|
Isn't the issue when streaming the overhead from game capture and then encoding (also network resources)? If the desktop outputs video as if it were to the monitor and something else captures it, then there should be no overhead for the desktop.
Looks like the elgato does hardware h.264 encoding (up to 1080p30 with 1080p60 input). That's what it says, and it's got to, considering the interface back to the comp is USB2. If you're willing to put up with presumably the lower quality/bitrate, that should work, and there really shouldn't be much left for the laptop to do, computationally. I mean, I don't think handling the broadcast software and overlays is all that much, but I've never tried that myself. (not enough upload bitrate or interest)
I think the linked computer is out of stock. This for $450 has a normal-voltage i3 (also last gen...), so it's faster anyway, though I don't think that really matters:
http://www.toshiba.com/us/computers/laptops/satellite/C50/C50-AST2NX2?src=MAKX&cm_mmc=Affiliates-_-Linkshare&siteID=dZCX6Je2w8Q-pnl2YwMR3D4ARJAyCo5LrQ
|
United Kingdom20326 Posts
Isn't the issue when streaming the overhead from game capture and then encoding (also network resources)? If the desktop outputs video as if it were to the monitor and something else captures it, then there should be no overhead for the desktop.
It seems intuitive like that, but destiny reported enough of a performance hit from using avermedia live gamer HD that it performed worse than OBS game capture.. really not sure what's going on. Basically >nobody< tests this and i don't have the resources to.
Hardware h264 encoding on every card that i've seen is significantly worse than x264 quality, as in you'd need to upload ~4mbits to match the quality of somebody CPU encoding @2mbit*. That's how the Live Gamer HD was, anyway. Nvidia's NVENC encoder on GPU's seems a lot better (still worse than x264, but it seems to be worse by a little bit, not by 2x, especially compared to veryfast preset streaming), but i don't know if there's an equivalent in capture card hardware. To be honest, i'd go so far as to say.. buy an Nvidia GPU, use OBS with NVENC lol**. Maxwell NVENC is twice as fast as Kepler's, i'm not sure if gm107 has that or if it's even at all relevant, but a 750ti would kick ass and cost far less than one of those laptops and a capture card.
*You mentioned that, but it's quite an extreme gap
**I didn't benchmark it to see if there's no real performance hit like using just Shadowplay to stream. I'l jump in and try that quickly~
+ Show Spoiler +Ok here we go, all tests on 4770k@4.5ghz, 770, game med settings (mainly for shaders) to enable high models to enable physics etc so that i could max CPU settings. Baseline, 20fps min, 50.75 avg, 80 max: ![[image loading]](http://i.imgur.com/M0Ls1g5.png) OBS gamesource+x264, set to 1080p, 48fps, 5mbit cbr - 15 min, 40 avg, 64 max: ![[image loading]](http://i.imgur.com/EyQXa0x.png) OBS gamesource+NVENC set to 1080p, 48fps, 5mbit cbr - 18 min, 48.25 avg, 79 max: ![[image loading]](http://i.imgur.com/WGRBtpi.png) Shadowplay set to 1080p60, 10mbit (this isn't the best option, some driver issues, still beta, limited control etc especially for streaming) - 18 min, 50 avg, 80 max: ![[image loading]](http://i.imgur.com/Y0bUFVr.png)
![[image loading]](http://i.imgur.com/4RoNxlW.png)
Best vs worst
Both of the options w/ NVENC seem good, game capture +x264 is very noticable and requires quite some CPU muscle to throw out ~1080p48fps stream. With this i'd say anyone trying to stream anything, particularly a game that doesn't run too well or loads CPU heavily.. this seems like a great option :D
With sufficient CPU, game capture + x264 = 1/5'th hit to performance and noticably disrupted experience, nvenc or shadowplay = pretty much invisible.
|
Interesting, so NVENC saving the day. I guess an Nvidia card is a lot cheaper than a whole new system.
Yeah, pretty much x264 crushes everything once you get past ultrafast, huh? There's way too much wizardry in that compared to original reference encoders and even most things people came up with afterwards.
|
How would you expect those results to scale with resolution? That is, would you expect better or worse results with one versus another at a higher resolution (1440p/1600p).
|
United Kingdom20326 Posts
On March 08 2014 08:37 Myrmidon wrote:
Interesting, so NVENC saving the day. I guess an Nvidia card is a lot cheaper than a whole new system.
Yeah, pretty much x264 crushes everything once you get past ultrafast, huh? There's way too much wizardry in that compared to original reference encoders and even most things people came up with afterwards.
Veryfast is much much much much better than ultrafast. (Oops that's actually what you said lol) That's the one mostly used for streaming, and comparison, though streaming a lower resolution/fps at medium preset is plausible and Veryslow is good for high efficiency encodes
On March 08 2014 08:42 Craton wrote:
How would you expect those results to scale with resolution? That is, would you expect better or worse results with one versus another at a higher resolution (1440p/1600p).
Game capture+x264 would quickly drop off, losing more performance and demanding more CPU. As far as i know, Maxwell NVENC is fast enough for live 1440p60 with no issues, kepler one maybe is (but maxwell nvenc is way faster) though i've not seen it tested, or even ran at all. If they can handle it, it should run well.
For live streaming, as 1080p60 would ideally have 5-20mbit bitrate.. there's not much reason to go to 1440p60, but for live encoding to a HDD it could be fun.
|
Can you do 1440p out? The limited information I found online was that you can do any resolution in, but only up to 1080@60 out. While we're on the topic, what about multi-card setups? How does that interact with ShadowPlay/NVENC?
|
United Kingdom20326 Posts
With shadowplay, you can't. With OBS control over NVENC you might be able to, i'm not sure.
While we're on the topic, what about multi-card setups? How does that interact with ShadowPlay/NVENC?
More performance hit but the SLI issue on shadowplay is fixed now (not sure where that issue actually was?). Not a crazy amount higher and more invisible than traditional software solutions which you can blind test that they are on - it feels like losing 100-200mhz off an OC as opposed to having screwed up game
|
On March 08 2014 04:50 Cyro wrote:
That's an ivy bridge dual core, low tdp processor (power limited to something like 18 watts between cpu and igpu before turbo turns off), guaranteed for only 1.8ghz clock speed, it's a very weak CPU compared to etc a 4670k, or even a desktop dual core, would definitely aim higher
I don't have actual performance numbers for working with capture cards etc, but my impression is that if you have sufficient CPU power available on the primary desktop.. they don't help at all, so i find it very hard to reccomend a $500+ laptop and like a $100 capture card for something that might not actually do anything at all.
hmm. so what do you reccomend. build a second tower with something like a i3 or i5. will the newer gen i3 cut it?
|
United Kingdom20326 Posts
What are your current specs?
|
My tower is 2500k 570gtx. but i wanna stream on a 2 computer setup.
|
United Kingdom20326 Posts
But for what goal? It doesn't make much sense to use a capture card and buy an entire second PC with at least a haswell i3 or ~fx6300 unless there's some gain for doing so. Like i said above, i have some doubts for gained performance, particularly vs using NVENC
|
the whole point of a second pc is to stream 1080p with no lag on my game play. I can go ahead and get a i7 4770k cpu but can still feel lag if i do. atleast i assume so
edit: also need a 2nd cpu for my gf
|
United Kingdom20326 Posts
On March 08 2014 15:48 CrayonSc2 wrote:
the whole point of a second pc is to stream 1080p with no lag on my game play. I can go ahead and get a i7 4770k cpu but can still feel lag if i do. atleast i assume so
edit: also need a 2nd cpu for my gf
^Read this page of the thread 
Point is, i don't know anybody who can say in detail how much better it would run, while this option is more easily tested and much, much cheaper, even if you're buying a second system anyway
|
but the purpose of the second system is for a second use and to stream my main computer. thats the WHOLE point.
|
United Kingdom20326 Posts
So you're building a second system as main goal but also want to use it for streaming if it helps out? What do you want the system for? (gaming etc, or facebook machine?)
I'm still thinking best option is likely streaming with nvenc, if you have a two gaming system setup it's easier because you can pass your 570 to that rig and grab a maxwell gpu for yourself
|
im happy with my computers performance. i just want to know the minimum for 1080p from the other tower to capture my current rig.
|
United Kingdom20326 Posts
Why would you want to use two systems when one would perform better though? If you already want the CPU for something and capture card is really cheap then maybe there's a point
|
cause my gf needs a computer? are you ignoring that whole point?
|
You could've simply said that. In fact, you never even hinted at needing a second machine.
|
i did state it, i'm not trying to be a dick....
|
|
|
|
|
|
EPT
