Your NVMe SSD was ripping through game installs and level loads a week ago. Now, after a Windows update (or a driver, firmware, or BIOS update), writes have collapsed to a crawl — file copies that took seconds now sit at “Calculating” or trickle along at double-digit megabytes per second, Shadowplay recordings stutter, and Steam downloads stall at the disk stage. The drive isn’t dead; reads might even feel normal. It’s specifically writes that tanked, often dropping to a small fraction of the drive’s rated speed, sometimes worse than an old SATA drive.
Don’t panic and don’t reformat yet. In the vast majority of cases this is a software-side regression — a Windows power-setting change, a driver replacement, a write-cache toggle flipping off, or HMB (Host Memory Buffer) allocation failing after an update — not a failing drive. Work through the fixes below in order; most people resolve it inside the first three. Budget 30–60 minutes. Before you start, back up anything irreplaceable, and download the latest storage and chipset drivers for your motherboard from the manufacturer’s site so you’re not hunting mid-fix.
Likely Causes
| Cause | How to confirm | Fix difficulty |
|---|---|---|
| Write caching disabled after update | Device Manager → drive → Policies tab; “Enable write caching” unchecked | Easy |
| Power plan throttling the drive (HIPM/DIPM, LPM) | Slow speeds appear mainly when idle or after bursts; power plan reset to Balanced defaults | Easy |
| Windows Search/Indexer or SysMain hammering the disk post-update | Task Manager → Performance tab shows 100% Active Time with low actual throughput; check Resource Monitor’s Disk tab | Easy |
| StorAHCI/NVMe driver replaced by a generic one (or vice versa) | Device Manager → Storage controllers shows “Standard NVM Express Controller” where your vendor driver used to be | Moderate |
| HMB (Host Memory Buffer) not allocated after update | DRAM-less drive (common on budget 1–2 TB models); vendor toolbox (e.g., Samsung Magician, WD Dashboard) shows HMB unavailable | Moderate |
| SSD firmware bug triggered by the update, or SLC cache misbehaving | Vendor toolbox flags a firmware update; speeds recover after a full idle/rest period or power cycle | Moderate |
| Drive overheating and throttling (heatsink removed, case airflow changed) | CrystalDiskInfo or HWiNFO shows controller temps above 70 °C under load | Easy |
Fix 1: Re-enable Windows Write Caching
What causes it
Feature updates occasionally reset device policies. With write caching off, every write is flushed to NAND immediately, and small random writes — exactly what games and Windows generate — suffer dramatically.
How to check
Right-click Start → Device Manager → expand Disk drives → right-click your NVMe drive → Properties → Policies tab. If “Enable write caching on the device” is unchecked, this is likely your problem.
What to do
Check “Enable write caching on the device.” If the option is grayed out, close Device Manager, reopen it as administrator, or check whether the drive is running in a RAID/VMD mode that locks the setting. Click OK and reboot.
How to undo it
Uncheck the same box. Leave it off only if you’re running a system without battery backup and need maximum data safety on sudden power loss — otherwise, keep it on.
Fix 2: Change the Power Plan and Disable Aggressive Link Power Management
What causes it
Windows updates can reset the active power plan or its hidden storage settings. Aggressive Link Power Management (HIPM/DIPM) puts the NVMe controller into low-power states between writes; on some drive/controller combinations, waking from these states introduces long latencies that crater sustained write throughput.
How to check
Control Panel → Power Options. Note which plan is selected and whether it changed after the update. Also check your vendor toolbox or HWiNFO for NVMe power state transitions if available. If slowness is worst right after idle periods and recovers during heavy activity, power management is a prime suspect.
What to do
Select the High Performance plan (or Ultimate Performance: run powercfg -duplicatescheme e9a42b02-d5df-448d-aa00-03f14749eb61 in an admin Command Prompt). Then, in your plan’s advanced settings, expand PCI Express → Link State Power Management and set it to Off. On laptops, also test with the power adapter plugged in, since battery mode enforces deeper power savings.
How to undo it
Switch back to Balanced and set Link State Power Management to Moderate Power Savings. You lose a little idle battery life with LPM off, which is the only real trade-off.
Fix 3: Stop Post-Update Background Disk Activity
What causes it
For hours (sometimes a day or two) after a cumulative update, Windows Search reindexes your files, SysMain (Superfetch) rebuilds its cache, .NET optimization jobs run, and OneDrive re-syncs. These background writes saturate the queue and make your drive look broken when it’s just busy.
How to check
Ctrl+Shift+Esc → Task Manager → Performance tab → select your disk and watch Active Time. Then open Resource Monitor (Run → resmon) → Disk tab and sort by Total B/sec to see exactly which processes are writing. Look for SearchIndexer.exe, MsMpEng.exe (Defender scanning post-update), or System.
What to do
First, just let it finish — leave the machine plugged in and idle overnight; this alone resolves many “my SSD died after updating” reports. If SearchIndexer is stuck for days: Settings → Search → Searching Windows → “Exclude folders” the large game libraries, or rebuild the index via Control Panel → Indexing Options → Advanced → Rebuild.
How to undo it
Re-enable anything you disabled: remove folder exclusions, and re-enable SysMain if you disabled it (Services.msc → SysMain → Startup type: Automatic).
Fix 4: Roll Back or Replace the NVMe/Storport Driver
What causes it
Windows Update sometimes swaps your vendor NVMe driver (Samsung NVMe Driver, etc.) for Microsoft’s generic “Standard NVM Express Controller,” or installs a problematic new storport.sys. Either direction can cause write-latency regressions on specific drive models.
How to check
Device Manager → Storage controllers. If you see “Standard NVM Express Controller” but you previously installed a vendor driver, the swap is your culprit. Note the driver version under Properties → Driver tab. The failure mode is typically a sudden large drop in write speeds specifically, with reads less affected — a signature of HMB allocation failing under the generic driver on DRAM-less drives.
What to do
If a vendor driver was displaced: download the current NVMe driver from your SSD manufacturer’s support page, uninstall the generic controller entry from Device Manager (check “delete driver” only if prompted during vendor install), install the vendor package, and reboot. If a vendor driver was already present and broke after updating it: Properties → Driver tab → Roll Back Driver (available for 10 days after an update). If Roll Back is grayed out, manually install an older driver version via Update Driver → Browse my computer → Let me pick.
How to undo it
Reverse the swap: Device Manager → Update Driver → Browse → Let me pick from a list, and select the alternative driver. Reboot after each change and retest so you know which driver behaves.
Fix 5: Update or Roll Back SSD Firmware
What causes it
Some firmware versions interact badly with new Windows builds or driver stacks — including bugs in HMB negotiation or garbage-collection scheduling that throttle writes. Manufacturers ship corrective firmware quickly when a regression is widespread.
How to check
Open your vendor toolbox: Samsung Magician, WD Dashboard, Crucial Storage Executive, Kingston SSD Manager, or the Corsair/ADATA/Seagate equivalent. It reports current firmware and flags available updates. CrystalDiskInfo also displays the firmware string if you don’t know it. Compare against the latest version listed on the manufacturer’s support page for your exact model.
What to do
Back up first. Close all applications, plug in a laptop, and run the firmware update from the vendor toolbox. Do not power off during the flash. Afterward, shut down fully (not restart — a full shutdown power-cycles the drive’s controller), wait 30 seconds, and boot.
How to undo it
Generally you can’t downgrade firmware — vendors rarely offer it. If a new firmware causes problems, check the vendor’s support page or forums for an archived older version, or wait for the next release.
Fix 6: Check Drive Temperature and Throttling
What causes it
An update can increase background load, pushing an already-marginal thermal situation over the throttling threshold. NVMe controllers (especially Phison-based DRAM-less models) throttle hard above roughly 70–80 °C, and sustained writes are the first casualty. If you removed your motherboard’s M.2 heatsink, added a hot GPU directly adjacent, or improved case airflow elsewhere, the drive may now be cooking.
How to check
Install HWiNFO64 (Sensors-only mode) or check CrystalDiskInfo. Watch the drive’s Composite/Controller temperature during a large copy. Sustained readings above 70 °C, or sudden speed drops that coincide with temperature spikes, confirm throttling. Touch the M.2 heatsink (carefully) — painfully hot to hold means airflow is inadequate.
What to do
Refit the motherboard’s M.2 heatsink if it was removed (many boards refuse to boot certain slots without it, and it works better than thin stick-on pads). Add or reposition a case fan to move air across the M.2 slot. If your motherboard has multiple M.2 slots, move the drive to a slot that isn’t sandwiched between the GPU and CPU. Thermal pads between the drive and heatsink should be the thickness the motherboard manufacturer specifies — usually 1–3 mm.
How to undo it
Restore your original airflow or slot configuration if the change hurts anything else. Thermal fixes have no software side effects.
Fix 7: Force a TRIM Pass and Give the Drive Rest
What causes it
Post-update churn can leave the drive’s internal mapping in a state where sustained writes trigger heavy garbage collection. A manual TRIM plus an idle period lets the controller consolidate and restore its write cache behavior.
How to check
Open an admin Command Prompt and run fsutil behavior query DisableDeleteNotify. If it returns DisableDeleteNotify = 1, TRIM is disabled — a serious problem in itself.
What to do
If TRIM is off, run fsutil behavior set DisableDeleteNotify 0. Then run defrag C: /L (this issues a retrim on SSDs — it does not defragment them) via an admin Command Prompt, and let the machine idle for a few hours or overnight. Full shutdown, wait 30 seconds, and cold boot afterward.
How to undo it
Nothing to undo — retrimming is safe and can be repeated monthly. Do not run traditional defragmentation repeatedly on the SSD beyond Windows’ own scheduled optimization.
Fix 8: Uninstall the Update or Use System Restore
What causes it
If the regression appeared immediately after a specific Windows update, uninstalling it confirms the diagnosis and buys time until a patched build ships.
How to check
Settings → Windows Update → Update history. Match the install date to when the slowdown started.
What to do
Settings → Windows Update → Update history → Uninstall updates. Remove the most recent cumulative or optional driver update. Alternatively, search “Create a restore point” → System Restore → pick a restore point dated before the slowdown. Afterward, pause updates for a week (Settings → Windows Update → pause) so it doesn’t reinstall immediately.
How to undo it
Simply reinstall the update from Windows Update when a fixed version is available — or accept it once you’ve verified a later driver/firmware update resolved the underlying issue.
If Nothing Works
Test the drive in another system, or another NVMe drive in your system — this cleanly separates drive problems from Windows problems. If the drive is slow elsewhere too, contact the manufacturer for warranty service; write-speed collapse can indicate genuine NAND or controller degradation, and vendor tools with a health percentage (CrystalDiskInfo’s health status, or the vendor toolbox) will often confirm it. If the drive is fine elsewhere, back up and consider an in-place repair install of Windows (run setup.exe from a mounted Windows 11 ISO and choose “Keep personal files and apps”), which replaces the entire Windows driver and system file stack without touching your data. Finally, rule out the motherboard side: check your BIOS/UEFI vendor’s site for a storage-related BIOS update, and confirm the drive isn’t sharing bandwidth or running in a degraded M.2 slot mode after a BIOS reset.
FAQ
Why are only write speeds slow while reads seem fine?
Writes depend on the drive’s cache, HMB allocation, power-state wake latency, and write caching — all software-controlled and all frequently disturbed by updates. Reads use simpler, more deterministic paths, so they’re rarely hit as hard.
Can a Windows update physically damage my SSD?
Not directly, but a bad update can expose a firmware bug, cause extreme thermal load, or generate heavy write amplification. If health metrics in CrystalDiskInfo or your vendor toolbox show declining health or growing reallocated sectors, the drive itself is the problem — claim warranty.
Should I use the manufacturer’s NVMe driver or Microsoft’s Standard NVM Express Controller?
Whichever performs correctly on your drive. DRAM-less drives in particular often need the vendor driver for proper HMB support. If the generic driver was fine before the update and broke after, try the vendor driver — and vice versa.
How long should post-update background activity last before I worry?
Indexing and cache rebuilding typically finish within a few hours to a day on a healthy system. If Resource Monitor still shows heavy background writes from system processes after two or three days of normal use, move on to the driver and firmware fixes.
