August 2025 Windows Updates Break NDI Streaming; Microsoft Urges RUDP Workaround

Live production crews and streamers are scrambling after Microsoft’s August 2025 cumulative updates inflicted severe stuttering, lag, and choppy audio on NDI-based video transport—a critical protocol used in multi-PC broadcast workflows. The root cause zeroes in on NDI’s default Reliable UDP (RUDP) transport, which becomes practically unusable after installing KB5063878 (Windows 11 24H2) or KB5063709 (Windows 10 21H2/22H2). Microsoft confirmed the regression late on August 15 and pointed to a vendor-recommended mitigation: switch NDI receivers to Single TCP or legacy UDP. No permanent patch has been shipped, leaving thousands of studios, corporate AV teams, and independent streamers to weigh a simple configuration tweak against the nuclear option of rolling back the security update.

What Broke and When

The trouble started on August 12, 2025—Patch Tuesday—when Microsoft deployed broad cumulative updates to Windows 11 24H2 (KB5063878, build 26100.4946) and Windows 10 21H2/22H2 (KB5063709). These rollups closed over 100 security vulnerabilities, including a zero‑day Kerberos elevation‑of‑privilege flaw, but within 48 hours video professionals noticed catastrophic frame drops and audio desyncs in NDI streams. The issue affects any application relying on NDI, most prominently OBS Studio, NDI Tools, and vMix, especially when Display Capture is used on the source machine.

Microsoft’s formal acknowledgment came in a Windows Release Health advisory that describes “severe stuttering, lag, and choppy audio/video” and explicitly ties the fault to NDI’s RUDP connections. “Traffic sent or received using UDP or Single‑TCP remains unaffected,” the advisory notes, confirming that the regression lives entirely inside the RUDP path. The NDI development team corroborated the finding, stating that RUDP traffic “dropped” after the August patch was applied. Independent outlets including BleepingComputer and Tom’s Guide reproduced the behavior on patched test beds, verifying that the workaround—changing the Receive Mode in NDI Access Manager—immediately restores clean performance.

Why RUDP, and Why Now?

NDI‘s architecture supports three transport modes to balance latency, loss resilience, and complexity:

• RUDP (Reliable UDP): The modern default. It layers application‑level sequencing, retransmissions, and pacing atop raw UDP to deliver low‑latency video with fewer visual artifacts. This hybrid design makes it sensitive to subtle changes in the OS networking stack—packet scheduling, timer precision, and buffer handling.
• UDP (Legacy): Simple datagrams with no reliability. It’s extremely low latency on lossless LANs but falls apart under any packet drops.
• Single TCP: Fully reliable, ordered stream. It eliminates stutter but can introduce head‑of‑line blocking and variable latency when the network hiccups.

The fact that only RUDP fails while UDP and Single TCP sail through strongly suggests that the August update altered a path‑specific timing or queuing behaviour deep in the Windows TCP/IP stack. Microsoft has not published a root‑cause analysis, but informed speculation from networking engineers points to modifications in the NDIS (Network Driver Interface Specification) or the kernel’s callback scheduling that upset RUDP’s delicate retransmission pacing. Until Redmond releases a technical bulletin, all causal statements remain provisional.

Who Hurts and How Bad

The blast radius is narrow in absolute numbers but outsized in operational severity. NDI is the de facto standard for multi‑machine live production: a capture PC sends pristine 4K video over gigabit Ethernet to a dedicated encoder, graphics machine, or switcher—all without the cost and weight of SDI cables. Affected environments include:

• Live broadcasters: Studios running control rooms where every video layer comes in via NDI.
• Corporate AV teams: Town halls, all‑hands meetings, and hybrid conferences that depend on remote capture and monitoring.
• Esports and gaming streamers: Those who split gameplay, camera, and overlays across two PCs.
• Educational and training facilities: Lecture capture systems rely on NDI for low‑cost, high‑quality feeds.

In the field, symptoms manifest as repeated micro‑freezes every few seconds, video glitching, and audio popping—even on pristine, uncongested LANs. For a live event, a single half‑second freeze can ruin a critical on‑air moment. For post‑production, the corrupted frames force a complete re‑ingest. “We had to cancel a live stream 20 minutes before air because every source hit the same stutter,” one broadcast engineer posted to the Windows Forum. “Switching to Single TCP on five receiving machines took an hour of frantic reconfiguration.”

The timing is cruel: the August patch is a cumulative security update, meaning any uninstalled system loses protection against dozens of vulnerabilities, including the publicly disclosed Kerberos exploit. Production managers are caught between a broken video pipeline and a known security risk.

Immediate Mitigation: Change the Receive Mode

Microsoft and NDI’s developer (Vizrt) jointly recommend the following low‑impact workaround, which preserves the security update and can be applied without rebooting Windows:

1. Install or open NDI Access Manager on every receiving machine. It’s part of the free NDI Tools pack.
2. Navigate to the Advanced tab.
3. Change Receive Mode from RUDP (default) to Single TCP or UDP (Legacy).
4. Click OK and restart all NDI‑receiving applications (OBS, vMix, NDI Studio Monitor, etc.). The applications pick up the new transport mode only at launch.

Which fallback to choose depends on the environment:

In practice, most production networks are engineered to handle TCP’s characteristics well, so Single TCP is the safer choice for mixed or complex LANs. Teams should run a test stream after the switch to verify that latency and stability meet their requirements. Note that the change must be applied on every receiving machine, not the source.

If reconfiguration is impossible—for example, when remote contributors cannot access NDI Access Manager—an organisation faces a hard trade‑off:

• Stay patched: Accept the video degradation or route around NDI entirely.
• Uninstall the August LCU: This instantly restores normal RUDP performance but removes all security fixes delivered in the August 2025 rollup. Uninstallation must be matched with compensating controls: isolate the host on a tightly controlled VLAN, enforce strict firewall rules, and accelerate deployment of any future fix.

Microsoft also released out‑of‑band (OOB) updates on August 19, 2025, for a separate critical bug that broke Windows’ Reset and Recovery features. KB5066189 (OS Builds 22621.5771 and 22631.5771) fixes that specific recovery failure and does not address the NDI RUDP regression. Administrators who were affected by both issues should apply the OOB for recovery while also applying the NDI transport workaround.

Microsoft’s Response and What to Expect Next

Microsoft moved quickly to acknowledge the bug and issue OOB fixes for the unrelated recovery problem, but it has not yet delivered a targeted patch for the NDI stutter. The Release Health advisory states that “Microsoft is investigating” and directs users to the vendor guidance. Historically, such transport‑layer regressions require deep stack analysis and careful regression testing, so a permanent fix could take weeks.

In the meantime, engineering teams should:

• Monitor the Windows Release Health dashboard for any update to the known issue.
• Watch the NDI knowledge base for vendor‑verified guidance.
• Test every new cumulative update in a staging environment with real NDI traffic before deploying to production hosts.
• Keep the switch to Single TCP or UDP in place until Microsoft explicitly declares the bug fixed. Do not revert prematurely.

Lessons for the Pro AV World

This episode underscores a growing tension in modern operating systems: massive monthly security rollups inevitably introduce regressions in specialized, performance‑sensitive domains. NDI’s dependence on a custom reliability layer magnifies the risk. As more professional video workflows move to IP‑based transport, both platform vendors and application developers need tighter cooperation.

For production teams, the incident is a loud reminder to formalize update validation procedures. Even a simple test stack—a spare PC running OBS with NDI Source and a one‑minute looped sequence—can catch show‑stopping bugs before they reach the control room. Pair that with a written rollback plan that has management sign‑off, and you convert panic into a controlled procedure.

The Bottom Line

The August 2025 Windows updates delivered necessary security patches but inadvertently choked one of the most important protocols in live video production. The fix—swapping NDI’s transport from RUDP to Single TCP or UDP—is straightforward, free, and does not compromise security. While Microsoft engineers hunt for the root cause, production teams can keep their streams live and their systems protected by making one configuration change per receiving machine. The real cost is not technical but logistical: the time, stress, and lost confidence that ripple through every broadcast studio when a trusted toolchain breaks without warning.