Today, i made a small test with my PC and BM quadlink. I used a hdsdi video from interview (to see the sync lips) and made a x4 output connected to the BM.
Here are the results :
- with one decklinck only, CPU is 3% and the sync is nearly non visible
- With 2x decklink, CPU is 11% and the lip sync is sensible
- With 3x decklink, CPU is about 21% and the lipsync is about 1/2 sec
- With 4x decklink, CPU is about 35% and the lipsync is about 1 sec

The good news is that the lipsync is the same on all the decklink at the same time (lipsync could have move). The question is with a different input card, shall i get the same ?

in my experience it is almost impossible to avoid latency on a vmix output. with mixing and recording a latency to the actual event can be observed. in the recording it is no problem but for mixing and projecting on a screen, latency will be a problem.

what could be done, and i sometimes do, is feed the video to the screen direcltly from the main camera to the screen and use a seperate feed for mixing and recording.

greetings stefan

Our church is currently using vMix for our IMAG system for our whole service.

Depending on what is "acceptable" latency means to an individual, I believe that vMix is definitely usable for an IMAG situation. We have been running multiple cameras through vMix for IMAG for a few months now and It has been working quite well. We are not even using the low latency capture option either. Our render time is between 8-15ms.

How much latency are you experiencing now?

Here is a link to a video of our church this morning. You can see our screens as well as our vMix window.



How much latency are you experiencing now?

Take a look at the video above. I would have to guess between 100-200ms. Ill have to measure it.

How many cameras, what model and how are they connected to what capture card?

Camera 1 - Canon HF R20 at 1080i60 into a Intensity Pro HDMI
Camera 2 - GoPro Hero 3+ Black Edition at 1080i60 DeckLink Recorder HDMI
Desktop Capture - 1600x900 30FPS Medium JPEG

What resolution and data rate are you using for capture?

Not sure of the data rate.

What is your IMAG display and how is it connected?

We have 2 different feeds in the sanctuary. One center screen that is used only for lyrics and movie displays. Its a Christie LX500
Side screens are the vMix feed and they are 60" LG TV's. They are currently hooked up over VGA but will switch to HDMI next week.

Are you recording and/or streaming as well as mixing?

We are recording at 1080p30
We are streaming at 864x486 at about 950kb
We are outputting 2 outputs feeds from our videocard.
-(Mon 1)DVI vMix Screen 1080 60Hz
-(Mon 2 Duplicated)VGA External Output 1080 59Hz
-(Mon 3 Duplicated)HDMI External Output 1080 59 HZ

Do you use any other inputs?

About 8 overlays for lower thirds

What is your PC system spec?

i7-4770
12gb ram
Sapphire Radeon 6950 2GB video card
Decklink Recorder
Intensity Pro
X-Fi Soundblaster (Optical Toslink input from AES/EBU audio feed)



Hope that helps someone! Let me know what you think :)

You bet!

We really didn't adjust too much. Pretty much started out great. Below is a link to a podcast of one of our services. The only thing that is added is an intro video. Everything else is basically how it is recorded through vMix

We did notice a decent improvement switching from a first gen i7 to a i7-4770.

How much latency do you think that you are getting? What motherboard do you have?

In my experience, the processor and video card have had the biggest difference in performance with streaming and external output. I could be wrong but that has been my history.

What are the specs of your computer?

I did tests today to see how to improve latency.
inputlag is a great clock tool, thank you to Rehan. But I had to do about 5 or 6 pictures at each time to be sure to a good result.

My test is :
Film the first PC screen with te clock (inputlag) with a JVC HM600 camera (HDSDI output),then a HDSDI multiplier, then scaler (to convert 1080p60 in 1080i/50 wich is my standard for recording, live events, etc.)
The camera has about 40 msec delay, and the scaler has also 31 msec delay.

I had a very very big delay on my PC shuttle i7 with zotac GT610 (nvidia) GPU and 4X HDSDI input black magic : 1,4 sec !!!! with 4x HDSDI connected to it and about 250 msec with only one. The CPU charge on vmix says 35% with 4x HDSDI !!
The GPU is the weakest element in my chain but i wanted to be sure.

So, I changed first the BM card for a magewell 4 HDSDI input card. The CPU charge was the same but the delay decreases to 700 msec with 4 HDSDI connected.
Conclusions :
- 4 HDSDI simuultaneously connected drain more delay than one.
- when your GPU is "out", the magewell drains lower power and has less delay than the black magic.

Then, I changed the computer for an old i7 x980 3,33 GHz 32 G RAM (3 years old config) and a GTX 480 GPU

The delay is between 220 and 250 msec. To have one or four HDSDI input simultaneously does not change much about the delay.

Then I changed for a 680 GTX wich is much more powerfull than the 480, and the delay is around 200 msec. Don't forget that you have 40 msec for the camera and 30 msec for the scaler included. So, the vmix delay in this config is actually 130 msec. You have to delay the sound if you want the lips to stay in sync in live event.

I will do some test with titan GPU and bigger i7 processor in a few days.

Olivier

I started doing some tests on latency recently, as it's something I get asked about quite often. My tests involved both Blackmagic and Magewell capture devices, and I tested using both vMix & VidBlaster. The result was typically about 4 frames (at 25fps), plus or minus a frame. Here's a shot of the setup using VMix with a total of six HD inputs (a mix of 720p50 and 1080i50) in which the latency through the PC is ~4 frames.

In a nutshell, timecode A (far right) is the source reference being fed into a capture device on the PC on the left (i7-930+GTX660), and timecode B is the result after the round trip back out of an HDMI o/p from the graphics card. The setup is more fully explained in a video clip that featured on That VidBlaster Guy - Latency Tests video excerpt. The full version (in which the audio sync is perfect, BTW), includes both vMix & VidBlaster tests using PCIe and USB3 devices, and will be on YouTube when I get round to finishing the edit.

Hi Martin,

Think you may have done some research for us already.
Which, have to say, was unexpected and extremely appreciated.


Curious to see your findings for Decklink Quad (Single input and fully loaded) vs The Magewell XI400DE-SDI (single input and fully loaded).

Thanks again,

Pete

Hi Guys,

I am fighting the latency beast too :)

I tested two of my systems, an 1155 socket i7 49xx 3gb Nvidia GTX 960 and a new i7 6700 w 2gb Nvidia GTX 670.

Both of the systems set to low latency capture... the best I can get is about 4-5 frames latency @ 59.97i. I calculate that to be 67 - 83 milliseconds (I tested counting back fields on a video delay, with deinterlace blending turned off.)

My question is, is this the best I can get? Is it GPU or CPU that I need to upgrade to make it more snappy?

Also... vMix site says that latency can be as low as 2-3 frames... (Latency) What frame rate is that measured at...

Also what is a good standard to stay under for latency... I find that this is not that noticeable on its own, but when you have to run it through another scaler/ mixer / system a frame or two more starts to look bad.

Thanks

Mike

I connected my camera as an input to vMix and recorded it as a video delay...
then I pointed my camera at the output screen of vMix...
I did a "clap" where I can see my clap and the clap in the vMix output on the screen.
Save the video delay.
You can ether watch the video delay at 4% or open the clip in an editor and count the frames between your initial clap and the clap and the clap in the monitor.

This measures the full round trip delay including camera and everything.