Microsoft has officially confirmed that its August 12, 2025 cumulative updates are causing severe stuttering, lag, and choppy audio/video in NDI-based streaming setups, particularly affecting OBS Studio users with Display Capture enabled. The company linked the regression to NDI’s default RUDP transport and pointed to a straightforward workaround: switch NDI Receive Mode to TCP or UDP on affected PCs.

Streamers and broadcast engineers began reporting erratic NDI performance almost immediately after the Patch Tuesday drop. Systems that had installed Windows 11 24H2’s KB5063878 or Windows 10 22H2/21H2’s KB5063709 exhibited jumpy feeds, broken lip-sync, and frame pacing so bad that productions became unreliable. Microsoft added the issue to its Windows Release Health tracker on August 21, confirming the problem while it investigates a permanent fix.

What’s Broken and What’s Not

The symptom list is specific: “Severe stuttering, lag, and choppy audio/video might occur when using NDI,” reads Microsoft’s advisory. It explicitly calls out OBS Studio and NDI Tools, and notes that “Display Capture” on the source PC makes the condition dramatically worse. Even low-bandwidth links can’t escape the jitter, pointing to a timing or buffering problem rather than raw throughput.

  • Impacted: NDI streams using the default RUDP transport after installing KB5063878 (Windows 11 24H2) or KB5063709 (Windows 10 22H2/21H2). Symptoms peak when Display Capture is in the encoder scene.
  • Not impacted: NDI flows using UDP (Legacy) or Single TCP transports. Swapping to either mode restores stability without removing the security updates.
  • Affected apps: OBS Studio and NDI Tools are named by Microsoft; any application relying on NDI over RUDP is potentially vulnerable.

The Culprit: RUDP Transport

NDI supports three transport protocols to balance latency and reliability. RUDP (Reliable UDP) is the modern default, adding application-layer reliability and pacing on top of UDP to avoid head-of-line blocking while keeping delays low. UDP (Legacy) sends raw, unacknowledged datagrams—fastest but without retransmissions. Single TCP is the most conservative, ordered, and reliable option but can suffer from head-of-line blocking under loss.

Microsoft’s release health entry and NDI’s vendor documentation both state that only RUDP connections are affected. While no root cause has been published, the restriction hints that something in the August update—possibly a subtle timing, buffering, or scheduler change in the networking stack or kernel—throws RUDP’s congestion and retransmission heuristics off-balance. That would explain why the issue appears even on uncongested LAN links and why TCP and plain UDP remain clean.

The Official Workaround: Switch NDI Receive Mode

The fix, which Microsoft directly endorses, requires no rollback of the security patches. Instead, users change the NDI Receive Mode away from RUDP on every receiving PC. NDI maintains a step-by-step guide, but here is the fast track:

  1. Download and install the free NDI Tools pack.
  2. Launch NDI Access Manager on the receiving computer.
  3. Open the Advanced tab.
  4. Change Receive Mode from RUDP (default) to either Single TCP or UDP.
  5. Click OK, then restart every NDI-receiving application (OBS Studio, Studio Monitor, vMix) so they read the updated setting.
  6. Repeat on other PCs that participate in the NDI workflow.

This can be done in minutes. If you manage distributed teams or remote guests, coordinate the change during a rehearsal or maintenance window. The configuration is reversible once Microsoft ships a fix.

Choosing Between Single TCP and UDP

Both alternatives work, but they present different trade-offs:

  • Single TCP: More reliable across complex networks and gear that mishandles UDP variants. It adds a few milliseconds of latency and can introduce head-of-line blocking if packet loss occurs. A safe first choice when you need to stabilize quickly.
  • UDP (Legacy): Lowest latency on a clean, dedicated LAN. No retransmissions mean any packet drop is visible unless your workflow buffers it. Best suited for environments where you control the switching fabric and can guarantee minimal loss and jitter.

Whichever you choose, retest your scenes—especially those with Display Capture—and validate lip-sync and frame pacing before going live.

Don’t Uninstall the Security Updates

The August patches fixed over 100 vulnerabilities, including CVE-2025-53779, an actively exploited Kerberos flaw that could allow privilege escalation to domain admin. Removing KB5063878 or KB5063709 reopens that and other critical holes. For environments that use Active Directory, the risk is too great to accept casually. If a rollback is absolutely unavoidable for a mission-critical live event, follow strict change-control procedures and reinstall the security updates immediately afterward. On Windows 11 24H2, combined servicing stack and cumulative update packages cannot be removed with the wusa.exe /uninstall command; you must use DISM /Remove-Package with the exact LCU package identity—an advanced operation that should be tested first.

August 2025: A Stormy Patch Tuesday

The NDI regression is not an isolated incident. Within days of the August 12 release, Microsoft also acknowledged a bug that broke Reset this PC and other recovery operations on some Windows 10 installations and older Windows 11 builds. An out-of-band update addressed those recovery failures. Separately, a WSUS delivery failure (error 0x80240069) for KB5063878 was resolved mid-cycle. Independent testers and media outlets have also reported storage anomalies on certain SSDs under heavy I/O after KB5063878, though Microsoft has not formally attributed a root cause there. The pattern of multiple high-impact issues in a single patch wave underscores the need for careful update staging, especially on production-critical machines.

Recommendations for Streamers and IT Admins

  • Apply the workaround immediately. Switching NDI Receive Mode to TCP or UDP restores streaming stability without sacrificing the crucial August security fixes.
  • Avoid Display Capture temporarily if possible, or at least deploy the workaround before using it. Window or game capture tends to be less sensitive to OS-level changes.
  • Stage your updates. Keep a “production ring” for broadcast and control-room PCs that receives cumulative updates only after a soak period on noncritical machines.
  • Document your NDI transport choices. For each studio or classroom, record whether Single TCP or UDP is in use, and why. This speeds troubleshooting if conditions change.
  • Monitor Windows Release Health. Microsoft has been updating its status pages quickly this month; a fix or Known Issue Rollback could arrive before the next Patch Tuesday.
  • For fleets: Inventory which production paths use NDI over RUDP. Push an operational advisory instructing staff to change Receive Mode and restart apps. Avoid wholesale uninstall of security updates; if rolling back, isolate devices and schedule immediate repatching.

What’s Next

Microsoft says it is investigating. Given the clear transport boundary (RUDP only) and cross-version impact, the permanent fix could land in the networking stack, kernel scheduling, or a compatibility shim tuned specifically for NDI’s RUDP behavior. NDI’s vendor documentation notes that the problem is Windows-side, so users should keep an eye on both cumulative updates and any NDI software adjustments that might add adaptive logic for changed platform behavior. An out-of-band fix or a documented resolution in the next cumulative update is plausible. Until then, the Receive Mode switch remains the safe route.