I doubt this is actually a vMix problem, more a NDI / network configuration / multicast issue.

I'm experiencing significant NDI freezes when roughly 15 or more devices are receiving a NDI feed from vmix.

Device often needs to look at another NDI feed and back to the main one to make it work again. I think that just immediately cuts off another device, although not had opportunity to check that for sure.

Available bandwidth is not a problem.

vMix > 10GBe > Managed switch, 2 x 10GB ports, 8 x 1GB ports (GS110EMX) > 1GBe > 4 x 8 port 1GB switches (GS308) > devices... (laptops)

If I instead share out the 4 vMix NDI outputs amongst all the devices, so each NDI feed is only received by a handful of devices, it's all fine.

My networking knowledge is limited, but as far as I know with multicast vMix wouldn't even know if 1 or 100 things were watching it's NDI feed anyway.

So what could be causing this?

Have followed Newtek's switch config guidelines as close as I can given the limited feature set in the 10GB switch, and the 1GB switches are unmanaged.

Jumbo frames off.
QoS off.
IGMP on.

Network has other NDI traffic, each of the 20 laptops are screensharing back to vMix, various production monitors, vmix calls etc. These all function properly.
Network also serves internet to devices that need it.
No firewalls.

I don't typically have 20 devices on hand to test this, but I have an opportunity next week so any suggestions anyone has would be most welcome!

Cheers

Ah, interesting, then I might be saturating the 1GB links between the switches.

Thank you, I'll look into that.

The bandwidth can very quickly be a problem.


To get a multicast you may need to go via HDMI into an NDI encoder that supports multicast. Make sure your network is multicast enabled. That isn#t switched on by default in most cases.

Try turning IGMP off, does this make any difference? IGMP trims where multicast packets go...if turning it off improves your issue then perhaps the IGMP rules need adjusting. I've also found that IGMP functionality isn't always enforced equally between different brands of switches.

We had a very similar issue yesterday when doing a vMix production where we were taking two speakers and their slide feeds from MS Teams using it's NDI broadcast feature, mixing in vMix with PiPs, branded backgrounds, namestraps, VTs, etc - and then broadcasting the mixed production on BlueJeans and to a backup microsite. We had primary and backup machines on the MS Teams call. At then time of the production we had vMix sessions open on 3 or 4 other machines and a couple of other machines turned on that put NDI on to the network.

What I didn't realise was:
- if NDI output is selected on a vMix machine, it gets output on to the network even when External isn't selected
- it seems it is the number of available NDI sources rather than the number of selected NDI sources that is significant for vMix machines. We were seeing jerkiness on the 3 NDI sources we were using and we fixed it by turning off the vMix machines we weren't using - that's what it seemed anyways

What we are going to do is migrate all our vMix machines to 10GB Ethernet and create a NDI production VLAN

regards, Scott

Cheers Andrew for the reply. I think I wasn't seeing the forest for the trees - the issue was more likely that we had too much NDI traffic on a Gigabit Ethernet network and we should migrate to 10 Gb - ie. nothing to do with the vMix machine itself