Monday, August 18th 2025

Microsoft Windows 11 24H2 Update May Cause SSD Failures

Microsoft issued an OS security update as part of the August 2025 Patch Tuesday cycle to counter the Lumma stealer and similar malware and exploits. The Windows 11 24H2 cumulative package KB5063878 was included in that rollout. Early deployments returned an installation error, 0x80240069, which Microsoft acknowledged and later addressed. Community testing has since identified a reproducible pattern in which certain SSDs, when running sustained sequential writes of approximately 50 GB, become inaccessible and display unreadable SMART telemetry. Follow-up analysis and vendor feedback to date primarily point to solid-state drives, although conclusions about traditional spinning hard disk drives remain unconfirmed.

A Japanese tester NichePCGamer published a 21-drive matrix and a repeatable procedure: copy a large game folder, write an approximately 62 GB archive composed of many video files to the target SSD, then decompress that archive on the same device. Failures were reported most frequently when controller utilization exceeded 60% during continuous writes. Reported devices include several Phison-based NVMe models and other drives using different controllers. Symptoms reportedly recur after a reboot, which often restores visibility temporarily but does not eliminate the underlying fault, creating a risk of file corruption if the same workload is repeated. Microsoft and SSD vendors are investigating and have not yet published a definitive root-cause report. Until vendors and Microsoft publish formal guidance, users with SSDs should back up important data, avoid long sequential transfers on systems updated to KB5063878, and monitor vendor and Microsoft advisories for SSD-specific firmware, driver, or OS fixes.

Update 14:02 UTC: Phison has posted an update, acknowledging the issue.

Update Aug 21st: Phison has reached out to us to let us know that some documents about this issue are fake.
Sources: via Neowin, Notebookcheck
Add your own comment

278 Comments on Microsoft Windows 11 24H2 Update May Cause SSD Failures

#176
remixedcat
R-T-BThe question I pose though is if disabling defender can act as a temporary protection against... whatever? At least until the firmware or update is modified in such a way that this ceases.

Note, that has a lot of its own drawbacks that should be obvious, but its something I have to ask.
I'm not gonna science that one, sadly. I allready have some SSD sciencing of my own I am doing soon,
Posted on Reply
#177
R-T-B
remixedcatI'm not gonna science that one, sadly. I allready have some SSD sciencing of my own I am doing soon,
Yeah it's too high risk to try it out on the employees here too, but at least I'm a datapoint (no failures or even slowness). Curious.
Posted on Reply
#178
unwind-protect
R-T-BNot really. NTFS is a patent nightmare. It is derived from HPFS which was coded by IBM and they likely would not consent to documentation either. So many other parties were involved at some point too... I think it's safe to say its probably easier to reverse engineer it than navigate the legal hurdles in the modern NTFS's case.
Reimplementing without documentation would not protect against patent issues tho
Posted on Reply
#179
R-T-B
unwind-protectReimplementing without documentation would not protect against patent issues tho
It historically has in a round about way. See wikipedia for information about the legalese around a "clean room" reimplementation. Its how projects like ReactOS and similar survive.

Also is the legal protective foundation to the Unity Analytics Killer Project on github, founded by yours truly. No lawsuits to date.
Posted on Reply
#180
EsaT
efikkanThe drives have a defect, they were defective products from the moment they were made, you just didn't know it yet.
You lost the rights to demand hardware manufacturers to make flawless product when you're defending "F*ck the users, quality can go to hell." MS&Nadella firing their testing department.
Do I need to remind you that they've pushed out seriously bugged updates even when knowing those issues?

You either put equal quilt on MS for not doing any testing, or then give also hardware manufacturers that Get Out of Jail Free card.
Posted on Reply
#181
unwind-protect
R-T-BIt historically has in a round about way. See wikipedia for information about the legalese around a "clean room" reimplementation. Its how projects like ReactOS and similar survive.

Also is the legal protective foundation to the Unity Analytics Killer Project on github, founded by yours truly. No lawsuits to date.
Clean room implementation are about copyright, not patents.
Posted on Reply
#182
R-T-B
unwind-protectClean room implementation are about copyright, not patents.
I may be conflating the two in my hurry to reply in between meetings at my job... lol. Good catch.

Copyright would be the chief concern here I'd think. There probably aren't many bona fide patents that still are valid on HPFS / NTFS. Both of those file systems are quite old now, dating back the early 90s if not the 80s.

Obligatory "I am not a lawyer" moment. lol.
Posted on Reply
#183
RJARRRPCGP
I thought that was to not run afoul of patent law.
Posted on Reply
#184
unwind-protect
RJARRRPCGPI thought that was to not run afoul of patent law.
If there is indeed a patent on a software mechanism you would violate it even with a clean-room implementation done via reverse engineering.

But as pointed out, patents on NTFS, if there have been any, are likely all expired. MS should just get over themselves and document the on-disk format or, even better, contribute an NTFS implementation to the Linux kernel.
Posted on Reply
#185
RJARRRPCGP
I think there's been fear of running afoul of patent law in the UEFI world! Because UEFI regularly is using a file system from the late-1990s! It was first introduced with Windows 95 OSR2. Say hello to FAT32!

Even when there's improvements over FAT16, it's still more prone to corruption, thus more limited case uses now. It was from the days where we weren't worried about a non-journalled FS.
Of course, it's less forgiving to power outages and severe brown outs.
Posted on Reply
#186
Doctor Who
I'll say this once and I'll say it again, Windows 11 is one of the worst unstable OSs M$ has come out with to date.
Hence why I am back to Windows 10 x64 Pro and ain't looking back.
Posted on Reply
#187
Sora
AleXXX666endoflife.date/windows

Enterprise, right. I ended up in using Pro editions of both 10 and 11 on all machines I maintain.:) Happy life without constant Minesweeper gameplay (will it break? lol),;):D

not enterprise, when MSFT says a product is end of support it means they aren't producing new features, so nothing turned on in 23h2 gets turned on in 22h2, but web experience and widget updates (task manager), security, defender, etc all still apply to 22h2. If 22h2 has a driver issue, its also on 23h2 and both get fixed in the update it gets fixed in.
efikkanMultiple Samsung SSD models had some firmware issues around a decade ago which lead to wear and corruption, which only happened in Linux, and it was covered but no outcry. And back then Samsung and Intel were known as the "good" consumer SSDs for Linux, as the vast majority of them were misbehaving, especially with the trim command. I don't remember anyone being foolish enough to call for lawsuits against the Linux community for defective hardware…
Your timeline doesn't line up with reality, SSD's produced close to recently have had abnormal wear and corruption failures.
remixedcatwhile this recent windows 11 thing affects all ssds regardless of controller/firmware/etc
No, no it doesn't.
Event HorizonWhy simp for Microsoft? Just stop. People are losing data. What is wrong with some people?
People lose data all the time because of their own foolishness.
R-T-BAnyways, a really big question I haven't seen answered is if disabling defender makes you immune to issues... because I do disable it on a few of my workstations, so that is certainly a datapoint, considering this update was designed to "address a new ransomware variant."
No, because there are no failures reported on VM's, which is telling really.
The nvme emulation in virtual machines aren't exactly robust and corruption is not demonstrated on those, only on bare metal,

There isn't any machine configuration data supplied to connect up to potential PCIE controller errors either.
This could as well be a mainboard chipset errata that only occurs with certain PCIE or SAS controller registers set.
Super XPI'll say this once and I'll say it again, Windows 11 is one of the worst unstable OSs M$ has come out with to date.
No, that award goes to Windows 10 1507.
11 is generally more stable than 10 unless you're one of those morons that carve everything out of the OS because you believe it gets you 1% more framerates (it doesn't)
Posted on Reply
#188
efikkan
RJARRRPCGPI think there's been fear of running afoul of patent law in the UEFI world! Because UEFI regularly is using a file system from the late-1990s!
It may be because it only requires a few lines of assembly to read FAT.
SoraYour timeline doesn't line up with reality, SSD's produced close to recently have had abnormal wear and corruption failures.
I was specifically replying about Linux related bugs, that is bugs triggered by running Linux. Samsung has had several "mishaps" since, including 980 Pro and 990 Pro which had firmware bugs causing abnormal wear and corruption (980 Pro had two or more instances if I recall correctly), but I don't know that these were limited to Linux users. And then there is Samsung 870 EVO (SATA) which has a horrible reputation for long-term reliability (on all systems)… As I've said in previous discussions; Samsung is known for making either very good or very bad SSDs. I'm still rocking three 970 Pros, which some regard as the last great consumer SSD.
And the frequency of issues with minor brands is probably much higher than we know, as these are mostly white label products which changes all the time and probably doesn't reach the critical mass.

-----

I find the response from Phison in the new article interesting. It would be interesting to see wheter the problem is exaggerated or if the assumptions about what triggers it is faulty.
Posted on Reply
#189
RJARRRPCGP
SoraNo, that award goes to Windows 10 1507.
11 is generally more stable
QFT, maybe Windows 10 1511 as well, as I noticed an unusual storahci bug back in 2016.

11 has generally been less likely to have update errors.
efikkanI'm still rocking three 970 Pros, which some regard as the last great consumer SSD.
Hearts go out to 970 Pro users! I have a 512 GB 970 Pro in one of my builds.
Posted on Reply
#190
AleXXX666
Soranot enterprise, when MSFT says a product is end of support it means they aren't producing new features, so nothing turned on in 23h2 gets turned on in 22h2, but web experience and widget updates (task manager), security, defender, etc all still apply to 22h2. If 22h2 has a driver issue, its also on 23h2 and both get fixed in the update it gets fixed in.



Your timeline doesn't line up with reality, SSD's produced close to recently have had abnormal wear and corruption failures.



No, no it doesn't.



People lose data all the time because of their own foolishness.



No, because there are no failures reported on VM's, which is telling really.
The nvme emulation in virtual machines aren't exactly robust and corruption is not demonstrated on those, only on bare metal,

There isn't any machine configuration data supplied to connect up to potential PCIE controller errors either.
This could as well be a mainboard chipset errata that only occurs with certain PCIE or SAS controller registers set.



No, that award goes to Windows 10 1507.
11 is generally more stable than 10 unless you're one of those morons that carve everything out of the OS because you believe it gets you 1% more framerates (it doesn't)
don't remember a THING 1507 was unstable lol. even recently i've installed 1607 to some office pc so not to mess with "buggy" updates, but even THIS version got BSOD-ed once with some stupid recent update. But 1507 is out of supoport with many modern sw. But I loved it's really old UI lol starting with 1607 they messed it stupidly up... (UI). Now, 11 seems a lil bit nicer, although it's a pure copy of linux ui...
Posted on Reply
#191
Sora
AleXXX666don't remember a THING 1507 was unstable lol.
i was tasked with the deployment testing of 1507 for a company, and things widely talked about at the time where the user profile would corrupt, dism(update deployment), app store corruption, among other things were easily triggered.

Windows 10 did not become widespread use ready until 1703.
Posted on Reply
#192
RJARRRPCGP
Soradism(update deployment)
Well, you could have very well saw such issues, since the July, 2025 patch-Tuesday. CBS reported the component store having unexpected content, and DISM with the restorehealth option also fails! :wtf:

23H2 usually just had an issue where Component Based Servicing (CBS) signals Bluetooth files being corrupted after a patch-Tuesday update.

The 24H2 issues are worse, multiple times now, DISM with restorehealth option failed on 24H2 with error message that the source can't be found, which was not a problem on 23H2.

However, it looks like the restorehealth option possibly has been fixed with the August, 2025 patch-Tuesday.

The July, 2025 patch-Tuesday has been the worst for me.
Posted on Reply
#193
remixedcat
efikkanI'm still rocking three 970 Pros, which some regard as the last great consumer SSD
Not really, not when you got sn7100,sn850x,firecuda,nytro,solidigm/sk hynix platinum p41/44,and GOOD OL OPTANE!!!
Posted on Reply
#194
Sora
RJARRRPCGP23H2 usually just had an issue where Component Based Servicing (CBS) signals Bluetooth files being corrupted after a patch-Tuesday update.
This is user fixable, the issue is that a hardlink cannot replace an already placed file - and the original 22h2 media PLACED the file into system32\drivers instead of hardlinking it there, run a trusted installer elevated command prompt and delete then hardlink the file to the location from the WinSxS location and it'll update correctly from then on.
RJARRRPCGPThe 24H2 issues are worse, multiple times now, DISM with restorehealth option failed on 24H2 with error message that the source can't be found, which was not a problem on 23H2.
www.elevenforum.com/t/24h2-how-to-remove-the-two-un-reclaimable-that-always-show-in-dism.34680/

This should fix that.
Posted on Reply
#196
Parn
RJARRRPCGPQFT, maybe Windows 10 1511 as well, as I noticed an unusual storahci bug back in 2016.

11 has generally been less likely to have update errors.


Hearts go out to 970 Pro users! I have a 512 GB 970 Pro in one of my builds.
1TB 970Pro here. I wish sammy would put real MLC (2 bits per cell) in their Pro lineups than misleading their customers with fake MLC (TLC in disguise) in 990Pro. Only found out that after I had purchased mine.
Posted on Reply
#197
RJARRRPCGP
Parn1TB 970Pro here.
I got my 512 GB 970 Pro during the "video card catastrophe", back in '21.

Fun fact: The 970 Pro is 512 GB, instead of 500 GB.
Posted on Reply
#199
RJARRRPCGP
A Computer GuyHere we go again...
This looks like possibly a Windows PCI-E subsystem fault! :wtf:

It looks extremely sus! It may be due to maximum utilization of PCI-E.
Posted on Reply
#200
Sora
RJARRRPCGPIt looks extremely sus! It may be due to maximum utilization of PCI-E.
ryzen pcie error rates rising from the dead :D
Posted on Reply
Add your own comment
Nov 19th, 2026 14:49 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts