present with just a handful on screen, again once there are about 12-16 people on a server.

i don't have tf2 up right now, net_graph is that just run from console?

Thanks again for your help on this lafinass

 

[Lafinass]

Yeah, just type 'net_graph 4' in the console, without the quotes. If it's network related, it should be pretty apparent from the display that will run on your screen. If not, it could possibly be something entirely client side. But I always start with network first.

And no problem, you people are more appreciative than the people I'm being paid to help.

Bloody law students.

 

[Lafinass]

Also, I've determined that the following is (at least for me) the best setup.

rate 25000
cl_updaterate "60"
cl_cmdrate "66"

Gives me 0 choke 99%.

 

[Groo]

Well, here are the screen shots, but after the update last night, I haven't been able to duplicate the problem. I played on multiple servers last night each with over 26 players without the usual stutter. I was just experiencing the occasional stutter. Which, depending on what school of thought you come from, is either good or bad. But I'll post them anyways in order to at least document what i am currently experiencing in case the problem returns.

Here's hoping this thread dies now =)

 

[Lafinass]

Glad the problem resolved itself. However, I can tell from that your net code could use a tweak. A choke of 50-60 is horrible.

See those four numbers on the right? 20, 19, 31.1, 30? Those are your cl_updaterate, actual update rate, actual cmdrate, and cl_cmdrate. I've found that our servers clamp a maximum that varies, but all much higher than what you have. I believe 20 and 30 are default rates, which probably means all of your rates are default. To change these...

Create, or edit if it exists, autoexec.cfg in ..\SteamApps\<USER>\team fortress 2\tf\cfg and add the following lines...

rate 25000
cl_updaterate "60"
cl_cmdrate "66"

I've done some trial and error and those work best for me on our servers. But results may vary. So if you do this and your choke is still high, you can adjust these values on the fly from console. Lower choke generally makes for smoother gameplay.

 

[Groo]

Thanks Lafinass for the information. I'll try adjusting those and keep you posted (I sure hope that is this issue and it never ever comes back as it is really hard to play tf2 with 1-2 second pauses every 4 seconds. Now, go back and kick those law students back in line! (and don't have too much fun while you are doing that =)
-Groo

 

[GTI-Steve]

also look at my post in the original source studder though u might have already done this