Windows August 2025 Update Triggers Severe NDI Stutter (RUDP)

Background​

What Microsoft and Vendors Confirmed​

• Microsoft’s Release Health entry describes the issue as “Problems with NDI streaming performance after installing the August 2025 update,” calling out severe stuttering, lag and choppy audio/video, and noting the effect is especially visible when the source uses Display Capture in tools such as OBS. (support.microsoft.com)
• Microsoft identified the interaction with NDI’s RUDP transport as the triggering pathway and pointed users to vendor guidance to change NDI’s Receive Mode to Single TCP or UDP (Legacy) as a temporary mitigation while engineering investigates. (bleepingcomputer.com, windowsforum.com)
• Independent reporting from industry outlets confirmed the same symptom pattern and mitigation approach, emphasizing the regression’s reproducibility across affected KBs and builds. (tomsguide.com, bleepingcomputer.com)

Technical primer: RUDP, UDP, and Single TCP — why transport choice matters​

• RUDP (Reliable UDP) — NDI’s modern default in many setups. It attempts to combine UDP‑style low latency with application‑level reliability (sequencing, retransmits, pacing) to reduce visual artifacts while keeping latency low.
• UDP (Legacy) — Simple datagrams sent without retransmission. Lowest latency in pristine LAN conditions but susceptible to packet loss.
• Single TCP — Ordered, reliable stream with retransmission. Robust but can suffer head‑of‑line blocking when packet loss occurs, increasing latency.

Timeline and Microsoft’s response​

• August 12, 2025 — Microsoft released its August cumulative updates (Patch Tuesday), including KB5063878 for Windows 11 24H2 and companion KBs for Windows 10. The rollups included many security fixes but were rapidly followed by community reports of regressions.
• Mid‑August 2025 — Reports from streamers, broadcasters, and vendor telemetry converged on a reproducible NDI stutter pattern caused by RUDP connections. Industry outlets documented the issue and NDI’s vendor guidance. (bleepingcomputer.com, tomsguide.com)
• Microsoft posted a Release Health advisory confirming the behavior, identified the transport fingerprint (RUDP only), and linked to vendor guidance recommending a Receive Mode change. Engineering is investigating. (support.microsoft.com)
• August 18–19, 2025 — Microsoft published out‑of‑band (OOB) cumulative updates to address related but separate regressions that affected Reset / Recovery flows (for example, KB5066189 for Windows 11 servicing families). Those OOB packages fixed recovery failures introduced by certain August patches; they do not claim to remediate the NDI RUDP regression specifically. Administrators were advised to apply the OOB updates if they were experiencing recovery failures. (support.microsoft.com)

Impact: who is hurt, and how badly​

• Live broadcasters and production studios that split capture and encode across machines.
• Corporate AV teams delivering lecture capture, town halls, or hybrid events.
• Streamers using separate capture and encoder PCs and those layering multiple high‑resolution sources.
• Education and training environments that use remote capture and monitoring via NDI.

• Repeated micro‑pauses and stuttering in live video layers.
• Audio desynchronization and choppy audio in feeds.
• Persistent failures even on low‑utilization, low‑loss LANs (indicating timing or retransmission pathology rather than congestion). (bleepingcomputer.com, windowsforum.com)

Practical mitigations and step‑by‑step remediation​

• Install or open NDI Access Manager on every receiving machine.
• In NDI Access Manager, go to the Advanced tab.
• Change Receive Mode from RUDP (default) to Single TCP or UDP (Legacy).
• Click OK, then restart any NDI‑receiving applications (OBS Studio, vMix, Studio Monitor, etc.). The apps read the Access Manager configuration on startup. (windowsforum.com)

• Single TCP: This typically eliminates stutter and is the safest fallback for complex network environments. It can add end‑to‑end latency and suffer head‑of‑line blocking during packet loss, so it’s best on reliable LANs with sufficient latency budget.
• UDP (Legacy): Keeps latency low on pristine LANs but forfeits RUDP’s retransmission/ordering benefits — acceptable on dedicated Ethernet segments.
• Restarting processes is required; changing the Receive Mode while applications are running will not take effect until the apps relaunch. Plan restarts during rehearsals or maintenance windows. (windowsforum.com)

• Consider uninstalling the August LCU on affected hosts as a last resort, but weigh the security implications carefully. Rollbacks may be acceptable for short maintenance windows but should be accompanied by compensating controls. (tomsguide.com)
• For other, unrelated August regressions (such as Reset/Recovery failures), apply Microsoft’s OOB fixes (for example, KB5066189 for Windows 11) if you are affected; they address separate, high‑impact recovery problems. Do not assume OOBs cure the NDI RUDP regression without explicit Microsoft confirmation. (support.microsoft.com)

Operational guidance for IT and AV teams​

• Test in a staging environment first. Reproduce the issue and the remediation (transport change) before the next live event or classroom session.
• Document all changes. If switching to Single TCP or UDP (Legacy), log the change and maintain a rollback plan.
• Coordinate with remote contributors. If you rely on external feeds, provide configuration instructions and a rehearsal window to apply changes and restart apps.
• Balance security and availability. Uninstalling the August LCU restores the previous NDI behavior but removes important security fixes. If you choose rollback, isolate the host or enforce strict network controls until the patch is replaced.
• Monitor Microsoft Release Health and vendor advisories. Microsoft is investigating and may publish a targeted patch; apply vendor‑tested updates when they appear rather than guesswork. (support.microsoft.com, bleepingcomputer.com)

Risk assessment and trade‑offs​

• Security risk vs. operational availability. Removing a cumulative security update is never a trivial decision. The August rollups patched a large set of vulnerabilities; one such example closed a publicly disclosed Kerberos elevation‑of‑privilege issue in the same release. For many organizations, the security risk of remaining unpatched outweighs temporary media quality problems.
• Workaround operational cost. Changing Receive Mode is low‑risk technically but has operational cost: per‑machine configuration, app restarts, and potential latency impacts on timing‑sensitive productions.
• Uncertain root cause. Community analysis and Microsoft’s note narrow the fault to the RUDP transport, but the precise kernel/stack interaction has not been publicly dissected by Microsoft. Until a root‑cause bulletin or fix is published, long‑term planning must assume uncertainty. Flag any definitive causal statements as provisional.

Why this matters beyond streamers​

• The complexity of shipping large cumulative updates across diverse hardware and software stacks means regressions can surface in narrow but critical workflows.
• Low‑latency protocols that implement custom reliability (like RUDP) are intrinsically more fragile to platform timing and scheduler tweaks.
• The cost of a rollback is not just technical — it’s organizational and security‑oriented.

Cross‑checks and verification​

• Microsoft’s Release Health and KB documentation describing the known issue and the OOB packages. (support.microsoft.com)
• BleepingComputer’s reporting of the NDI/RUDP fingerprint and the NDI vendor guidance recommending Receive Mode changes. (bleepingcomputer.com)
• Tom’s Guide and Windows Central confirmation of the issue and Microsoft’s acknowledged response, including guidance on removal or mitigation tradeoffs. (tomsguide.com, windowscentral.com)

Recommended checklist for streamers and admins (short, actionable)​

• Back up configuration and scene collections before making changes.
• Update NDI Tools to the latest vendor release.
• Change NDI Receive Mode to Single TCP or UDP (Legacy) on receiving hosts and restart receiving applications. (windowsforum.com)
• If you experience Reset/Recovery failures, apply Microsoft’s OOB KBs (for example, KB5066189) for affected servicing families. Verify the specific KB against your OS build before installing. (support.microsoft.com)
• Keep affected hosts isolated from high‑risk networks if you choose to uninstall the August LCU temporarily.
• Monitor Microsoft Release Health for a permanent fix and vendor advisories for tested guidance.

Longer view: lessons for vendors and enterprise​

• Vendors that rely on nonstandard transport stacks (RUDP, custom reliability layers) should maintain a robust compatibility matrix and test harnesses that exercise kernel and network scheduler changes at scale.
• Platform vendors should consider targeted, staged rollouts with rapid telemetry funnels for specialized verticals (e.g., broadcasting and pro AV) where even small regressions are mission‑critical.
• IT teams should formalize update windows that include validation tests for critical workflows (live events, lecture capture, remote proctoring) before broad deployment.

Conclusion​