logo

Live Production Software Forums


Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

Options
Go to last post Go to first unread
Ph0en1x  
#1 Posted : Saturday, November 1, 2014 9:31:50 AM(UTC)
Ph0en1x

Rank: Advanced Member

Groups: Registered
Joined: 7/31/2014(UTC)
Posts: 38

Thanks: 1 times
Was thanked: 2 time(s) in 2 post(s)
So i just found out that i actually do need a dedicated graphics card to livestream 1080p or even 720p...

I tested my vMix preset on my laptop and there i saw no difference between using the Intel HD Graphics 4000 of my 3rd gen i7 and the dedicated Geforce GT 750m GPU on my laptop, so i purchased a new desktop build without dedicated GPU and the i7 4790 @ 3.60hz, trusting i could ride it's integrated graphics.

Now, with exactly the same inputs and settings, i run into horrible render times on the new pc, even when just recording to the SSD or simply viewing the inputs in vMix.

Two questions:

1. Can someone explain why a) i didn't notice any difference on my laptop between the dedicated GPU and the 3d gen integrated graphics, and b) why the Intel HD Graphics 4600 of my i7 4790 perform far worse than the Intel HD Graphics 4000 in my laptop?

2. If there is a logical explanation for this, which graphics card is recommended? Of course i'm looking for best bang for buck.

I'll be using the following inputs:

1. A primary 1080p camera feed through a Blackmagic Intensity Shuttle (possibly extended with some RTSP feeds from secondary cams)
2-6. A handful of 1080p video files
7. A 1080p lossless stinger animation with transparency
8-13. A handful of 1080p text overlays
14. A video delay input that i use for live replays (and saving them for future use)

Many thanks in advance for any answers and/or suggestions!
tdurhamjr  
#2 Posted : Saturday, November 1, 2014 11:14:16 AM(UTC)
tdurhamjr

Rank: Advanced Member

Groups: Registered
Joined: 9/8/2014(UTC)
Posts: 244
Location: Southaven MS

Thanks: 12 times
Was thanked: 25 time(s) in 21 post(s)
Have you considered installing a Black Magic Decklink Mini in your desktop? Those are only about $150 and you will plug in into your PCIe slot. I have the same BM Int Shuttle as well as the Elgato HD60. I think the Decklink Mini will give you much better results and performance.
Ph0en1x  
#3 Posted : Saturday, November 1, 2014 11:26:03 AM(UTC)
Ph0en1x

Rank: Advanced Member

Groups: Registered
Joined: 7/31/2014(UTC)
Posts: 38

Thanks: 1 times
Was thanked: 2 time(s) in 2 post(s)
I already have the Shuttle, which i like for the fact i can just unplug it whenever, so what would be the added value of the Black Magic Decklink Mini? It doesn't help with the high render time, does it?
admin  
#4 Posted : Sunday, November 2, 2014 11:13:32 PM(UTC)
admin

Rank: Administration

Groups: Administrators
Joined: 1/13/2010(UTC)
Posts: 5,214
Man
Location: Gold Coast, Australia

Was thanked: 4301 time(s) in 1523 post(s)
Hi,

Intel Graphics 4000/4600, while a bit slow should still be fast enough for small HD productions.
What sort of render times are you seeing and what frame rate and resolution is set in vMix settings?

Try adding the inputs you need one by one to identify what is causing the increase in render times.

Regards,

Martin
vMix
Ph0en1x  
#5 Posted : Monday, November 3, 2014 7:13:18 AM(UTC)
Ph0en1x

Rank: Advanced Member

Groups: Registered
Joined: 7/31/2014(UTC)
Posts: 38

Thanks: 1 times
Was thanked: 2 time(s) in 2 post(s)
I have two different presets, one at 720p and one at 1080p. Both at 25 fps. For 720p i have downscaled the input video's accordingly, except for some static overlayed titles.

For 720p: When i'm just playing in vMix, render time is <10, but once i start streaming, it goes towards 40, depending on the amount of movement in the output. It doesn't drop frames though, which is also seen when i record. The render time is about the same, but there are no source frames dropped.

For 1080p: The render time is about the same as for 720p when just playing, but once i start streaming it peaks at 100 and the input framerate is only 12-15. When recording, there are indeed a lot of source frames dropped.

I don't think specific inputs are to blame, since i experience the same problems with different ones.

I'm still wondering why my old Intel Graphics 4000 performed better than my Intel Graphics 4600. Perhaps the 4000 actually did profit from the dedicated GPU that was sitting beside it? I did set it up so that vMix should use the integrated graphics, but perhaps the graphics card was still added in the mix somehow?
admin  
#6 Posted : Monday, November 3, 2014 7:20:30 AM(UTC)
admin

Rank: Administration

Groups: Administrators
Joined: 1/13/2010(UTC)
Posts: 5,214
Man
Location: Gold Coast, Australia

Was thanked: 4301 time(s) in 1523 post(s)
What does the CPU usage look like when streaming?
If the CPU reaches above around 65%, you will start to see an impact to the render times, so try some different streaming/recording
settings to see if things get a bit more manageable.

Regards,

Martin
vMix
h2video.nl  
#7 Posted : Monday, November 3, 2014 9:30:20 AM(UTC)
h2video.nl

Rank: Advanced Member

Groups: Registered
Joined: 3/7/2012(UTC)
Posts: 503
Man
Location: The Netherlands

Thanks: 9 times
Was thanked: 45 time(s) in 39 post(s)
you write:

or 720p: When i'm just playing in vMix, render time is <10, but once i start streaming, it goes towards 40, depending on the amount of movement in the output. It doesn't drop frames though, which is also seen when i record. The render time is about the same, but there are no source frames dropped.

so likely your streaming program is to blame....

just to be sure: where do you record your cpu load from: vmix or the windows machine, and what are the different percentages of all processes running.

2 in my experience the workflow should be alligned: downscalong is ok but if you keep everything the same scale rendering times are minimum.

3. i prefer to seperate recording and streaming machines. this way i can restart a stream machine without stopping a recording and the load is optimala for recording and for streaming.

basically i am always wondering how on earth Martin managed to get such a small programme ( now approaching 75 mb) to perform so well, with such a minimal impact. if impact is high it's usually something i did wrong or a logical explanation.

p.e. when using a camera input there is no extra load but when i use a webcam the load will increase significantly. webcam, recording mp4 and streaming... usually a little bit too much for my systems...

so i suggest you build your system up from scratch, one input at a time and one action like recording of streaming at a time and try to pin point the likely cause.
stefan
Ph0en1x  
#8 Posted : Monday, November 3, 2014 10:12:57 AM(UTC)
Ph0en1x

Rank: Advanced Member

Groups: Registered
Joined: 7/31/2014(UTC)
Posts: 38

Thanks: 1 times
Was thanked: 2 time(s) in 2 post(s)
h2video.nl wrote:
you write:

or 720p: When i'm just playing in vMix, render time is <10, but once i start streaming, it goes towards 40, depending on the amount of movement in the output. It doesn't drop frames though, which is also seen when i record. The render time is about the same, but there are no source frames dropped.

so likely your streaming program is to blame....
I think you misread: when i use record (and don't use FMLE), the render time is the same as when i stream.

admin wrote:
What does the CPU usage look like when streaming?

Quote:
just to be sure: where do you record your cpu load from: vmix or the windows machine, and what are the different percentages of all processes running.

The CPU loads on my new system as displayed by vMix:

For 720p: when playing in vMix around 10%, when recording to mp4 around 20-25%, when streaming around 15-20%.

For 1080p: when playing in vMix around 30%, when recording to mp4 around 45-50%, when streaming around 25-30%.

Give or take a few %, these numbers correspond with the load that Windows attributes to vMix. When streaming, FMLE takes up an additional 45% for both 720 and 1080p.

Quote:

2 in my experience the workflow should be alligned: downscalong is ok but if you keep everything the same scale rendering times are minimum.

As far as know, I'm not scaling anything during my tests. The video inputs i'm trying to stream in 720p are prerendered in that resolution. The same goes for 1080p. Only the static titles are always 1080p, but i'm not using those yet.

Quote:

3. i prefer to seperate recording and streaming machines. this way i can restart a stream machine without stopping a recording and the load is optimala for recording and for streaming.

Normally I wouldn't be recording anything, just streaming. But when i use the record function i can see where the dropped frames are coming from, and for the 1080p preset (where render times are >100 when recording) they are 'due to source frame rate' and not 'due to hard disk/cpu'.

Quote:

so i suggest you build your system up from scratch, one input at a time and one action like recording of streaming at a time and try to pin point the likely cause.
stefan

I don't feel doing that will change anything though, since the exact same presets and inputs perform better on my laptop with a lesser cpu and lesser integrated gpu.

The only edge my laptop has over my desktop is a dedicated graphics card, which i thought i wasn't using. However, i tried recording the output of my 1080p preset on my laptop, using the integrated graphics for vMix, and the render time didn't go above 10ms, whereas on my desktop, using the exact same preset and 1080p input, it exceeded 100. So i think somehow the dedicated GPU on my laptop is still being used for rendering even though i want to use the integrated graphics, and that dedicated GPU is the missing link on my desktop.

What remains is my original question: which card should i get for reliable 1080p vMixing?

hexx  
#9 Posted : Tuesday, November 4, 2014 5:11:48 AM(UTC)
hexx

Rank: Newbie

Groups: Registered
Joined: 11/4/2014(UTC)
Posts: 6
Location: greece

Thanks: 4 times
If yoy are using mp4 video files try to transcode them to mpeg files and make a test , this should reduse rendering times and cpu usage in Gvmix (G for great)
Ph0en1x  
#10 Posted : Tuesday, November 4, 2014 9:43:27 AM(UTC)
Ph0en1x

Rank: Advanced Member

Groups: Registered
Joined: 7/31/2014(UTC)
Posts: 38

Thanks: 1 times
Was thanked: 2 time(s) in 2 post(s)
hexx wrote:
If yoy are using mp4 video files try to transcode them to mpeg files and make a test , this should reduse rendering times and cpu usage in Gvmix (G for great)

The video file i'm using as input is mpeg. And again: the exact same file in the exact same preset plays better on my inferior system with a dedicated GPU (which i thought i wasn't using) than on my new desktop without dedicated GPU.

Shall i just go for the Nvidia GeForce GTX 650 Ti 1GB?

hexx  
#11 Posted : Tuesday, November 4, 2014 10:56:21 AM(UTC)
hexx

Rank: Newbie

Groups: Registered
Joined: 11/4/2014(UTC)
Posts: 6
Location: greece

Thanks: 4 times
If you could , try this too , do not open a preset project , create an new one with the same inputs and do the test again , we had a performance issue once with a preset project (corrupted file?) , we created a new and the problem gone !
Ph0en1x  
#12 Posted : Tuesday, November 4, 2014 3:52:47 PM(UTC)
Ph0en1x

Rank: Advanced Member

Groups: Registered
Joined: 7/31/2014(UTC)
Posts: 38

Thanks: 1 times
Was thanked: 2 time(s) in 2 post(s)
hexx wrote:
If you could , try this too , do not open a preset project , create an new one with the same inputs and do the test again , we had a performance issue once with a preset project (corrupted file?) , we created a new and the problem gone !

I tried that. A new preset with only one video file input performs better on my 3d gen i7 laptop than on my new 4th gen i7 desktop.
Users browsing this topic
Guest
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.