Microsoft’s latest cumulative update for Windows 11 24H2, KB5063878 (OS Build 26100.4946), is causing severe disruptions in NDI-based live production workflows. The August 12, 2025 Patch Tuesday rollout introduces a networking regression that triggers stuttering, lag, and choppy audio/video for broadcasters, streamers, and AV teams who rely on NDI’s high-bandwidth IP video transport. Within days of release, reports flooded in from production studios and independent creators, prompting Microsoft to add a Release Health advisory acknowledging the issue and pointing to a vendor-supplied workaround.

The advisory, posted on Microsoft’s support site, confirms “severe stuttering, lag, and choppy audio/video” when using NDI streams between machines that have installed the August updates. The problem is especially pronounced when the source uses Display Capture in tools like OBS. Microsoft has not yet published a root-cause analysis, but the symptoms and the immediate mitigation are well-documented by both the NDI vendor and independent testers.

The August Update and Its Fallout

KB5063878, released as part of the August 2025 security rollup, includes fixes and quality improvements alongside a servicing stack update. However, it also introduces a change that destabilizes NDI’s default transport mechanism. NDI is a widely adopted IP video protocol in live production, remote contribution, lecture capture, and multi-PC streaming rigs. It delivers high-quality, low-latency video and audio across local networks, enabling workflows that combine capture PCs, encoder machines, graphics engines, and software like OBS or vMix. Consequently, the regression has an outsized operational impact, even though it does not affect consumer video playback on services like Netflix or YouTube.

Microsoft deployed the update on August 12, and by August 14, community forums and social media buzzed with reports of sudden performance drops. Producers described smooth video for a few seconds, then abrupt stutter, audio drift, and dropped frames — exactly the sort of intermittent failure that can ruin a live show and is notoriously difficult to diagnose.

Symptoms and Scope

The observable failure mode is consistent across affected systems: severe stuttering, frequent frame drops, noticeable lag, and desynchronized audio. Microsoft’s advisory notes that the problem is most visible when using Display Capture in OBS, but reports extend to any NDI flow where the default Reliable UDP (RUDP) receive mode is active.

The scope encompasses Windows 11 version 24H2 after installing KB5063878 (OS Build 26100.4946). A companion update for Windows 10, KB5063709, appears to trigger similar behavior on some servicing branches. Affected users are predominantly those with two-PC NDI workflows — for example, a gaming or capture PC sending a Display Capture feed to an encoder PC for mixing and streaming. Single-PC setups that encode locally and push directly to a platform are generally unaffected because they do not rely on NDI’s PC-to-PC transport.

Inside the Break: RUDP vs. UDP vs. TCP

To understand why the update broke NDI and why the workaround works, it’s necessary to look at NDI’s transport options. NDI 5 defaults to Reliable UDP (RUDP), an application-layer protocol built on top of standard UDP. RUDP adds sequencing, retransmission, congestion control, and pacing to offer a balance between low-latency delivery and packet recovery. It avoids the head-of-line blocking inherent in TCP while providing more reliability than plain UDP.

When RUDP behaves as designed, it delivers smooth, low-latency video even on imperfect networks. The problem arises because RUDP depends on specific timing, buffering, and socket behaviors from the operating system’s network stack. A cumulative update can alter kernel-level scheduling, socket buffering, or packet-timing heuristics in ways that disrupt RUDP’s assumptions. If the sender or receiver detects spurious packet losses due to these changes, it may trigger aggressive retransmissions — a retransmission storm — and the stream degrades into severe stutter even on lightly loaded local area networks.

Microsoft has not detailed exactly what changed in the TCP/IP stack, but the vendor guidance and community testing clearly show that the issue is isolated to RUDP. Flows using legacy UDP or Single TCP remain stable. Legacy UDP offers no retransmission; if the LAN is clean (as it often is within a single switch), it continues to deliver packets without the RUDP-related side effects. Single TCP provides reliable, ordered delivery with slightly higher latency and potential head-of-line blocking, but its well-established behavior is unaffected by the update.

The Workaround: Switching NDI Receive Mode

Both Microsoft and NDI’s vendor documentation recommend switching the NDI Receive Mode from the default RUDP to either Single TCP or UDP (Legacy). This change does not require uninstalling the security update, so it preserves all August patches while restoring production stability for most environments.

To apply the workaround on each receiving machine:

  1. Confirm that KB5063878 is installed (check Settings → Windows Update → Update history, or run winver to see OS Build 26100.4946).
  2. Download and install NDI Tools for Windows if not already present.
  3. Launch NDI Access Manager.
  4. Open the Advanced tab.
  5. Under Receive Mode, change the setting from RUDP to Single TCP or UDP (Legacy).
  6. Click OK.
  7. Restart any NDI-receiving applications — OBS, Studio Monitor, vMix — because NDI only reads configuration at application start.

Repeat these steps on every machine that receives NDI flows. The senders do not need reconfiguration; the change affects only how the receiver handles incoming streams. After switching, perform a full rehearsal under live conditions to validate latency, lip-sync, and switching behavior.

Trade-offs and Risks

Choosing Single TCP or UDP (Legacy) is a temporary fix, not a perfect replacement for RUDP. Each alternative carries performance implications:

  • Single TCP: The safest universal fallback. It provides reliable delivery but can introduce a few milliseconds of additional latency. During bursts of packet loss, TCP’s head-of-line blocking may cause momentary pauses that are still far less disruptive than the stutter seen with broken RUDP.
  • UDP (Legacy): Preserves low latency on clean networks but offers no retransmission. Any packet loss will cause visible artifacts or frame drops. This mode is best only if the local network is thoroughly tested and known to be loss-free.

For most production teams, Single TCP is the recommended interim choice because it eliminates the severe stutter while maintaining acceptable quality.

Operational complexity also weighs heavily. Coordinating Receive Mode changes across dozens or hundreds of endpoints — especially remote or guest-controlled machines — demands change-control windows and thorough testing. Rolling back the security update is an alternative only when reconfiguration is infeasible, but it removes critical August security patches, including fixes for several Kerberos vulnerabilities. Microsoft strongly advises against uninstalling the update, and any rollback should be explicitly authorized with compensating controls in place (e.g., network segmentation, additional monitoring).

Collateral Damage?

While the primary focus is NDI, several outlets have reported separate regressions from the same August rollup, including failures in recovery tools and SSDs disappearing under heavy I/O. These claims are not uniformly reproducible and may depend on specific hardware and firmware. Microsoft has not acknowledged these as widespread issues, so treat storage-related anomalies as a distinct track requiring device-specific validation from OEMs. For NDI users, the immediate priority is the Receive Mode change.

What’s Next and Recommendations

Microsoft’s engineering team is investigating the root cause. A future servicing update or Known Issue Rollback should restore RUDP stability without requiring NDI configuration changes. The NDI vendor may also release a Tools update that adds automatic fallbacks or tuning to tolerate the altered Windows behavior.

In the meantime, studios, AV managers, and IT teams should adopt these practices:

  • Establish a production ring: Delay broad deployment of cumulative updates on show-critical NDI hosts until a short soak period proves stability with representative hardware and network topologies.
  • Document configurations: Record whether endpoints use RUDP, UDP, or Single TCP, so rapid remediation is possible if future updates change behavior again.
  • Automate where feasible: The NDI Access Manager configuration is stored in %programdata%/NDI/ndi-config.v1.json. Use management tools to deploy the desired Receive Mode and orchestrate application restarts during maintenance windows.
  • Monitor Release Health: Microsoft’s Release Health page is the authoritative source for known issue status. Subscribe to official update channels and vendor advisories.
  • Keep rollbacks as a last resort: Document the risk decision and implement compensating security controls if a rollback is unavoidable for a critical live event.

The August 2025 NDI stutter incident underscores a persistent challenge: cumulative security updates can silently destabilize specialized real-time protocols, even when the broader ecosystem remains unaffected. For production teams that depend on flawless NDI performance, a disciplined update ring, automated configuration management, and a rehearsed fallback plan are no longer optional — they are essential.

Practical checklist for immediate action:

  1. Confirm KB5063878 (Windows 11 24H2) or KB5063709 (Windows 10) on production hosts.
  2. If NDI is used, change NDI Access Manager → Advanced → Receive Mode to Single TCP and restart receiving applications.
  3. Run a full rehearsal to verify latency and lip-sync.
  4. If remote endpoints cannot be reconfigured, evaluate a controlled rollback only with explicit authorization and compensating controls.
  5. Stage future updates in a pilot ring to avoid show-day surprises.