EPTWould anyone be able to troubleshoot/help me with this?
[G] Streaming with OBS (Open Broadcaster Software) - Page 55
| Forum Index > Tech Support |
|
Poo
Canada536 Posts
Would anyone be able to troubleshoot/help me with this? | ||
|
ZeroTalent
United States297 Posts
On January 26 2013 19:10 Tehkilla wrote: Hello! Can anyone help me figure out the best stream settings for me? i7 2600k @ 4.2 ghz MSI Geforce 560 Ti Twin Frozr OC ASUS P something Deluxe Motherboard 8192 MB of Corsair Vengeance Blue Windows 7 Home Premium 64-bit Internet speed: ![[image loading]](http://www.speedtest.net/result/2463451486.png) I don't really have any current settings. Max out everything ^^. I have a nearly identical setup (CPU is at 4.3 Ghz, GPU is at MSI factory speed, different mobo) and stream at OBS with the following settings:
This is on the edge of playable (my goal is > 30 game FPS I'm okay if it goes below it sometimes), and max battles can get rough even though I play T (Terran is the least CPU-intensive race for streaming). Keep in mind I am going for stream quality over gameplay, so if you don't like the input lag drop down to faster or veryfast preset. | ||
|
Cyro
United Kingdom20326 Posts
Terran is the least CPU-intensive race for streaming o.0? Also you shouldnt see much if any difference from changing presets in game performance unless your CPU usage was too high before | ||
|
felisconcolori
United States6168 Posts
Also, OBS just keeps getting better and better... you guys are the best. Really. | ||
|
Genesis Brood
United States193 Posts
| ||
|
ZeroTalent
United States297 Posts
On January 28 2013 07:22 Cyro wrote: o.0? Also you shouldnt see much if any difference from changing presets in game performance unless your CPU usage was too high before I don't have super hard numbers -- I'm just looking at the game FPS when I run various tests. And it stays higher in TvT than in any other matchup. I assume that what's going on is that since marines and tanks don't shoot visible projectiles, and creep is annoyingly hard to render either on the screen, in the encoder, or both, it's just easiest to play while streaming if you're terran. Also for the same reason, the video quality stays higher in TvT, holding all other streaming factors constant. Just less stuff on the screen. At 3000 Kbps, going from veryfast to fast (skipping faster) raises SSIM of my TvT test footage from .896 to .912, which is a 17% improvement according to the formula in this thread. The ZvP max battle test goes from .858 to .873, which is a 11% improvement. For veryfast->medium the improvements are 23% and 17%, respectivelly. So, nothing fantastic but nothing to sneeze at either. I don't think it's anything viewers would consciously notice, but it would increase the subconscious ooh, ahh factor and he certainly has the resources to do give it a shot. | ||
|
Cyro
United Kingdom20326 Posts
Those numbers are a bit higher than what ive seen, and you also have to remember the extreme changes in CPU requirements between presets ![[image loading]](http://4.bp.blogspot.com/_kmp1khwwjgM/TQk9Aafv5mI/AAAAAAAAACA/Uhb_tgOKlgE/s1600/ssim_chart.png) Maybe not the best data, but trend for red line seems to be atleast somewhat correct. What resolution are you streaming at? On a 2500k/3570k at stock, you dont really have any margins to comfortably drop presets at 720p60/1080p30 standard, its really adding HT and overclocking or notably lowering resolution that gives you those margins. In these days its not really uncommon to overshoot on CPU settings and be hitting CPU limits occasionally or even pretty often on 1 or more cores (which you can do even if you are using like 50-60% CPU on average) | ||
|
ZeroTalent
United States297 Posts
On January 29 2013 12:53 Cyro wrote: What resolution are you streaming at? On a 2500k/3570k at stock, you dont really have any margins to comfortably drop presets at 720p60/1080p30 standard, its really adding HT and overclocking or notably lowering resolution that gives you those margins. In these days its not really uncommon to overshoot on CPU settings and be hitting CPU limits occasionally or even pretty often on 1 or more cores (which you can do even if you are using like 50-60% CPU on average) I (and PP) both have OC'd 2600Ks. I'm streaming at 1080p30. Sorry, what's the red line in the chart, game FPS or encoding FPS? And what's the bitrate or vbv-maxrate for these tests? And source video, if you've got it ... | ||
|
Cyro
United Kingdom20326 Posts
I doubt you have stable 30fps at that resolution at medium or even fast 100% of the time (though you probably wouldnt notice, unless stuttering and missing frames got really bad) with a CPU like the 2600k (thats the kind of workload you would put an an OC'd 3930k) The same as putting something like 1080p60@faster or whatever that ive seen people do on random stock sandy/ivy bridge CPU's, the output is usually very suboptimal and game performance is probably impacted by CPU maxing out in worst case but everything runs fine | ||
|
ZeroTalent
United States297 Posts
On January 29 2013 13:57 Cyro wrote: I doubt you have stable 30fps at that resolution at medium or even fast 100% of the time (though you probably wouldnt notice, unless stuttering and missing frames got really bad) with a CPU like the 2600k (thats the kind of workload you would put an an OC'd 3930k) ... the output is usually very suboptimal and game performance is probably impacted by CPU maxing out in worst case but everything runs fine I certainly don't have stable 30fps at medium (I think at 25 fps it stays stable until both armies are at about 100 supply). Fast is, like you said, fine for almost all gameplay but you can start to feel it in really high supply (180+) battles. | ||
|
ZeroTalent
United States297 Posts
On January 29 2013 12:53 Cyro wrote: I meant difference in game framerate from presets Those numbers are a bit higher than what ive seen, and you also have to remember the extreme changes in CPU requirements between presets Maybe not the best data, but trend for red line seems to be atleast somewhat correct. More data is always nice, thanks for posting this ... I found the original blog post too. Since he's 2-pass encoding and has no VBV constraint, I'm not surprised that he had much smaller SSIM differences as the preset changes. It's a very different scenario from streaming; 2-pass encoding makes it way easier for the encoder to allocate bits efficiently. He's also got a fairly high bitrate for 720p content, blunting the impact of presets. At 3000 Kbps CRF 27, these are the encoding FPS I got before overclocking with 1-pass CRF encoding for a max battle ZvP medium 34.8 fast 37.2 faster 39.5 veryfast 42.1 I didn't run the full test suite after overclocking but on the tests where I have an apples to apples comparison, the encoding FPS was 15-20% faster. So "fast" encoding FPS should be in the 42-44 FPS range, up where "veryfast" was before OCing, plus/minus weird threading interactions while playing. I can still feel it in max battles, but just barely ... if I were pro I'd feel it more probably. | ||
|
Cyro
United Kingdom20326 Posts
First test i did of a video (960x540) i got 85fps at Medium, 150 at Veryfast - but the CPU is barely at 50% load on Veryfast while being 100% on the Medium test, so i have CPU utilization issue somewhere + Show Spoiler + More data is always nice, thanks for posting this ... I found the original blog post too. Since he's 2-pass encoding and has no VBV constraint, I'm not surprised that he had much smaller SSIM differences as the preset changes. It's a very different scenario from streaming; 2-pass encoding makes it way easier for the encoder to allocate bits efficiently. He's also got a fairly high bitrate for 720p content, blunting the impact of presets. This is why i said red line only, the SSIM in the chart isnt really relevantI dont think you should be able to feel a preset change in game performance, i havent seen that unless you are going too far with CPU settings | ||
|
yokohama
United States1116 Posts
![[image loading]](http://www.speedtest.net/result/2470485521.png) I've also tested pingtest.net and I have 0 packet loss. Even in offline games and no other devices in my apartment using my network I still start dropping frames and get a red box on OBS when I go over 1500kbps. I recently upgraded to this 2pc stream setup so I could start a very high quality stream and the fact that this is happening is very disheartening. I really want to have a nice 3000kbps+ 1080p 30fps stream. In addition, when I do play online games such as Battlefield 3 when I try to stream, I get extreme cases of rubberbanding even at this bitrate. Details in OBS are as follows: Quality 10 Bitrate/Buffer: 1500 kbps FMS URL: Singapore (I've tried every server and nothing gets better) Base Resolution:1920x1080 Res. Downscale: 1280x720 FPS: 30 CPU Preset: Medium (I've tried all and no performance change, I have an i7-3770k @ 4.7Ghz http://www.twitch.tv/esportsjapan/b/361908792 I've uploaded a video so you can see that the quality of my stream isn't near what I want, and it still has an annoying stutter to it. Any help is appreciated. | ||
|
R1CH
Netherlands10342 Posts
| ||
|
yokohama
United States1116 Posts
On January 29 2013 22:22 R1CH wrote: Do a speed test to somewhere realistic like California, Twitch has no ingests in Japan. ![[image loading]](http://www.speedtest.net/result/2470696558.png) Good call. Though shouldn't my ingest point be Singapore, since that is the server I had picked? ![[image loading]](http://www.speedtest.net/result/2470702706.png) A quick question on top of that, how can pros in Korea stream so well using twitch.tv, wouldn't they also be ingesting to California? | ||
|
Az0r_au
Australia385 Posts
I had to set it up so the Live Gamer was my 2nd monitor (video card > HDMI Cable > LiveGamer) and set my gaming monitor as Primary (videocard > DVI cable > Monitor) Because I wasn't getting a signal using the passthru method (Video card > HDMI cable > Live gamer > HDMI cable > Monitor) So I currently have the Live gamer set to mirror my gaming monitor under the Windows resolution screen. | ||
|
amd098
Korea (North)1366 Posts
also, i get about a 10 or so second delay in what im doing and what i see, anything i can change to fix that? im on a i7 2670qm / 8g ram / gf555m | ||
|
felisconcolori
United States6168 Posts
On January 30 2013 06:34 amd098 wrote: so i just tried to dl the installer and i got a trojan alert, is that normal? the separate binaries were fine, just the installer came up as a trojan with comodo firewall also, i get about a 10 or so second delay in what im doing and what i see, anything i can change to fix that? im on a i7 2670qm / 8g ram / gf555m When I'm streaming, I notice there is a delay between my screen and the stream. I'm pretty sure that this is a combination of network latency (between me and Twitch and back to me again), and the various processing that I and/or Twitch are doing. It shouldn't really be much of an issue in practice, though, because I can't think of anyone I've seen streaming that doesn't have some kind of delay in the stream, either purposely (in the case of partners) or for the reasons above (everyone else). Makes stream sniping/cheating just a little more difficult. (Also, nice raw specs. I'm jelly.) | ||
|
RemarK
United States452 Posts
Thanks in advance  | ||
|
ZeroTalent
United States297 Posts
On January 30 2013 06:34 amd098 wrote: also, i get about a 10 or so second delay in what im doing and what i see, anything i can change to fix that? (1) If you are streaming to twitch, use JTVPing to pick the best ingest location, instead of using the Global Loadbalancing Service. (2) Make sure Settings->Advance->Network->Use Send Buffer is unchecked, though this may result in more frequent dropped frames. | ||
| ||
