It's not limiting my FPS when I stream. I have an NVidia card with VSync forced off in my drivers. In fact, I actually can't force it on if I try for windowed applications.

EDIT: I just read your thread and I think you are confused. The "FPS" option you set in OBS is not your desired in-game FPS, and it has no bearing on your in-game FPS. It is the FPS that your stream will run at.

Try it with Aero off. I've heard that if you have an ATI card, Aero results in forced VSync.

I got some more data points and didn't have a way to elegantly put them in the original thread (character limit exceeded). New thread ¯\_(ツ)_/¯

placeholder

That's what "Software Capture" is.

When FFsplit has more features than a rock, sure. As it is now I wouldn't use it, so I don't care about testing it.

OBS can do almost anything Xsplit can, and the things it can't do are planned features for future updates.

I believe you can specify a custom RTMP url, so yeah. (Though TBH I think TCP relay is kind of bogus)
Also: http://obsproject.com/forum/viewtopic.php?f=5&t=56

I only tested it for a little bit, but using OBS's screen capture did seem to alleviate the input lag problems I was having. I still need to play with it more though and see how it feels in a 6v6.

As far as I can tell this input lag is a strange problem specific to me, so you guys shouldn't worry about it.

EDIT: I've been told that reducing the "Send Buffer Size" to a low value (such as 4096) can eliminate input lag in some cases. If you have input lag please try this fix.

There are three main ways you can capture video for OBS:
1) Raw screen capture, like the "screen regions" in XSplit everyone knows and loves.
2) Window Capture WITH AERO ENABLED (!!) will ONLY capture the specified window, and no windows layered on top of it.
3) Dxtory

There is no reason that OBS shouldn't be satisfactory for professional casts, though at this moment in time XSplit may be slightly better suited for it. However, I foresee that changing quickly.

What I learned:
1) OBS has overall higher performance than Xsplit.
2) Turning Aero off isn't as great as people think, and in fact OBS has significantly increased performance when using screen regions if Aero is on.
3) OBS's built-in Window Capture is more efficient than Dxtory.
NOTE: OBS can function EXACTLY like Dxtory (only showing your game window) if you play in a window, use "Window Capture", and have Aero enabled.

Caveats:[/b]
These results are specific to my computer and my setup. Your results may vary.
TF2 was the only game tested. Other games may perform better or worse in these configurations.
Window Capture and Monitor Capture in OBS are NOT the same thing, and can behave very differently. It is NOT recommended to use Aero if using Monitor Capture.
Even if using Aero results in higher performance, it may also result in higher input lag. Use caution.

Conclusion:
The data speaks for itself, OBS w/ Software Capture & Aero enabled has the best performance. I am going to stop using Dxtory to capture TF2, and stop using Xsplit altogether. OBS is only in an early beta phase, and it is already a superior program. OBS, in my experience, also has better image quality than Xsplit at similar bitrates. After further testing, OBS and XSplit appear to have nearly identical quality

Other tests:
There was a rumor that streaming at 29.97 fps would result in significantly higher performance than streaming at 30 fps. In my tests, this was not the case. Compare the result below to the "Xsplit w/ Dxtory & No Aero" result above.

Xsplit w/ Dxtory & No Aero (29.97 FPS)
AVERAGE: 41486 frames 211.896 seconds 195.81 fps ( 5.11 ms/f) 49.1 fps variability

Image Quality:
It's really hard to tell XSplit and OBS apart in terms of IQ.

Comparison 1: OBS | XSplit
Comparison 2: OBS | XSplit

If you would like to conduct your own tests and compare them against mine, here is the demo file I used for the timedemos: https://dl.dropbox.com/u/6089084/Research/the_timedemo.dem

I wanted to find out what software configuration would yield the highest performance when streaming. So I did.

Preface:
OpenBroadcasterSoftware (OBS) is beta software that has only recently been developed. It is a free and open source alternative to Xsplit. It can be downloaded @ http://obsproject.com/
Aero is the desktop composition module in Windows 7 responsible for the fancy "glass" window borders and taskbar. Most streaming programs and guides reccomend you to turn it off. However, my tests don't completely agree.

Test settings:
Stream resolution of 960x600 and a framerate of 30.
1700 video bitrate, 10 quality, "veryfast" preset, 128 audio bitrate (AAC)
TF2 running in a native resolution noborder window (1680x1050).

Test procedure:
Open programs needed for test.
Open TF2.
Begin streaming.
Run three "timedemo"s in TF2 to benchmark performance. "Timedemo" is a command that lets you play a demo, and afterward the game prints a line with some performance statistics to the console. A "timedemo" always plays with a host_framerate of 66. The same demo file was used for every test.

Data set:
Each scenario was run three times (more if anomalies were observed), and the results averaged.

Results:
Sorted by performance (highest at top)

Control (Not streaming, Aero off)
AVERAGE: 41486 frames 144.422 seconds 287.26 fps ( 3.48 ms/f) 31.2 fps variability

Control (Not streaming, Aero on)
AVERAGE: 41486 frames 156.161 seconds 265.66 fps ( 3.76 ms/f) 36.1 fps variability

OBS Window Capture & Aero
AVERAGE: 41486 frames 177.971 seconds 233.11 fps ( 4.29 ms/f) 40.5 fps variability

OBS w/ Dxtory & No Aero
AVERAGE: 41486 frames 188.321 seconds 220.29 fps ( 4.54 ms/f) 45.7 fps variability

OBS w/ Dxtory & Aero
AVERAGE: 41486 frames 193.119 seconds 214.83 fps ( 4.66 ms/f) 42.9 fps variability

XSplit w/ Dxtory & Aero
AVERAGE: 41486 frames 212.022 seconds 195.68 fps ( 5.11 ms/f) 42.3 fps variability

Xsplit w/ Dxtory & No Aero
AVERAGE: 41486 frames 214.834 seconds 193.12 fps ( 5.18 ms/f) 42.9 fps variability

OBS Window Capture & No Aero
AVERAGE: 41486 frames 220.966 seconds 187.75 fps ( 5.32 ms/f) 49.3 fps variability

XSplit Screen region & Aero
AVERAGE: 41486 frames 252.391 seconds 164.38 fps ( 6.08 ms/f) 46.2 fps variability

XSplit Screen region & No Aero
AVERAGE: 41486 frames 257.593 seconds 161.05 fps ( 6.21 ms/f) 60.4 fps variability

http://www.hidemyass.com/ is supposed to be one of the best.

-dxlevel 81

you will notice it

and it is funny

http://www.youtube.com/watch?v=cnCRhpc5xe4

That's not true. Some servers do have mods (like HLStatsX and shit) that will broadcast your IP to chat though.

The only way that I know for someone to get your IP is for them to have rcon access to the server.

its like you scotch taped brown asparagus to your temples

http://i.imgur.com/pJffQ.png

I'm wary of having TF.TV host their own game servers.

I don't think it actually accomplishes all that much for the amount of effort put into them, and frequently these kind of endeavors don't really work out. How many times have you guys actually gone to Skial.com or NewbsTF2.com just because you played on their servers?

MGE hosted many of their own servers, and the MGEMod servers were the only ones to fill up (and that was only because they had 'the' stats database back before people hosted their own). eXtv has several of their own servers, yet only the DM servers fill up regularly (and this is because they were sort of inherited from AG, and at one point those were the 'official' DM servers back before tons of people started hosting their own).

That said, this could be done with little monetary risk and I would not mind setting up the server(s) if we got them.

I'm always down to pound, John.