AV-over-IP spent the last decade fragmented across three open standards and one proprietary cluster. ISE 2026 was the inflection point: the AIMS Alliance launched IPMX as a fully developed, certifiable standard with 48 products approved on day one. Tender language for new control rooms started shifting within weeks. This article lays out what IPMX, SMPTE ST 2110, and SDVoE actually carry, where each one wins, and how to recognise the tender language that locks you into one of them.
The 2026 inflection point
AV-over-IP started as a wedge between two worlds. Broadcast facilities had SMPTE ST 2110 — a SMPTE-shepherded standard for uncompressed video, audio, and ancillary data over IP, with PTP-locked timing. Corporate ProAV had SDVoE — Software Defined Video over Ethernet, pushed by the SDVoE Alliance, anchored on the Marvell AQrate (formerly Aquantia) chipset family. The gap in the middle — control rooms, NOC and SOC operations, corporate AV, education, situation centres — relied on NDI or proprietary KVM-over-IP, neither of which is a true open standard with multi-vendor certification.
IPMX (Internet Protocol Media Experience) targets exactly that gap. It is built on the SMPTE ST 2110 family — same packet formats, same NMOS management plane — but modified so it runs on the cables ProAV already owns: 1 GbE, optional PTP, native HDCP. AIMS officially launched the standard on the ISE 2026 show floor, with Cobalt Digital, Matrox, Evertz and 45 other certified products listed in the AIMS product registry the same week.
Most tender forecasts in early 2026 converge on the same point: by end-2026, more than half of new control-room and situation-centre procurements will require IPMX plus NMOS in the technical specification. That is a category pivot, not a minor adjustment. Spec'ing the wrong standard now means either re-doing the build in 2028 or living with vendor-locked stacks the buyer was explicitly trying to avoid.
What each standard actually carries
SMPTE ST 2110 — the broadcast standard
ST 2110 is not a single protocol. It is a suite — ST 2110-20 for video, -30 for audio, -40 for ancillary data, -21 for timing — that splits a media stream into separate IP flows locked to a common PTP clock (IEEE 1588). The whole point of the split is that broadcast facilities can mix and match: route the audio one way, the video another, and recombine downstream without re-encoding.
Video in ST 2110 is uncompressed, which is the standard's biggest strength and its biggest cost. A single 4K60 stream eats about 12 Gbit/s of network capacity. That forces 10 or 25 Gigabit Ethernet end to end, plus switches with strict QoS, PTP boundary clocks, and enough port density to handle the source count. The total infrastructure cost rises fast above ten sources.
Adoption keeps accelerating in broadcast. The Haivision Broadcast Transformation Report 2025 surveyed roughly 900 industry professionals across 110 countries: 37% reported running ST 2110 in 2025, up from the high 20s a year earlier, and 51% reported a hybrid SDI-plus-IP stack with ST 2110 as the IP layer. Greenfield broadcast builds default to ST 2110; replacing legacy SDI routers with ST 2110 is the standard refresh path.
SDVoE — uncompressed video for sub-frame latency
SDVoE is the chipset-defined alternative. Every certified endpoint — encoder, decoder, transceiver — runs the same AQrate silicon family from Marvell. The transport is uncompressed 4K60 over 10 GbE, with end-to-end latency under a single video frame. USB, serial, and audio ride alongside the video over the same link, so SDVoE doubles as a KVM-class fabric where it is deployed.
The fabric is simpler than ST 2110 by design. There is no PTP requirement — endpoints synchronise via the chipset itself. The control plane is proprietary to the SDVoE Alliance, and while some vendors layer NMOS on top, NMOS is not the canonical management path. This makes SDVoE faster to commission than ST 2110 and harder to mix with non-SDVoE gear.
The cost story is also simpler and steeper. Switches must be 10 GbE throughout; SDVoE endpoints sit at a premium relative to NDI or H.264-based AV gear because the chipset licensing flows through. For workflows where sub-frame latency end-to-end is a hard requirement — trading floors, military command and control, some live production — that price is the table stake. Outside those workflows, the premium does not always pay for itself.
IPMX — ST 2110 adapted for ProAV
IPMX keeps the ST 2110 packet structure and the NMOS management plane but changes three things that matter for ProAV deployments:
- JPEG-XS compression. Visually lossless, low latency (single-frame end to end), and roughly 10× the bandwidth efficiency of uncompressed ST 2110. A 4K60 stream drops from ~12 Gbit/s to roughly 1 Gbit/s, which fits on standard 1 GbE switches.
- PTP becomes optional. IPMX endpoints can run without a facility-wide PTP grandmaster, which removes the single most expensive piece of ST 2110 infrastructure. PTP is still available when synchronised playback across a wall demands it.
- Native HDCP. Commercial HDMI sources — set-top boxes, BD players, console outputs, copy- protected production gear — flow through the fabric without bridging. ST 2110 has no native HDCP path; SDVoE handles it via chipset; IPMX puts it in the standard itself.
Management on IPMX is NMOS IS-04 for discovery and IS-05 for connection control, both mandatory. This is the same NMOS that some ST 2110 broadcast facilities run, which means the two standards interoperate cleanly at the control plane. A facility can run ST 2110 in the master control room and IPMX in the adjacent operations spaces without building two management universes.
Side-by-side
| Property | SMPTE ST 2110 | SDVoE | IPMX |
|---|---|---|---|
| Transport network | 10/25/100 GbE | 10 GbE only | 1 GbE and up |
| Video payload | Uncompressed | Uncompressed | JPEG-XS (visually lossless) |
| End-to-end latency | Sub-frame | Sub-frame | Sub-frame to single frame |
| PTP synchronisation | Mandatory (IEEE 1588) | Not used (chipset sync) | Optional |
| Discovery | NMOS IS-04 (recommended) | Proprietary | NMOS IS-04 (mandatory) |
| Connection management | NMOS IS-05 (recommended) | Proprietary | NMOS IS-05 (mandatory) |
| HDCP | Not in standard | Native via chipset | Native in standard |
| Switch requirements | Multicast L3, IGMP querier, PTP boundary clock, strict QoS | Specific to AQrate endpoints, IGMP, multicast | Standard Ethernet, IGMP |
| 2026 tender presence | Default for broadcast | Niche for low-latency operator workflows | Rising in ProAV / control room |
| Vendor diversity | High (50+ products listed in SMPTE registry) | Constrained by chipset licensing | Growing (48 certified at launch, more in pipeline) |
The honest summary of that table: ST 2110 sets the broadcast bar, SDVoE sets the low-latency operator-control bar, and IPMX sets the bar for everything else that wants open multi-vendor AV-over-IP without either of those infrastructure commitments.
When each one is the right answer
Pick ST 2110 when
The build is a broadcast facility (greenfield or refresh), the existing partners are SMPTE-aligned (Grass Valley, Sony, EVS, Imagine Communications, Lawo, Riedel), and the budget can absorb 10/25 GbE switches plus PTP infrastructure. If a sports OB truck or a master control room is on the BOM, ST 2110 is the default choice. Trying to substitute IPMX in a full broadcast facility means losing the uncompressed-video guarantee that broadcast-grade colourists and editors rely on.
Pick SDVoE when
A sub-frame end-to-end latency is the single non-negotiable requirement and the environment is small enough (typically under 64 endpoints) that the AQrate chipset premium is acceptable. Classic fits: financial trading floors with time-sensitive operator action; some military command-and-control rooms; live production environments where the operator's cursor must feel attached to the source PC. The price tag and the chipset lock-in are the real reasons SDVoE does not win bigger control-room bids.
Pick IPMX when
Most other ProAV control rooms, NOC and SOC builds, situation centres, higher-education AV networks, and large corporate AV estates fit IPMX better than either of the alternatives. Specifically:
- New build or refresh in 2026-2028 where the buyer wants an open standard but cannot justify 10/25 GbE everywhere.
- Mixed AV-and-IT infrastructure on the same Ethernet backbone, with standard switches the IT team already knows how to manage.
- Tender language calling out NMOS IS-04/IS-05 management. IPMX is the only AV-over-IP standard where NMOS is mandatory rather than optional.
- Workflows that need HDCP from commercial sources without an out-of- band copy-protection bridge.
- Procurements explicitly trying to avoid chipset lock-in or single-vendor stacks.
The honest caveat for IPMX in 2026: the vendor pool is smaller than ST 2110. 48 certified products is enough for most bids, but a tender that requires niche gear (specific encoder models, exotic I/O formats) may still need ST 2110 or SDVoE as the practical answer until the IPMX certified list grows.
The 2026 tender shift in numbers
Three data points anchor where the category is heading:
- 48 certified products at ISE 2026. AIMS Alliance announced the IPMX-conformant product list on launch week. Cobalt Digital, Matrox, and Evertz were among the first vendors with full certification; the list now extends to encoders, decoders, KVM endpoints, video wall controllers, and capture cards.
- >50% of new control-room procurements requiring IPMX+NMOS by end-2026. Most industry forecasts in early 2026 sit between 50% and 65% for this number. The exact cut depends on segment (federal vs corporate vs higher-ed) but the trend line is consistent across analyst voices.
- NMOS becomes a separate line-item in tender language. ProAV tenders in late 2025 already started listing NMOS IS-04 and IS-05 as required deliverables, often independent of the underlying transport. That language signals a deliberate move away from proprietary control planes — which IPMX satisfies natively and SDVoE does not.
Cross-referenced against the broadcast side: ST 2110 stays default for new broadcast facility builds (the Haivision 2025 numbers above confirm this), and that is not changing. The shift is at the boundary between broadcast and ProAV — where the two used to share nothing and now share the NMOS control plane through IPMX.
A practical decision tree
Strip the previous sections down to four questions, in order:
- Is this a broadcast facility (master control, OB truck, sports production, news studio)? If yes, spec ST 2110. The rest of the tree does not apply.
- Is sub-frame latency end to end an absolute hard requirement, and is the endpoint count under ~64? If yes, look at SDVoE. Validate the chipset premium against the budget before committing.
- Does the tender call for NMOS IS-04/IS-05 or open multi-vendor AV-over-IP? If yes, spec IPMX. ST 2110 is overkill for this language, SDVoE does not satisfy NMOS natively.
- Does the project run on standard 1 GbE IT infrastructure already? If yes, IPMX again. ST 2110 forces a network refresh you may not need.
If none of those questions returns "yes", the project likely does not need a SMPTE- family transport at all — NDI or H.264 streaming over standard switches is usually the right answer, and the cost story stays cleaner.
Where Craft Wall fits
Craft Wall is the software composer above the transport layer — the engine that takes whatever sources arrive over the network and lays them out on the wall. Today the supported source mix is NDI (full and HX), AV-over-IP via RTSP, browser/HTML5, IP-KVM, and screen-capture from any source PC. IPMX endpoint support is on the roadmap; the JPEG-XS decode side is the work in progress at the time of writing.
Two things matter for any IPMX-related procurement decision involving Craft Wall:
- Because Craft Wall sits above the transport, the IPMX vs ST 2110 vs SDVoE decision happens at the encoder / decoder layer, not at the composer. The same Craft Wall installation ingests an NMOS-discovered IPMX stream and a legacy NDI stream side by side once the IPMX decode path lands.
- Standard Linux hardware means no chipset lock-in at the wall controller tier. The IT budget for the composer stays on commodity x86 with a GPU, regardless of which AV-over-IP standard the encoders speak.
If a tender requires IPMX-conformant encoders today, the right path is to spec certified IPMX gear (from the AIMS product registry) at the source side and bring Craft Wall in once the IPMX decode path lands. The composer layer is the more flexible piece of the stack — switching wall software is cheaper than switching transport.
Closing
AV-over-IP in 2026 is not "pick the winner" — it is "match the spec to the room". ST 2110 owns broadcast. SDVoE owns sub-frame latency at moderate scale. IPMX is the open standard for everything else, and the 2026 tender shift is real enough that buyers who ignore it now risk re-doing the build in 2028. The safest default for a new control-room procurement is IPMX plus NMOS, with a clear escape valve to ST 2110 if any part of the facility crosses into broadcast.
Read next: IPMX, NMOS, SMPTE ST 2110, SDVoE glossary entries, or the Craft Wall vs VuWall TRx comparison for how this transport question plays out at the wall-management tier.