NDI Stutter After August 2025 Windows Update: Fix with Receive Mode Change

The August 2025 Windows 11 cumulative update (KB5063878, OS Build 26100.4946) introduced a networking regression that is causing severe stuttering, lag, and choppy audio/video in NDI-based streaming workflows — a problem Microsoft has confirmed and that currently requires configuration workarounds in NDI Tools and OBS to restore reliable performance. (support.microsoft.com)

Background​

Microsoft shipped the August 12, 2025 cumulative security update for Windows 11 — identified as KB5063878 (OS Build 26100.4946) — as part of the regular Patch Tuesday rollup. The package includes fixes and quality improvements, plus a servicing stack update. Within days of distribution, broadcasters, streamers, and AV teams reported consistent and reproducible performance regressions when using Network Device Interface (NDI) streams between machines. Those reports prompted Microsoft to add a Release Health advisory describing the behavior and referencing a vendor-supplied workaround. (support.microsoft.com, bleepingcomputer.com)
NDI is a widely used IP video protocol in live production, remote contribution, lecture-capture, and multi‑PC streaming rigs. It moves high-quality, low-latency video and audio across local networks and powers workflows that combine capture PCs, encoder machines, graphics engines, and OBS or vMix switchers. The regression therefore has an outsized operational impact even if it doesn’t affect consumer playback of content on services like Netflix or YouTube. (docs.ndi.video)

---

The essence of the problem​

• Symptom: Severe stuttering, dropped frames, lag, and choppy audio/video when using NDI flows between PCs that have installed the August 2025 updates. Microsoft says the problem is especially visible when the source uses Display Capture in tools such as OBS. (bleepingcomputer.com, tomsguide.com)
• Scope: Reported on Windows 11 version 24H2 after installing KB5063878 (OS Build 26100.4946), and in parallel on some Windows 10 servicing branches with companion updates (for Windows 10 the parallel patch is KB5063709). Microsoft marked the NDI regression as a confirmed Release Health issue while engineering investigates. (support.microsoft.com, windowsforum.com)
• Root trigger (as identified so far): The interaction occurs when NDI is using its default Reliable UDP (RUDP) transport mode for receive connections; traffic using legacy UDP or Single TCP has not shown the same failure pattern. Microsoft and NDI’s vendor guidance therefore centers on switching the NDI Receive Mode away from RUDP as a temporary mitigation. (bleepingcomputer.com, docs.ndi.video)

These facts — update ID, build number, symptom set, and vendor-recommended mitigation — are consistent across Microsoft’s Release Health page, NDI documentation, and independent reporting. (support.microsoft.com, docs.ndi.video, bleepingcomputer.com)

---

Why this matters to creators and production teams​

Live production workflows are fragile because they rely on a chain of timing-sensitive components: capture, network transport, mixing/encoding, and delivery. NDI was designed to keep video/audio synchronized and latency low across commodity LANs; RUDP is a transport engineered to preserve low latency while adding application-layer reliability.
When RUDP’s behavior is disrupted at the OS/network-stack level, the observable result is not an application crash but unpredictable buffering, retransmission storms, and repeated micro-pauses in the media stream — exactly the problem producers reported: smooth video for a few seconds, then sudden stutter, audio drift, and dropped frames. That pattern is far more disruptive to live shows than a single dropped frame or occasional jitter. (docs.ndi.video)
Key operational impacts include:

• Lip‑sync and timing breakage for multi‑camera shows.
• Repeated, unrecoverable frame drops that force reruns or show aborts.
• Difficulty diagnosing live because the problem appears as intermittent stutter rather than a straightforward loggable error.
• Administrative burden to reconfigure fleets of machines or to roll back security updates — both expensive and risky.

---

Technical deep dive: RUDP, UDP, TCP — what changed and why it breaks NDI​

What is RUDP and why NDI uses it​

NDI 5 defaults to Reliable UDP (RUDP): an application-level protocol layered atop UDP that adds sequencing, retransmission, congestion control, and pacing to provide a balance between low-latency delivery and packet recovery. RUDP aims to avoid head‑of‑line blocking inherent in TCP while avoiding the lossy behavior of plain UDP on imperfect networks. NDI’s official documentation describes RUDP as the recommended default for NDI 5 because it offers the best real-world performance across a broad set of networks. (docs.ndi.video)

Why a Windows update can destabilize RUDP​

RUDP depends on specific timing, buffering, and socket behaviors from the OS to correctly pace retransmissions and avoid retransmission storms. A cumulative update can change kernel network scheduling, socket buffering, or packet timing heuristics in ways that are subtle but consequential for protocol stacks built on top of UDP. If those platform-level assumptions shift, the sender or receiver might detect spurious losses, trigger aggressive retransmissions, or behave as though the network is congested — and the stream exhibits severe stutter even on low‑bandwidth LANs. Microsoft’s messaging implies a transport-layer interaction tied to RUDP, but a full root-cause analysis remains pending. Treat speculation about low-level internals as provisional until Microsoft publishes technical details. (docs.ndi.video)

Why UDP or Single TCP are less affected​

• UDP (Legacy): No retransmission logic. If the LAN is stable (as ISPs within a single switch often are), UDP will keep delivering packets and avoid RUDP’s retransmission-related side effects. The trade-off is that packet loss is unrecoverable and can visibly degrade video quality.
• Single TCP: Offers reliable, ordered delivery at the cost of potential head-of-line blocking and slightly higher latency. TCP’s behavior is stable and well understood by OS stacks, so it does not rely on RUDP-specific timing heuristics that appear affected by the update.

Vendors and Microsoft observed that flows using UDP or Single TCP remained unaffected in reproduced tests, which is why they recommend switching Receive Mode as an interim step. (docs.ndi.video)

---

Who’s affected — scope and risk profile​

• Primary impact: Two‑PC NDI workflows where one machine captures (often using Display Capture) and another receives that feed for encoding or mixing (for example, a capture PC sending a Display Capture to an encoder PC running OBS or vMix).
• Enterprise risk: AV labs, lecture‑capture deployments, esports arenas, and small production houses that use NDI across managed networks and cannot quickly reconfigure remote endpoints.
• Less impacted: Single‑PC hobby streamers who encode locally and submit a stream directly to a platform (those setups typically do not rely on NDI high‑bandwidth PC‑to‑PC transport).
• Geographic or hardware correlation: Reports have been widespread and reproducible across multiple setups; some side reports also mention SSD and recovery-tool regressions tied to the same August rollups, but those storage claims remain vendor-hardware-specific and should be validated at the device‑firmware level before acting. Flag: storage anomalies are reported but not uniformly reproduced. (tomshardware.com)

---

Microsoft and vendor responses (status summary)​

• Microsoft added a Release Health entry acknowledging “Severe stuttering, lag, and choppy audio/video” for NDI streaming after the August update and classified it as a known issue while engineering investigates. The company explicitly tied the problem to NDI’s default RUDP receive mode and recommended vendor guidance. (support.microsoft.com, bleepingcomputer.com)
• NDI (vendor) documentation and vendor KB articles (vMix, NDI docs) provide a step‑by‑step mitigation: change NDI Access Manager → Advanced → Receive Mode from RUDP to Single TCP or UDP (Legacy), then restart NDI-consuming applications so they re-read configuration. (docs.ndi.video, m.vmix.com)
• Independent outlets including BleepingComputer, Tom’s Guide, and others have reproduced the symptom set and corroborated Microsoft’s advisory. Those outlets also emphasize the operational tradeoffs of the workaround versus uninstalling updates. (bleepingcomputer.com, tomsguide.com)

---

Step-by-step mitigation (what to do now) — concise guide​

• Identify affected hosts:
• On each machine, open Settings → Windows Update → Update history → Uninstall updates to confirm whether KB5063878 (Windows 11 24H2) is installed. Or run winver to check OS Build 26100.4946. (support.microsoft.com)
• Preferred fix (does not remove security updates):
• Download and install NDI Tools (free).
• Launch NDI Access Manager.
• Open the Advanced tab.
• Under Receive Mode, change from RUDP to Single TCP or UDP (Legacy).
• Click OK.
• Restart any NDI-receiving apps (OBS, Studio Monitor, vMix) to force them to read the new config — NDI reads configuration only at application start. Repeat on every receiving machine in the flow. (docs.ndi.video, m.vmix.com)
• Alternative (only if reconfiguration is infeasible): Roll back the security update on the affected machine using Windows’ Uninstall updates path. Note: on combined SSU+LCU packages you may need DISM /Remove‑Package and reinstallation policies must be considered. Rolling back reintroduces the security vulnerabilities fixed in the August rollup and should be used only as a last resort with compensating controls. (support.microsoft.com)
• Test: Run a full rehearsal under live conditions after any change to validate latency, lip‑sync, and switching behavior. Single TCP usually restores stability but can add a few milliseconds of latency; UDP preserves latency on clean LANs but offers no retransmission. Pick the mode appropriate to your environment and test. (m.vmix.com)

---

Operational tradeoffs and risks​

• Security vs. Availability: Removing a security rollup that patched many CVEs (including serious Kerberos vulnerabilities patched in August) increases attack surface. For production teams, the tradeoff must be documented and approved by security governance. Microsoft’s advisories emphasize mitigation over rollback where possible. (tomshardware.com)
• Latency and quality: Single TCP is the safest universal fallback but can introduce head‑of‑line blocking during loss bursts; UDP (Legacy) keeps latency low but forfeits reliability. RUDP is normally the best balance — hence the urgency for Microsoft to restore RUDP behavior via a platform fix. (docs.ndi.video, m.vmix.com)
• Scale and human error: Coordinating Receive Mode changes across dozens or hundreds of endpoints, some of which may be remote or guest‑controlled, is operationally expensive. Change control and rehearsal windows are essential.
• Collateral reports (storage / recovery): Several outlets have reported separate regressions linked to the same August rollup (recovery tool failures, SSDs disappearing under heavy I/O). Those reports complicate the decision to reinstall or remove updates and require device‑specific validation from OEMs and Microsoft before taking destructive remediation steps. Treat storage reports as hardware-specific until confirmed by vendor testing. (tomshardware.com)

---

Recommended policy for studios, AV managers, and IT​

• Establish a production ring for AV/NDI hosts: Delay wide deployment of critical cumulative updates on show-critical machines until a short soak window proves stability in representative hardware and network topologies.
• Document network and NDI configuration: Record whether endpoints use RUDP, UDP, or Single TCP and why. That makes rapid remediation faster if a future update changes behavior again.
• Automate the mitigation where possible: Use management tools to deploy NDI Access Manager configuration (the config is stored in %programdata%/NDI/ndi-config.v1.json) and orchestrate restarts of NDI-consuming processes during maintenance windows. The NDI config is read at app start so automation must include application restarts. (docs.ndi.video)
• Keep patch rollbacks as a last resort: Document the risk decision and implement compensating controls (network segmentation, additional monitoring, temporary firewall rules) if a rollback is necessary for a critical live event.
• Monitor Release Health: Microsoft’s Release Health page is the authoritative status for known issues and will note when a fix or Known Issue Rollback is deployed. Subscribe to official update channels and vendor advisories. (support.microsoft.com, bleepingcomputer.com)

---

What to watch next​

• A Microsoft servicing update or Known Issue Rollback that restores RUDP stability without requiring NDI config changes.
• Vendor updates from NDI/NDI Tools that might add automatic fallbacks or tuning to tolerate the changed Windows behavior.
• OEM and storage vendor advisories if you rely on heavy sequential transfers: validate any SSD disappearance claims against firmware revisions and vendor test results before taking dramatic actions. Flag those claims as not yet uniformly reproducible and proceed cautiously. (tomshardware.com, docs.ndi.video)

---

Strengths and weaknesses of the public response​

Strengths:

• Microsoft acknowledged the problem publicly and posted a Release Health advisory, giving IT teams a clear signal and a working mitigation path.
• NDI and production-vendor documentation provides an immediate, reversible workaround that preserves the security update while restoring production continuity in most cases.
• Community testing reproduced the failure mode quickly, which accelerated vendor collaboration and guidance.

Weaknesses / risks:

• Lack of a published root‑cause analysis leaves vendors and integrators guessing whether future updates might reintroduce similar regressions.
• The workaround imposes nontrivial operational overhead on distributed production teams.
• Concurrent reports of unrelated regressions (storage, recovery tools) in the same cumulative window increase the complexity of remediation decisions and the risk of widespread disruption.

These strengths and weaknesses point to a pragmatic interim approach: accept the mitigation that preserves security (change NDI Receive Mode), stage future updates in a controlled ring, and insist that Microsoft publish a low-level analysis so vendors can harden their stacks against similar platform shifts. (m.vmix.com)

---

Practical checklist for immediate action​

• Confirm whether KB5063878 (Windows 11 24H2) or KB5063709 (Windows 10) is installed on production hosts.
• If affected and NDI is used, change NDI Access Manager → Advanced → Receive Mode to Single TCP or UDP (Legacy) and restart receiving applications. Verify end-to-end timing and latency in a rehearsal.
• If you cannot reconfigure remote or guest endpoints, evaluate a controlled rollback only under explicit, documented authorization and with compensating security controls in place.
• Deploy the update pilot‑ring policy to avoid show‑day surprises.
• Track Microsoft Release Health and vendor KBs for an official patch that restores RUDP functionality.

---

Conclusion​

The August 2025 Windows cumulative update (KB5063878) introduced a transport-layer interaction that destabilizes NDI’s default RUDP behavior, producing severe stuttering and choppy audio/video in multi‑PC streaming setups that use Display Capture and OBS or similar tools. Microsoft has publicly confirmed the issue and pointed to NDI vendor guidance: change the NDI Receive Mode to Single TCP or UDP (Legacy) while engineering investigates and develops a permanent fix. That mitigation preserves the security patch while restoring production stability in most environments, but it carries measurable tradeoffs in latency and operational complexity that production teams must manage. Administrators should stage the patch in production rings, automate configuration where possible, and prioritize rehearsals before live events to avoid interruptions. (bleepingcomputer.com, docs.ndi.video)
Caution: additional, separate reports about storage and recovery-tool regressions associated with the same August rollups have appeared; those claims require device‑specific verification with OEMs and Microsoft before taking corrective action. Treat storage anomalies as a distinct track while keeping the NDI workaround as the first-line operational response. (tomshardware.com)

---

Source: ITC.ua The latest Windows 11 update slowed down video streaming