Phison has confirmed it is investigating reports of SSDs disappearing or becoming permanently inaccessible after heavy file transfers on Windows 11 systems running Microsoft’s August cumulative update KB5063878. The controller maker’s statement, issued exclusively to Tom’s Hardware, acknowledges “industry‑wide effects” across multiple drive models and firmware combinations. Independent testing by X user Nekorusukii reproduced the failure on 12 out of 21 consumer SSDs, with one Western Digital SA510 2TB drive rendered unrecoverable even after a cold reboot. Separately, Microsoft has acknowledged a streaming regression in KB5063878 that causes severe stuttering in OBS and NDI‑based workflows—a bug serious enough that the NDI team and Microsoft are jointly recommending an immediate protocol workaround.

For users whose production machines or daily drivers are affected, the instinctive response is to uninstall the update. But that attempt frequently ends with installer error 0x800f0905, leaving many stuck between a malfunctioning system and a rollback that won’t complete. This guide explains exactly why the error appears, which steps Microsoft and the community have validated to remove the update, and how organizations can apply a safer, policy‑based mitigation while a permanent fix is developed.

What KB5063878 Changed

i
KB5063878 is the August 2025 cumulative update for Windows 11 version 24H2, build 26100.4946. It bundles a servicing stack update (SSU) and a latest cumulative update (LCU) into a single package—a combination that Microsoft explicitly warns cannot be uninstalled via wusa.exe /uninstall. The update delivers security fixes, quality improvements, and under‑the‑hood changes to the storage stack and networking subsystems.

Within days of its release, two distinct failure patterns emerged:

  1. Storage subsystem crashes during sustained writes. When a drive is more than roughly 60% full and a continuous transfer exceeds 50 GB, the SSD can disappear from Windows. In many cases a reboot restores visibility, but multiple reports describe corrupted file systems and at least one unrecoverable unit. The issue spans SATA and NVMe drives using controllers from Phison, Silicon Motion, and in‑house designs, suggesting a kernel‑mode change rather than a vendor‑specific firmware bug.
  2. NDI streaming regression. Microsoft’s Windows Release Health dashboard now lists a known issue where NDI‑based streaming (common in OBS Studio, vMix, and NDI Tools) experiences “severe stuttering and choppy audio/video” when the default RUDP (Reliable UDP) transport is used alongside Display Capture. The guidance: switch NDI Receive Mode to Single TCP or UDP until a code fix ships.

Phison’s public statement, sent to Tom’s Hardware on August 19, reads: “Phison has recently been made aware of the industry‑wide effects of the ‘KB5063878’ and ‘KB5062660’ updates on Windows 11 that potentially impacted several storage devices, including some supported by Phison. We understand the disruption this may have caused and promptly engaged industry stakeholders … the controllers that may have been affected are under review and we are working with partners.”

Why Uninstall Fails with 0x800f0905

Error 0x800f0905 is a generic servicing failure, but when it appears during removal of KB5063878 the root cause is almost always one of three conditions:

  • The combined SSU+LCU architecture resists wusa uninstalls. Microsoft’s KB page states, “Combined SSU and LCU packages cannot be uninstalled by running wusa.exe /uninstall. You must use DISM to remove the LCU component.”
  • Corrupted Windows Update caches (SoftwareDistribution or catroot2) prevent the servicing stack from locating and reversing the package.
  • The component store (WinSxS) is damaged, so even DISM cannot enumerate or remove the package until repairs are performed.

These conditions are well‑documented across Microsoft Learn, support forums, and independent validation. The following sequence resolves each in turn.

Step‑by‑Step: Safely Removing KB5063878

Back up critical data first. If your drive has already exhibited the disappearance behavior, do not initiate any large write operations on the affected volume until you have either removed the update or applied the enterprise workaround. Should you see signs of corruption (missing files, volumes failing to mount), stop all write activity immediately and collect SMART telemetry and Event Viewer logs for vendor support.

1. Run the Windows Update Troubleshooter

Navigate to Settings → System → Troubleshoot → Other troubleshooters and launch the Windows Update troubleshooter. It performs permission repairs, clears stuck service states, and resolves basic download cache issues. Reboot and test the uninstall again. This step alone clears a surprising number of 0x800f0905 cases.

2. Reset Update Components

If the troubleshooter doesn’t resolve the error, open an elevated Command Prompt and run:

net stop wuauserv
net stop bits
net stop cryptsvc
ren %SystemRoot%\SoftwareDistribution SoftwareDistribution.old
ren %SystemRoot%\System32\catroot2 catroot2.old
net start wuauserv
net start bits
net start cryptsvc

Restart the PC. These commands purge corrupted payloads and force Windows Update to rebuild its internal databases from scratch. Many 0x800f0905 failures originate from a poisoned catroot2 store, and this reset is the recommended first‑line fix.

3. Use DISM to Remove the LCU by Package Identity

The definitive removal method, documented by Microsoft, bypasses the Settings app entirely.

dism /online /get-packages | findstr KB5063878

Note the exact Package Identity string (it will resemble Package_for_RollupFix~31bf3856ad364e35~amd64~~26100.4946.1.1). Then:

dism /online /remove-package /packagename:<PackageIdentity>

Reboot when prompted. This targets only the cumulative update payload, leaving the SSU in place—which is expected and safe. If DISM itself fails with a servicing error, proceed to repair the component store.

4. Repair the Component Store and System Files

Run the following two commands consecutively:

DISM /Online /Cleanup-Image /RestoreHealth
sfc /scannow

DISM /RestoreHealth fetches fresh copies of corrupted system files from Windows Update (or a mounted ISO if offline). sfc /scannow then validates and replaces protected files. After both complete without errors, retry the dism /remove-package command.

5. Clean Boot and Driver Considerations

If the uninstall still fails, temporary disable all non‑Microsoft services and startup items via msconfig (General → Selective startup, then Services → Hide all Microsoft services → Disable all). Reboot and attempt removal in the clean environment. Also check Device Manager for any storage controller or network adapter drivers that may have been updated concurrently—rolling back or reinstalling OEM drivers before retrying the uninstall has resolved interference in several reported cases.

Blocking the Update After Removal

The servicing mechanism will re‑offer KB5063878 unless you actively hide it.

  • Consumer devices: Download and run Microsoft’s official Show or Hide Updates troubleshooter (wushowhide.diagcab). Select Hide updates, tick KB5063878, and confirm. Then pause Windows Update temporarily from Settings → Windows Update.
  • Managed environments (WSUS, Intune, SCCM): Decline the update or remove approval for the affected deployment rings. Use Group Policy to target the Known Issue Rollback (KIR) described in the next section if you cannot block the patch entirely.

Enterprise Mitigation: Known Issue Rollback (KIR)

For organizations that cannot afford to strip security fixes fleet‑wide, Microsoft offers a Known Issue Rollback. A KIR deploys a policy that disables the specific code change causing the regression—leaving the rest of the update intact. Microsoft has published a KIR MSI containing ADMX/ADML templates for KB5063878’s storage and networking regressions. Administrators import the templates into Group Policy Central Store, apply the policy to affected OUs, and force a policy refresh. The rollback takes effect on the next reboot and can be reversed by removing the policy when a permanent servicing fix is released.

This path preserves security coverage and is Microsoft’s recommended enterprise response. Pilot the KIR on a representative set of devices before broad rollout, particularly if your fleet mixes multiple SSD models and firmware revisions.

What to Watch After Rollback or Mitigation

  • Streaming: Validate OBS/NDI workflows. Even if you reverted the update, consider switching NDI Receive Mode to Single TCP as a long‑term stability improvement—Microsoft and NDI’s guidance suggests the protocol change is a safe default for many setups.
  • Storage: Gradually reintroduce heavy write workloads, monitoring disk latency and event logs. If you previously saw drive disappearances, start with transfers under 20 GB and verify that SMART attributes remain stable.
  • Microsoft channels: Monitor the KB5063878 support page and the Windows Release Health dashboard for confirmation of a resolved update package. Only install a corrected cumulative update once hardware vendors have validated it against your drive models.

Risks and Real‑World Trade‑offs

Uninstalling a cumulative update always removes the security patches it contained. For a gaming PC or a streaming rig, the operational risk of losing a drive mid‑session usually outweighs the theoretical exploit risk, especially if the machine is behind a firewall and runs only trusted software. On a domain‑joined laptop or server, however, the calculus flips: gaps in the security boundary are unacceptable. That’s why enterprise admins should prefer the KIR or policy‑based holding to wholesale LCU removal.

It’s also critical to understand that the SSU component of KB5063878 cannot be reverted. The SSU updates the servicing stack itself and remains even after you remove the LCU. This is by design and does not cause additional harm.

Finally, not every storage glitch reported in August is attributable to KB5063878. Independent testing by Nekorusukii and corroboration from Japanese outlet NichePCGamer demonstrates a reproducible trigger, but the failure depends on drive firmware, driver version, and the presence of other system‑level filters (encryption, deduplication). Treat the rollback as a focused intervention rather than a blanket diagnosis.

Looking Ahead: What Microsoft and Partners Must Deliver

The storage corruption bug—dubbed “disappearing drive” in community shorthand—has the hallmarks of a scheduler or flush‑ring change that misbehaves under boundary conditions. Phison’s investigation and the diversity of affected controllers suggest the fix must come from Microsoft’s kernel, not from SSD firmware alone. Given the severity, a targeted out‑of‑band update is plausible before the September Patch Tuesday. For streaming, Microsoft has already published a workaround and is expected to ship the corrected networking stack in the next cumulative update.

In the meantime, the troubleshooting sequence documented here—built from Microsoft’s own servicing guidance and verified by thousands of community reports—remains the reliable path for anyone stuck with error 0x800f0905. Back up, follow the steps in order, and block the re‑offer until the all‑clear is given.

This article incorporates technical steps validated by Microsoft documentation and independent testing from Tom’s Hardware, Nekorusukii, and community troubleshooting threads. Always test mitigations in a non‑production environment before deploying to critical systems.