Phison has confirmed it is investigating a storage regression tied to Microsoft’s August cumulative update for Windows 11 24H2 that makes NVMe SSDs disappear during heavy sustained write operations. The issue, which can corrupt files in a minority of cases, has been reproduced by independent testers across multiple drive models and controllers, raising fresh concerns about the fragility of the Windows storage stack.

The KB5063878 update, released August 12, 2025 (OS Build 26100.4946), was intended as a routine security and quality patch. Within days, however, enthusiast testers and tech outlets began documenting a consistent failure profile: when performing large sequential writes—commonly exceeding 50 GB—some NVMe drives vanish from File Explorer, Device Manager, and Disk Management. In some instances, files written during the failure window are corrupted or missing even after a reboot. The issue is not limited to a single vendor; drives using Phison, Silicon Motion, and other controllers have been implicated, along with a handful of SATA SSDs and even hard drives in isolated tests.

Phison’s statement, shared with Tom’s Hardware, acknowledges “industry-wide effects” and says the company is working with partners to review affected controllers. The language stops short of attributing blame to Windows or specific firmware, a common pattern when cross-ecosystem bugs require joint telemetry. This report synthesizes community findings, vendor responses, and technical analysis to guide affected users and IT professionals.

The August Cumulative Update Breaks Storage

Microsoft’s KB5063878 arrived as part of the regular Patch Tuesday cycle with no mention of storage issues. By August 15, popular hardware forums and X (formerly Twitter) lit up with reports of disappearing drives. One prominent reproduction came from tester Nekorusukii @Necoru_cat, who documented the failure while updating Cyberpunk 2077 on a system with the update installed. The game’s SSD simply vanished. After rebooting, the drive reappeared, but repeating the large file transfer triggered the same disappearance.

The pattern quickly gained traction. Japanese outlet NichePCGamer collated at least eight additional user reports. A SanDisk Extreme Pro M.2 NVMe user recounted that a 50 GB Honkai: Star Rail update made the drive inaccessible multiple times; uninstalling KB5062660 (a precursor update) eliminated the problem. Community testers then designed a reproducible test: copy a Steam library folder, create a compressed archive, and copy the expanded archive to the target drive. Under this sustained write load, drives consistently dropped offline.

Timeline: From Release to Widespread Reports

  • August 12, 2025 – KB5063878 released for Windows 11 24H2.
  • Within days – Enthusiast testers and outlets publish reproducible test cases; drives vanish during large continuous writes.
  • Mid-August – Phison issues a statement acknowledging investigation and partner engagement. Independent outlets urge caution.
  • Concurrent enterprise issues – Administrators using WSUS/SCCM report unrelated installation errors (0x80240069), prompting Microsoft to issue Known Issue Rollback (KIR) mitigations for managed environments.

Symptoms, Trigger Profile, and Affected Drives

The symptoms are unmistakable:
- NVMe SSD disappears from File Explorer, Device Manager, and Disk Management during a write operation.
- Event Viewer logs show NVMe or storage controller errors and kernel I/O failures.
- Controller and SMART telemetry become unreadable to vendor tools.
- In a subset of cases, files written during the incident are corrupted, truncated, or missing after a reboot.

The trigger is sustained sequential writes—large file copies, archive extractions, game installs/updates, or cloning tasks. Community reproductions commonly cite a threshold around 50 GB, though the exact point varies by model, firmware revision, and host platform. Nekorusukii’s testing of 21 drives found 12 became inaccessible; only a Western Digital SA510 2TB could not be recovered after a reboot. Drives from Corsair, Kioxia, Samsung, and others also exhibited the behavior, indicating a broad OS‐host interaction rather than a single-vendor bug.

Phison-based controllers (PS5012-E12, E21T, E31T families) appear over-represented, but non-Phison controllers and even a few hard drives have been implicated in isolated repros. DRAM-less or HMB-reliant designs are common among affected units, hinting at a host memory buffer allocation change as a probable co-factor.

Phison’s Response: Investigating but Not Pointing Fingers

Phison’s statement to Tom’s Hardware reads: “Phison has recently been made aware of the industry-wide effects of the 'KB5063878' and ‘KB5062660’ updates … We understand the disruption this may have caused and promptly engaged industry stakeholders. … At this time, the controllers that may have been affected are under review and we are working with partners. We will continue to provide updates and advisories.”

The measured language is typical for controller vendors: confirm awareness, promise partner coordination, and hold off on public root-cause attribution until joint telemetry confirms the exact failure mechanism. Western Digital and SanDisk had previously issued firmware fixes for HMB-allocation BSODs during the earlier 24H2 rollout, demonstrating that vendor-side firmware updates can resolve such regressions.

Microsoft’s KB page for KB5063878 did not initially list a storage-device known issue. The company has tools like Known Issue Rollback and targeted update blocks to manage rollout, but public communication often lags behind community reports. With enterprise deployment errors already triggering KIR mitigations, administrators are advised to treat this update cautiously.

Technical Root Causes: HMB, Firmware, and OS Interaction

Three plausible mechanisms recur in community analysis:

  1. Host Memory Buffer (HMB) allocation changes – Many budget NVMe drives lack onboard DRAM and use HMB to borrow system RAM for caching flash translation layer structures. If the OS increases allocation size or alters timing, DRAM-less controllers can enter unstable states under heavy write stress. Earlier 24H2 bugs linked HMB changes to BSODs on WD/SanDisk models; the current failures may represent a subtler deadlock rather than a crash.

  2. Buffered I/O or NVMe command ordering regression – A kernel-level change in buffering, command ordering, or timing could create command sequences that certain firmware cannot handle. When the firmware fails to respond to admin commands, the device effectively disappears from the PCIe/OS topology—matching the observed symptoms.

  3. Controller firmware edge cases under high utilization – SSD controllers juggle garbage collection, translation tables, and wear-leveling. Prolonged sequential writes might trigger an internal metadata update or recovery that deadlocks or corrupts controller state, making the drive unreachable until a power cycle or firmware reset. The fact that SMART data becomes unreadable after failure points toward a firmware lockup scenario.

Important caveat: definitive root cause requires joint telemetry from affected SSD vendors and Microsoft. Community reproductions provide strong hypotheses but not vendor-confirmed diagnoses.

Data Loss Risk: Real and Documented

This is not just an inconvenience. Some testers reported files written during the failure window were corrupted or missing. For content creators, system imaging, and large game updates, this represents a tangible data-loss threat. The severity is heightened because the drive can become inaccessible even after a reboot—as demonstrated by the Western Digital SA510 that remained unrecoverable in the test suite. However, occurrence is not universal; variables like firmware revision, UEFI version, and host driver all influence whether a given drive will fail under identical conditions.

For enterprise fleets, the parallel deployment error (0x80240069) compounds the risk: administrators must inventory drives, withhold the update on suspect hardware, and rely on Microsoft’s Known Issue Rollback where applicable.

Mitigations for Users and IT Admins

Immediate steps for end users:
- Back up critical data to a physically separate drive or cloud immediately.
- Avoid heavy sustained writes (game installs, large clones, archive extraction) on systems with KB5063878 installed until your drive vendor publishes a fix.
- Check your SSD vendor utility (Corsair iCUE, Western Digital Dashboard, SanDisk Toolkit, Kioxia Toolbox, etc.) for firmware updates and apply any vendor-recommended updates after backing up.

Advanced temporary workaround: disable HMB via Registry – A stopgap that reduces performance but may prevent HMB-related failures. Use with caution:
1. Open Registry Editor (regedit).
2. Navigate to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\stornvme\Parameters\Device
3. Create or modify the DWORD HMBAllocationPolicy and set it to 0.
4. Reboot.

For IT administrators:
- Inventory storage devices and flag systems with models identified in community testing.
- Use WSUS, SCCM, or Intune to withhold KB5063878 from at-risk fleets until validated.
- If a drive fails, power off the system to preserve state, image the drive, gather logs, and open a support case with the SSD vendor.

Why This Keeps Happening

Modern operating systems and SSD controllers are tightly coupled. Features like HMB improve performance but create dependencies on host memory allocation and driver timing. When a platform update alters these variables—even in a supposedly compatible manner—previously untested firmware edge cases can emerge. The sheer diversity of controller firmware revisions, OEM BIOS versions, and system configurations makes exhaustive pre-release testing impossible. The result is that low-probability, high-impact regressions like this still surface in the field, requiring coordinated remediation across multiple vendors.

Outlook: Coordinated Remediation Underway

Phison’s public acknowledgment and engagement with partners mark the first step toward a fix. History suggests that once root cause is confirmed, SSD vendors will release firmware updates, and Microsoft may issue an out-of-band patch or broaden Known Issue Rollback protections. Western Digital’s prior HMB-related firmware fixes demonstrate that such coordinated responses can succeed. However, the immediate burden falls on users and IT teams to practice cautious update management and thorough backups.

The incident underscores a broader lesson: storage is a co-engineered system where OS updates, driver behavior, controller firmware, and platform firmware all interact. Transparent vendor advisories, prompt firmware remedies, and robust telemetry sharing are essential to restoring confidence—and those processes are now in motion.