Microsoft has delivered a fix for the NDI streaming performance regression that plagued video production workflows following the August 2025 Patch Tuesday security updates. The resolution arrives via September’s cumulative updates—KB5065426 for Windows 11 and KB5065429 for Windows 10—officially eliminating the severe stuttering, dropped frames, and audio desynchronization that forced broadcasters, streamers, and AV teams to implement temporary workarounds for nearly a month.

The August 12, 2025 rollups (KB5063878 for Windows 11 24H2 and KB5063709 for Windows 10 22H2/21H2) patched over a hundred vulnerabilities, including a publicly disclosed Kerberos elevation-of-privilege flaw. But they also broke NDI—the widely used IP video transport protocol—when streams relied on its default Reliable UDP (RUDP) mode. Microsoft quickly acknowledged the issue on its Release Health dashboard, classifying it as “Confirmed” and pointing users to vendor-provided mitigation steps.

“The result is severe stability issues for all NDI sources on the affected computer,” noted vMix in its knowledge base, describing laggy video, low frame rates, and massive frame drops. The problem struck even on lightly loaded gigabit LANs, which typically sail along with single-digit milliseconds of latency. For live event producers, the timing was disastrous: many had just applied the critical security patches, only to discover their multicamera setups and remote contributor feeds had become unusable.

NDI underpins modern video-over-IP productions, from small game streamers using OBS to full-scale broadcast operations. A typical setup captures video on one PC via Display Capture and transports it across the network to a mixing or encoding rig. The protocol’s low latency and decent image quality over commodity hardware made it a de facto standard. When those feeds began to stutter dramatically after a routine Windows update, the impact rippled across the industry.

At the heart of the trouble was NDI’s default transport, RUDP. To understand the breakage, it helps to look at how the three available NDI transports differ:

  • UDP (Legacy): Simple, unreliable datagrams. No retransmissions or ordering, meaning lost packets create visible artifacts. Latency is extremely low, but any network hiccup takes a toll. Best on pristine, dedicated LAN segments.
  • Single TCP: The classic ordered, reliable byte stream. It retransmits lost data and adjusts to congestion, but head-of-line blocking can introduce bursts of lag—a dealbreaker for fast-paced live switching.
  • RUDP (Reliable UDP): An application-level reliability layer that layers selective retransmission and pacing over bare UDP. It aims to preserve UPD’s speed while cleaning up occasional packet loss, making it the preferred mode for most NDI installations.

RUDP achieves this balance by maintaining precise timing between packets, sending acknowledgements, and retransmitting only what’s necessary. That intricacy makes it acutely sensitive to changes in the underlying OS networking stack. If kernel buffers, socket scheduling, or timer behavior shifts—perhaps due to a security patch adjusting low-level code—RUDP’s congestion window, retransmission logic, and pacing heuristics can fall out of step. The result is not a loss of connectivity but a pathological stutter: periodic stalls, retransmission storms, or audio drift that appears even when bandwidth utilization remains low.

Microsoft has not published a full root cause analysis, but the symptom pattern strongly points to a transport-layer interaction. Only RUDP flows suffered; switching to UDP (Legacy) or Single TCP immediately restored smooth video. That exclusivity, and the fact that the problem reproduced on networks with negligible packet loss, suggests the August updates inadvertently altered a kernel behavior that RUDP critically depends on.

Within days of the reports surfacing, NDI’s developers and Microsoft aligned on a temporary workaround: change the Receive Mode on all affected receiving machines. The steps, detailed by vMix and echoed in Microsoft’s Release Health note, were straightforward:

  1. Download and install the free NDI Tools pack from ndi.video/tools if not already present.
  2. Launch NDI Access Manager.
  3. Switch to the Advanced tab.
  4. Under Receive Mode, select Single TCP or UDP (Legacy).
  5. Click OK and restart any NDI-receiving applications (OBS, vMix, Studio Monitor) so they pick up the new setting.

Single TCP became the recommended universal fallback because it reliably eliminated the stutter. However, it introduced a trade-off: head-of-line blocking could add latency jumps under suboptimal network conditions, which crews accustomed to ultra-tight switching tolerance found problematic. UDP (Legacy) kept latencies low but left feeds unprotected against packet loss. Production teams had to choose their poison, often redesigning show flows to accommodate the less forgiving transport.

For those who couldn’t reconfigure every remote contributor’s machine—a common scenario in live streaming with on-camera guests—some chose to uninstall the August security updates entirely. That decision invited its own risks, since KB5063878 and KB5063709 closed 107 to 119 vulnerabilities (counts vary across sources), including a zero-day under active exploitation.

The good news arrived on September 9, 2025. Microsoft released KB5065426 (Windows 11) and KB5065429 (Windows 10) that directly address the NDI regression. The vMix knowledge base confirms: “As of 9th of September 2025 this issue is resolved if you have the following Windows Updates installed.” NDI’s official FAQ now points to the same fix, signaling that the RUDP transport can again be used safely. Users can verify installation by checking Settings > Windows Update > Update History; after installing the September cumulative, they can revert NDI Access Manager’s Receive Mode back to its default (RUDP) or leave it on the workaround if they prefer.

For production environments that deferred the August patches to keep shows running, the path forward is clear: install the September rollups immediately. They contain all the August security fixes and the NDI correction, so there’s no longer a trade-off between safety and performance. As always, a rehearsal under full load before a live event is essential—even with the fix confirmed, configuration drift in a complex workflow can still bite.

The NDI mess was not the only headache from August Patch Tuesday. Microsoft also had to rush out-of-band fixes for “Reset this PC” failures and cloud recovery regressions. Enterprises pushing updates via WSUS hit installation errors with code 0x80240069. And parallel reports emerged of NVMe SSDs—particularly some Phison-controlled drives—disappearing or becoming inaccessible after large sequential writes on KB5063878. While storage issues remain under active investigation and vary by firmware, they added to the operational anxiety for IT departments managing fleets with production hardware.

That broader context underscores a recurring dilemma for Windows administrators: monthly rollups address critical security flaws but sometimes at the cost of introducing new bugs. The NDI case illustrates a best-practice response from Redmond: rapid acknowledgment, collaboration with vendors on a stopgap, and a targeted fix delivered in the next monthly cycle. Still, the month-long gap forced live event crews to grapple with imperfect mitigation, highlighting the need for robust testing rings, especially in production environments where downtime translates to lost revenue.

Looking ahead, NDI developers may introduce hardening to prevent similar regressions—perhaps automatic fallback logic or better resilience to OS-level timing changes. Microsoft’s Release Health dashboard remains the authoritative spot for tracking such issues. Production teams should also subscribe to vendor advisories from NDI, vMix, and OBS to catch early warnings before a Patch Tuesday update reaches their critical systems.

For now, the immediate takeaway is simple: update to the September cumulative. The stuttering that plagued live feeds throughout August is resolved, and the RUDP transport that keeps video crisp and in sync is once again trustworthy. The incident serves as a potent reminder that even the most battle-tested protocols can falter when the ground shifts beneath them—and that a well-rehearsed fallback, combined with timely official fixes, keeps the show on the air.