Free forever·no account, no tracking, no third-party requests ·open data, free API
showstack / compare / ndi-vs-smpte-st-2110

NDI vs SMPTE ST 2110

Is NDI good enough, or do I need ST 2110? Below is every field we hold on both, side by side. Each one is sourced on the individual entry, and anything nobody could source is left blank rather than guessed.

Side by side

NDISMPTE ST 2110
Categoryvideo-transportvideo-transport
FlowMedia — large, continuous, on time and in orderMedia — large, continuous, on time and in order
PortsUDP 5353, TCP 5959, TCP 5960, TCP 5961
Transporttcp, udpudp
Multicastyesyes
Spec availabilityproprietary-documentedpublished-paid
StewardVizrt (NDI AB), originally NewTekSMPTE (Society of Motion Picture and Television Engineers)
First published20152017
Statuscurrentcurrent
Our confidenceverifiedreported

Both are media flows. Anything that starves the link — usually a management transfer nobody scheduled — and jitter that outruns the receive buffer. That applies equally to either, so it is not a discriminator here.

What each one carries

Capability defined by the specification, not by any one product. A blank means the specification does not fix it, which is itself the answer on a transport like SRT.

NDISMPTE ST 2110
Sample rates48 kHz48 kHz
Audio bit depth16-bit, 32-bit16-bit, 24-bit
Audio encodingFloating point or 16-bit PCM, carried with the videoST 2110-30, which is AES67
ResolutionNot fixed by the specificationNot fixed by the specification
Frame rateNot fixed by the specificationNot fixed by the specification
Colour sampling4:2:2, 4:2:04:2:2, 4:4:4
Video bit depth8-bit, 10-bit8-bit, 10-bit, 12-bit, 16-bit
Video compressionVisually lossless (NDI High Bandwidth) or H.264/H.265 (NDI HX)None — uncompressed essence

NDI, audio: Audio and video travel together, unlike ST 2110 where they are separate essences.

NDI, video: NDI caps neither resolution nor frame rate; 3840x2160 is where most hardware decoders stop and software goes past it. The two families are very different things sharing one name: High Bandwidth is lightly compressed at 100 to 250 Mbit/s with low latency, HX is long-GOP at a fraction of that and correspondingly more delay.

SMPTE ST 2110, audio: Essence streams are separate, so audio and video are independently routable — which is the whole point of 2110 and the reason lip sync becomes something you have to manage rather than something you get.

SMPTE ST 2110, video: ST 2110-20 is resolution and frame-rate agnostic by design; what limits you is the link. Uncompressed HD is around 1.5 Gbit/s and UHD around 12 Gbit/s, so the practical ceiling is the network rather than the standard. Audio travels separately in ST 2110-30 as AES67.

Speaks both (5)

These are the safe choices if you have not committed to either, or need to bridge the two.

NDI only (35)

SMPTE ST 2110 only (2)

What goes wrong with NDI

Discovery is mDNS, which is link-local. NDI sources do not appear across a router. If you need cross-subnet sources you must run the NDI Discovery Server on TCP 5959 and point every device at it explicitly.
Opening only 5960 and 5961 in a firewall gets you discovery and no video. Each connection also grabs ports from the 6960 and 7960 blocks, so the rule has to cover those ranges.
NDI HX is a different thing to full NDI. It is hardware H.264 or HEVC, much lower bitrate and noticeably more latency, and it is what most cheap PTZ cameras actually send.
Full-bandwidth NDI is roughly 100 Mbit/s for 1080i per stream. A gigabit uplink shared with anything else runs out faster than people expect, and NDI degrades by dropping frames rather than by failing cleanly.
The SDK is royalty-free but proprietary, and the licence requires visible NDI attribution in your application. There is no published wire format, so every implementation goes through Vizrt's binary SDK.

What goes wrong with SMPTE ST 2110

Video, audio and ancillary data are separate streams with separate multicast addresses. Losing one leaves the others running, so "no audio" or "no timecode" can present as a perfectly healthy video path.
Everything hangs off PTP. ST 2110-10 defines the system timing model, so a grandmaster that steps, or two grandmasters fighting, takes the whole plant out at once rather than degrading gracefully.
ST 2110-20 video is genuinely uncompressed. Bandwidth is set by raster and frame rate, not by an average bitrate, so plan link capacity from the video format and leave headroom before you commit to a switch.
ST 2110-21 defines sender traffic shaping types. A sender and a receiver with mismatched shaping expectations gives intermittent picture breakup under load, not a clean failure you can point at.
ST 2110-30 audio is closely related to AES67 but the two do not automatically interoperate. Packet time and channel count profiles have to match on both ends before anything passes.
The suite is a dozen separate documents (ST 2110-10, -20, -21, -22, -30, -31, -40, -41, -43, plus RPs) and they are priced individually. Budget for the whole set, not one part number.

Sources

Disagree with any of this?

Every claim above comes from one of the two entries and carries a source. If one is wrong, correcting it is a single file and it fixes this page automatically. Edit NDI or edit SMPTE ST 2110.