Streamers and live production teams relying on NDI discovered a nasty surprise after installing Microsoft’s August 2025 cumulative updates: previously smooth video and audio streams suddenly stuttered, dropped frames, and produced choppy audio on local networks. The culprit: a low-level regression in how Windows handles NDI’s default Reliable UDP (RUDP) transport, triggered by KB5063878 (Windows 11, OS Build 26100.4946) and KB5063709 (Windows 10). The good news is that a simple, non-destructive workaround—changing the NDI Receive Mode to Single‑TCP or UDP Legacy—restores full performance without rolling back any security fixes.

Background: NDI and the Patch that Started It All

NDI (Network Device Interface) is the backbone of countless multi‑PC streaming rigs, enabling high‑quality, low‑latency video and audio transport over standard Ethernet. From OBS Studio to vMix and professional broadcast suites, NDI connects game capture machines, graphics render boxes, and encoder PCs. So when Microsoft’s August Patch Tuesday delivered security fixes and quality improvements, no one expected a protocol‑level breakdown.

Shortly after the updates rolled out, forums and support channels lit up with reports of severely degraded NDI streams. Video stuttered relentlessly, frames dropped even on gigabit LANs, and audio became choppy. Microsoft quickly added the symptom set to its Release Health dashboard, confirming the issue and promising an investigation. KB5063878 for Windows 11 (OS Build 26100.4946) and its Windows 10 sibling KB5063709 are the definite offenders.

What’s Actually Failing: RUDP Under Fire

NDI’s default transport is Reliable UDP (RUDP)—an application‑layer protocol that blends UDP’s low latency with custom retransmission logic to recover lost packets. It’s a clever design that worked well, until the August updates.

The updates appear to alter socket behaviour, packet scheduling, or driver interactions in a way that upsets RUDP’s assumptions. Instead of smooth delivery, users see retransmission storms and delayed packet handling. The result: video that looks like it’s buffering every second, and audio that breaks up like a bad VoIP call. Importantly, NDI traffic using plain UDP or Single‑TCP does not exhibit the same failures, which points directly to RUDP as the regression target.

Affected applications include OBS Studio when using NDI sources, NDI Tools (Access Manager, Studio Monitor), vMix, and any other NDI‑aware production software. The problem is especially visible when the sending machine uses Display Capture—a staple for gamers and presenters—rather than a camera feed.

The Immediate Fix: Switch Receive Mode

For the vast majority of users, the best response is to change the NDI Receive Mode on every receiving machine. This avoids the RUDP regression entirely, keeps the Windows security updates in place, and restores reliable streams. Here’s exactly how to do it:

  1. Download and install NDI Tools from the official NDI distribution page if you haven’t already.
  2. Launch NDI Access Manager from the Start menu.
  3. Go to the Advanced tab.
  4. Under Receive Mode, change the selection from Reliable UDP (RUDP) to either Single‑TCP or UDP (Legacy).
  5. Click OK to save.
  6. Restart every NDI‑receiving application (OBS, Studio Monitor, vMix, etc.) so they read the updated configuration file. This step is critical—NDI apps load settings only at startup.
  7. Repeat on every receiving host in your NDI chain. Leaving even one machine on RUDP can cause asymmetric behaviour.

Single‑TCP vs. UDP (Legacy): Choosing the Right Fallback

  • Single‑TCP: Replaces RUDP’s custom reliability with TCP’s ordered, guaranteed delivery. It’s the most robust option across complex networks, NATs, and firewalls. The trade‑off is a slight latency increase and potential head‑of‑line blocking during packet loss—still far better than the current stutter.
  • UDP (Legacy): Strips all retransmission logic. It delivers the lowest possible latency on a clean Ethernet LAN, but any packet loss translates directly into visual or audio glitches. It avoids the retransmission storm entirely, which is often preferable for live production.
  • RUDP: Previously the ideal balance, but now broken by the Windows update. Once a permanent patch arrives, you can revert to it.

After switching, always reboot or restart the NDI processes on each host. NDI reads Access Manager’s configuration (%programdata%/NDI/ndi-config.v1.json) only at application launch.

Why Changing Transport Works: A Technical Explainer

RUDP operates by wrapping UDP datagrams with sequence numbers and retransmission requests. It tries to keep latency low by requesting only missing packets instead of waiting for full in‑order delivery like TCP. The August update likely disrupts the timing or buffering of these retransmissions, creating a vicious cycle where the sender and receiver keep retransmitting the same data or processing it too late.

By switching to Single‑TCP, you delegate all reliability to the operating system’s mature TCP stack, which is not affected by the regression. Choosing UDP (Legacy) simply removes the reliability layer, so the broken RUDP code path is never touched. Both methods sidestep the fault, and both have been validated by NDI vendors and the community.

Caveat: Microsoft has not yet issued a detailed root‑cause analysis. The explanation above synthesises vendor findings and network engineering principles—it is not a verbatim statement from Microsoft engineering. Flagging uncertainty is important until a formal write‑up is published.

Risks, Trade‑offs, and Live‑Show Hygiene

Changing Receive Mode is a configuration change, not a silver bullet. Understand the operational tradeoffs and always run rehearsals.

  • Security vs. Availability: Keeping the patch installed is always safer. Uninstalling the KB removes multiple security fixes and should be a last resort only for machines that absolutely cannot be reconfigured (e.g., remote appliances you cannot touch). If you must roll back, isolate the machine on a private VLAN with no internet access.
  • Latency and Lip Sync: Single‑TCP may introduce a small latency penalty. For multi‑camera productions, test and adjust audio delays to maintain sync.
  • Packet Loss Sensitivity: UDP (Legacy) exposes every dropped packet, but many production LANs are clean enough that the occasional glitch is preferable to retransmission storms. Test on your actual switch fabric.
  • App Restart Requirement: Already‑running NDI apps will not pick up the new Receive Mode until you restart them. Plan a restart window and notify all contributors.

Broadcasters’ Operational Checklist

  1. Identify affected hosts: Run winver or check Settings > Windows Update > Update history for KB5063878/KB5063709.
  2. On each receiving machine:
    • Install NDI Tools if missing.
    • Set Receive Mode to Single‑TCP or UDP (Legacy) via Access Manager.
    • Restart all NDI apps and verify source discovery.
  3. Run a full show rehearsal: Check switching, transitions, OBS scenes, and audio sync under load.
  4. If a remote guest machine cannot be reconfigured, consider a controlled rollback only for that host, with network segmentation.
  5. Monitor Microsoft Release Health and vendor advisories for a permanent fix; plan to test and redeploy when ready.

Enterprise and Advanced Options

Large organisations should look to Windows Update for Business, WSUS, or SCCM to hold back the problematic LCU from production rings until a fix is validated. Microsoft has previously released Known Issue Rollback (KIR) policies for regressions of this type; watch for one targeting this NDI issue. In the meantime, the Access Manager workaround can be pushed via deployment tools across your estate.

For mission‑critical broadcast operations, consider maintaining a staging ring that mirrors production. Run new cumulative updates there first and validate with a full NDI workflow before wide deployment.

What Vendors and Microsoft Are Saying

Microsoft’s Release Health dashboard and the KB5063878/KB5063709 support pages now list the symptom set, confirming the investigation. Vendors like NewTek (the creator of NDI) and vMix have published practical guidance that centres on switching Receive Mode as the safe temporary mitigation. Community knowledge bases echo the same advice.

When a permanent fix arrives, expect these steps:

  • Validate the patch on a test machine first.
  • Revert NDI Receive Mode to Auto or RUDP only after confirming stability across your topology.
  • Rehearse a full production show to be sure.

Long‑Term Hardening: Avoiding Update Surprises

  • Stage updates: Maintain a ring of non‑critical machines that gets patches 1–2 weeks ahead of live production.
  • Isolate NDI traffic: Use separate NICs or segmented VLANs for NDI so that internet‑facing or management traffic cannot interact with media flows.
  • Automate configuration: Use deployment tools to enforce the desired Receive Mode across your estate, and verify after every patch cycle.
  • Keep vendor tools current: NDI Tools, capture card firmware, and NIC drivers can all influence transport stability. Coordinated updates reduce multi‑factor regressions.
  • Document a recovery playbook: Steps to change transport, restart apps, roll back a KB, and apply network segmentation should be tested quarterly.

Final Assessment and Recommendations

The August 2025 Windows updates have created a serious, high‑impact regression for anyone relying on NDI in live production—but the problem is solvable without sacrificing security. For the vast majority of streamers and broadcasters, the clearest, safest path is:

  1. Immediately change NDI Receive Mode to Single‑TCP or UDP (Legacy) on every receiving host. This restores reliability while keeping the August security updates intact.
  2. Reserve KB uninstallation only for truly unreconfigurable machines, and treat it as a temporary, controlled emergency measure with network isolation.
  3. Implement a staging/test ring and rehearse patch windows so future regressions are caught before they affect a live show.

The community and vendors have converged on this practical mitigation. A permanent Microsoft patch is expected, but for your next live event, you should plan on using the workaround unless you can fully validate a fix in advance. Print the operational checklist, walk your team through the steps, and test everything in a rehearsal. With the Receive Mode switched, your NDI workflows will be back to smooth, reliable operation—and your systems will remain protected.