Interactive Troubleshooter

Published: Nov 4, 2026

NVMe SSD Not Showing Full Speed In CrystalDiskMark: PCIe Lane Diagnostic

Time: 15 mins Level: Beginner

By StepDiag Engineering Team

Reviewed by Ibraheem Taofeeq Opeyemi

An exhaustive root-cause diagnostic tree for troubleshooting low NVMe SSD read/write speeds in CrystalDiskMark. This guide systematically pinpoints physical PCIe lane bifurcation limits, M.2 slot bandwidth sharing, thermal throttling states, NVMe driver discrepancies, and Windows Power Plan configurations.

Full Diagnostic Tree & Step-by-Step Overview

What is the primary symptom or benchmark discrepancy observed in CrystalDiskMark?

Sequential Read/Write speeds are exactly 50% or 25% of rated spec (e.g., ~3500 MB/s instead of ~7000 MB/s, or ~1750 MB/s instead of ~3500 MB/s)
Drive achieves full speeds initially, but drops severely to SATA-like speeds (< 500 MB/s) mid-benchmark or under sustained load
Sequential speeds are fine, but Random 4K Q1T1 / Q32T1 IOPS and low-queue-depth reads are abnormally low
Drive reads at full rated speed, but Sequential Write speeds are permanently capped below 1000 MB/s across all profile tests

What link parameters are reported by HWiNFO or CrystalDiskInfo under 'Transfer Mode'?

Drive operates at PCIe 3.0 x4 on a PCIe 4.0 drive, or PCIe 4.0 x2 instead of x4
Drive operates at PCIe 3.0 x2 or PCIe 4.0 x1 mode due to M.2 slot lane-sharing with SATA/SATA Express/PCIe slots
Drive is seated in a secondary motherboard slot connected via Chipset (PCH) instead of Direct CPU Lanes
Drive link reports supported PCIe 4.0 x4 but active mode switches dynamically to PCIe 1.1 / 2.0 and fails to ramp up under load

PCIe Generation or Lane Allocation Downgrade Fix

Solution:

Root Cause: PCIe Generation or Lane Count Negotiation Misconfiguration

When a PCIe 4.0 x4 NVMe SSD (capable of ~7000 MB/s) operates at PCIe 3.0 x4 (~3500 MB/s) or PCIe 4.0 x2 (~3500 MB/s), the root complex has negotiated a restricted bus width or generation interface. This occurs when the drive is installed in a legacy CPU/chipset slot, or BIOS Auto-Negotiation fails to initialize max Gen speed during the PCIe Link Training and Status State Machine (LTSSM) sequence.

Diagnostic Verification:

Download and launch HWiNFO. Navigate to Bus -> PCI Bus -> [PCIe Root Port] -> NVMe Controller.

Compare Original Link Width (e.g., x4) and Maximum Link Speed (e.g., 16.0 GT/s for Gen4) against Current Link Width and Current Link Speed.

Execute the following PowerShell command to view physical link status via Windows Storage Management:

powershell
  Get-PhysicalDisk | Get-StorageReliabilityCounter | Select-Object DeviceId, Temperature, Wear
  Get-InitiatorPort | Select-Object NodeAddress, PortAddress

Step-by-Step Fix:

1Re-seat Drive and Verify M.2 Slot Bandwidth Assignment:

Power off PSU and ground yourself. Relocate the NVMe drive to the primary M.2 slot located closest to the CPU socket (M2_1). On modern platforms (Intel LGA1700 / AMD AM5), secondary slots route through the PCH chipset, which may cap speeds at lower PCIe generations or share bandwidth with other expansion slots.

2Configure BIOS PCIe Link Generation:

Reboot and enter BIOS setup (Del / F2).

Navigate to Advanced -> Onboard Devices Configuration or PCI Express Configuration.

Locate the target M.2 slot configuration parameter and explicitly change the link speed from [Auto] to [Gen4] or [Gen5] (matching your drive's hardware spec).

3Clean M.2 Connector Contacts and Inspect Pins:

Inspect the physical gold fingers on the NVMe M.2 edge connector. Remove dust or residue using 99% Isopropyl Alcohol and a lint-free swab. Bent pin contacts inside the M.2 slot can disrupt physical lanes 2 and 3, forcing the drive to negotiate in down-shifted x1 or x2 modes.

Prevention & Long-Term Monitoring:

Avoid installing secondary PCIe expansion cards (such as x4 capture cards or sound cards) in slots that force CPU lane bifurcation (e.g., splitting a primary x16 GPU slot into x8/x4/x4).

Reference the motherboard block diagram in your manual to confirm slot routing.

M.2 Slot Lane-Sharing Throttling Fix

Solution:

Root Cause: Motherboard Chipset Lane Sharing & Multiplexing Constraints

Motherboard vendors multiplex high-speed I/O lanes between SATA ports, secondary M.2 slots, and PCIe x1/x4 expansion slots. When specific SATA ports are populated or secondary PCIe slots contain expansion cards, the motherboard switch IC dynamically disables or bifurcates M.2 slots from x4 down to x2 or x1 modes, severely capping throughput.

Diagnostic Verification:

Consult your motherboard vendor manual section on Storage Interface Specifications and PCIe Expansion Slot Tables.

Open Device Manager, switch view to Devices by connection, and trace the NVMe controller hierarchy under ACPI x64-based PC -> PCI Express Root Complex to identify shared parent bridges.

Step-by-Step Fix:

1Identify Shared SATA and PCIe Ports:

Review the motherboard layout diagram. Common configurations disable M.2_2 x4 bandwidth if SATA ports 5 and 6 are populated, or drop M.2_3 to x2 mode if the lower PCIe x4 slot is in use.

2Move SATA Cables to Non-Conflicting Ports:

Disconnect SATA storage drives from disabled/shared ports (typically SATA5/SATA6 or SATA1/SATA2 depending on vendor) and plug them into dedicated unshared ports (e.g., SATA3/SATA4).

3Reconfigure BIOS M.2 Bandwidth Switch Setting:

Enter BIOS Setup -> Advanced -> Onboard Devices Configuration.

Find M.2_2 Bandwidth Configuration or PCI Express X4_2 Slot Selection.

Toggle setting from [X2 Mode] to [X4 Mode]. Note: This will disable linked shared onboard SATA ports or x1 slots.

Prevention & Long-Term Monitoring:

Map all onboard storage device assignments prior to expansion. For platforms requiring extensive NVMe storage, utilize dedicated PCIe-to-M.2 add-in cards in CPU-attached x16 slots supported by PCIe bifurcation in BIOS settings.

Chipset DMI Bus Bottleneck Fix

Solution:

Root Cause: PCH Chipset DMI Bus Saturation & Overhead Latency

NVMe SSDs installed in M.2 slots connected to the motherboard Chipset (Platform Controller Hub / PCH) must share a single uplink bus (Direct Media Interface / DMI) back to the CPU. If multiple background devices (Gigabit/10G LAN, USB 3.2 Gen2x2 controllers, secondary drives, audio) transfer data simultaneously over the chipset, DMI link contention causes throughput degradation and elevated IOPS latency in CrystalDiskMark.

Diagnostic Verification:

Download and run HWiNFO. Check PCI Bus -> Host Bridge / PCH Root Port bandwidth and DMI link speed (e.g., DMI 4.0 x4 vs DMI 3.0 x4).

Run CrystalDiskMark while monitoring System Interrupts and DMI Bus utilization in Task Manager / Performance Monitor.

Step-by-Step Fix:

1Relocate Primary NVMe SSD to Direct CPU Slot:

Ensure your primary OS and high-performance benchmark drive is inserted into slot M.2_1 (usually closest to CPU socket with a direct x4 CPU lane connection).

2Update Motherboard Chipset Drivers:

Download the latest chipset drivers directly from the platform manufacturer (Intel or AMD) to optimize PCH controller buffer management and ASPM latency. Refer to Microsoft Official Support Guide for Windows device driver deployment best practices.

3Disable Unused Integrated Onboard Controllers in BIOS:

Access BIOS -> Advanced Settings.

Disable unused secondary ASMedia/Promontory SATA controllers, extra USB 3.2 hubs, and secondary Ethernet/Wi-Fi chips to clear DMA buffer allocations across the DMI bus.

Prevention & Long-Term Monitoring:

Reserve chipset-connected M.2 slots for secondary storage, cold backups, or media drives where maximum sustained burst sequential IOPS are not required.

PCIe Dynamic Link State (ASPM) Stuck Fix

Solution:

Root Cause: Aggressive ASPM L1/L1.1/L1.2 Substates Failure to Exit Low-Power State

Active State Power Management (ASPM) forces the PCIe link into lower power states (L0s, L1, L1.Substates) during brief idle periods. Misconfigured NVMe firmware or Windows power policies can cause the drive controller to remain locked in L1 power-saving states during benchmark execution, preventing the link from returning to L0 full-speed operation.

Diagnostic Verification:

Open PowerShell as Administrator and query current ASPM status:

powershell
  powercfg /query SCHEME_CURRENT SUB_PCIEXPRESS ASPM

Check if the returned hexadecimal value corresponds to 0x0 (Off) or 0x2 (Maximum Power Savings).

Step-by-Step Fix:

1Configure Windows Power Plan to Maximum Performance:

Press Win + R, type powercfg.cpl, and press Enter.

Select High Performance or Ultimate Performance plan.

Click Change plan settings -> Change advanced power settings.

Expand PCI Express -> Link State Power Management and set both On battery and Plugged in to Off.

2Force ASPM Settings in Windows Registry:

Press Win + R, type regedit, and navigate to:

HKLM\SYSTEM\CurrentControlSet\Control\Power\PowerSettings\54533751-0701-4d6b-9a92-b43529b16c0f\ee12f906-d277-4e1b-b6da-0276d71a0563

Set the Attributes DWORD value from 1 to 2 to expose hidden ASPM controls in Power Options.

3Update Drive Firmware via Official Utility:

Download your drive vendor's SSD management software (e.g., Samsung Magician, WD Dashboard, Crucial Storage Executive) and flash the latest drive firmware to resolve known L1.2 exit latency bugs.

Prevention & Long-Term Monitoring:

Keep laptop and desktop power profiles set to high performance when executing storage-heavy benchmarks or compiling massive codebases.

What physical or operational conditions occur when the speed drop takes place?

Drive temperature exceeds 70°C–75°C in HWiNFO during CrystalDiskMark Sequential tests
Drive temperature is low (< 50°C), but write speeds drop off after 10–30 seconds of sustained test runs
Speeds drop severely and Windows logs 'stornvme Event ID 129' or 'Reset to device, \Device\RaidPort0, was issued'
System is a laptop running on battery power during the benchmark test

NVMe Controller Thermal Throttling Fix

Solution:

Root Cause: Thermal Throttling Triggered by NVMe Thermal Sensor (T2 / T3 Junction)

Modern High-Performance Gen4 and Gen5 NVMe controllers generate significant localized heat during sequential burst transfers. When the main controller IC reaches its thermal limit (typically 75°C to 85°C), internal thermal protection throttling forcibly drops clock frequencies and asserts dummy wait states, reducing transfer rates to SATA-like levels to protect hardware integrity.

Diagnostic Verification:

Run HWiNFO in background while launching CrystalDiskMark.

Observe Drive Temperature 1 (NAND Flash) and Drive Temperature 2 (Controller Junction).

Check for Thermal Throttling Status = Yes under the drive's SMART health tree.

Step-by-Step Fix:

1Remove Protective Plastic Film from Motherboard Heatsinks:

Turn off computer and remove motherboard M.2 metal heatsink shield. Ensure the factory protective blue/clear plastic film covering the thermal pad underneath has been removed.

2Replace or Re-align Thermal Pads:

Ensure thermal pad makes firm, direct physical contact across both controller chip and NAND flash packages. Replace compressed/torn thermal pads with high-conductivity silicone pads (e.g., 1.5mm / 6 W/mK).

3Install Active Cooling for PCIe 4.0 / 5.0 Drives:

PCIe 5.0 drives require dedicated high-surface-area heatsinks with active fan cooling or substantial chassis airflow over the CPU socket region.

Prevention & Long-Term Monitoring:

Set custom case fan curves in BIOS linked to the PCH/Drive temperature sensors to ramp up intake fan RPM when storage loads spike.

SLC Cache Exhaustion Fix

Solution:

Root Cause: Dynamic Pseudo-SLC Cache Exhaustion on TLC/QLC NAND Flash

Consumer NVMe drives utilize a dynamic Pseudo-SLC (pSLC) cache buffer to achieve high advertised write speeds. Once continuous write benchmarks write enough data to completely fill this pSLC buffer, write operations fall back to the drive's slow native multi-level cell speed (Direct-to-TLC or Direct-to-QLC), which can be as low as 100–500 MB/s on QLC drives.

Diagnostic Verification:

Run CrystalDiskMark using varying test file sizes (1GiB vs 8GiB vs 32GiB).

If the 1GiB test shows full speed, but the 32GiB test experiences a massive speed drop halfway through, the drive is encountering pSLC cache depletion.

Step-by-Step Fix:

1Execute Manual TRIM Command in Windows:

Open PowerShell as Administrator and run:

powershell
     Optimize-Volume -DriveLetter C -ReTrim -Verbose

2Provide Adequate Idle Time for Garbage Collection:

Leave the machine idle at the desktop for 15–30 minutes after heavy write cycles to allow the drive's internal flash controller to run background garbage collection and clear the pSLC buffer.

3Increase Free Over-Provisioning Space on Drive:

Maintain at least 15% to 20% unallocated free space on the drive. Dynamic pSLC cache size scales directly with available unallocated capacity on TLC/QLC SSDs.

Prevention & Long-Term Monitoring:

Select high-end SSDs featuring dedicated DRAM caches and static SLC buffers for sustained video editing or heavy write workstation workloads.

NVMe Controller Crash and Reset (storport / stornvme) Fix

Solution:

Root Cause: Kernel NVMe Driver Timeout & Bus Reset (stornvme Event ID 129)

The default Windows stornvme.sys driver initiates a controller reset when an I/O request exceeds the default hardware completion queue timeout. This manifests as a sudden benchmark lockup, followed by speed drops down to 0 MB/s, system stutter, and logged storage bus reset events.

Diagnostic Verification:

Open Windows Event Viewer (eventvwr.msc).

Navigate to Windows Logs -> System.

Filter for Source stornvme or storport and check for Event ID 129: Reset to device, \Device\RaidPort0, was issued.

Step-by-Step Fix:

1Replace Generic Windows Driver with Vendor-Specific NVMe Driver:

Install dedicated manufacturer drivers (e.g., Samsung NVMe Driver or Intel/Solidigm NVMe Driver) if supported by your drive family.

2Disable Power Management Link State (HIPM/DIPM):

Run Command Prompt as Administrator and execute:

cmd
     reg add "HKLM\SYSTEM\CurrentControlSet\Services\stornvme\Parameters\Device" /v "ForcedPhysicalSectorSizeInBytes" /t REG_DWORD /d 4096 /f

3Disable Host Memory Buffer (HMB) Allocation Allocation Regkey (DRAM-less drives):

For DRAM-less drives experiencing allocation faults, adjust HMB registry parameter:

cmd
     reg add "HKLM\SYSTEM\CurrentControlSet\Control\StorPort" /v "HmbAllocationPolicy" /t REG_DWORD /d 0 /f

Reboot system for settings to take effect.

Prevention & Long-Term Monitoring:

Ensure controller firmware is fully patched and avoid using legacy AHCI/RAID storage modes in BIOS when running single standalone NVMe drives.

Mobile Platform Battery Throttling & PCIe Power Policy Fix

Solution:

Root Cause: Laptop ACPI Power Budget Throttling & Embedded Controller Constraints

Mobile platforms restrict peak current draw on the 3.3V rail supplying M.2 slots while operating on battery power. The laptop System Management Controller (SMC) and Intel Dynamic Tuning Framework force the PCIe bus to lower link speeds and generation modes to extend battery life and reduce thermal load.

Diagnostic Verification:

Run CrystalDiskMark while connected to AC power, then repeat test on battery power.

Observe link mode differences in HWiNFO between both states.

Step-by-Step Fix:

1Connect Laptop to AC Power Supply:

Always execute storage performance benchmarks with the official OEM high-wattage power adapter connected.

2Adjust OEM Power Profile (e.g., Lenovo Vantage, Asus Armoury Crate, Dell Power Manager):

Switch device profile from Silent or Balanced to Turbo / Performance Mode.

3Enable Maximum Performance in Windows Control Panel:

Ensure Windows Power Mode slider in Taskbar settings is dragged to Best Performance.

Prevention & Long-Term Monitoring:

Perform high-throughput file transfers and disk-intensive workflows while tethered to wall power.

Which specific benchmark metrics are reporting low performance in CrystalDiskMark?

Random 4K Read/Write speeds at Queue Depth 1 Thread 1 (RND4K Q1T1) are unexpectedly low (< 30-50 MB/s)
All Random 4K IOPS metrics are throttled, and CPU usage is heavily pinned on a single core during execution
Write caching warnings or write performance suffers due to Windows Flush Policy settings
Drive is configured in Intel VMD or AMD RAID mode in BIOS rather than native AHCI/NVMe passthrough

Low Queue Depth 4K Random Read Performance Fix

Solution:

Root Cause: CPU Core C-State Power States & Latency Overhead

Low-queue-depth random 4K performance (Q1T1) measures raw latency rather than overall bus throughput. When CPU cores enter deep energy-saving sleep states (C6/C7/C8/C10) during small I/O operations, the microsecond delay required to wake the CPU core directly degrades 4K random read scores.

Diagnostic Verification:

Launch HWiNFO and observe CPU C-State Resiliency metrics during CrystalDiskMark execution.

Notice if CPU frequencies drop significantly between short test intervals.

Step-by-Step Fix:

1Disable Deep C-States in BIOS (for dedicated benchmarking):

Reboot into BIOS -> Advanced CPU Configuration.

Locate Global C-State Control or CPU C3/C6/C7 States and set to Disabled or C1e Only.

2Adjust Windows Processor Power Management:

Open Power Options (powercfg.cpl).

Edit current plan -> Processor power management -> Minimum processor state.

Set value to 100% to prevent clock down-throttling during I/O wait periods.

3Turn Off Core Isolation / Memory Integrity (VBS) in Windows 11:

Open Windows Security -> Device Security -> Core Isolation Details.

Toggle Memory Integrity (HVCI) to Off. Virtualization-based Security introduces hypervisor memory translation overhead that can lower 4K low-queue random IOPS by 10–20%.

Prevention & Long-Term Monitoring:

Re-enable VBS and CPU C-States after testing if running on a battery-constrained laptop or enterprise environment requiring memory protection features.

NVMe Queue Interrupt Steering & Core Affinity Optimization Fix

Solution:

Root Cause: CPU Core Interrupt Saturation (DPC/ISR Bottleneck)

NVMe storage devices utilize parallel hardware I/O submission and completion queues. If all storage completion interrupts are routed to a single logical CPU core (Core 0), that thread becomes fully saturated with Deferred Procedure Calls (DPCs), limiting high-queue-depth 4K IOPS scaling.

Diagnostic Verification:

Open Performance Monitor (perfmon.msc).

Add counters for Processor Information -> % DPC Time and % Interrupt Time across all individual core instances.

Run CrystalDiskMark and check if a single core spikes to 100% load while others remain idle.

Step-by-Step Fix:

1Enable RSS (Receive Side Scaling) and MSI-X in Windows:

Ensure Message Signaled Interrupts (MSI-X) are active for the NVMe controller via Device Manager.

2Optimize NVMe Queue Depth in CrystalDiskMark Settings:

In CrystalDiskMark, click Settings -> Settings.

Adjust NVMe settings to use NVMe SSD profile rather than Default (SATA) profile. This configures tests to utilize 64 Queues with 32 Threads to properly saturate parallel controller channels.

3Update Storage Controller Driver Stack:

Ensure the latest platform-specific storage drivers are active to enable multi-queue I/O routing across all available CPU threads.

Prevention & Long-Term Monitoring:

Select profile NVMe SSD under Settings -> Settings in CrystalDiskMark when benchmarking NVMe drives.

Windows Write-Cache Buffer Policy Configuration Fix

Solution:

Root Cause: Windows Write-Cache Buffer Flushing Constraints

To prevent data corruption during sudden power loss, Windows forces write commands to be committed directly to non-volatile physical flash memory (FUA - Force Unit Access), disabling volatile onboard RAM write caching. This security mechanism drastically lowers Random 4K Write speeds in benchmark utilities.

Diagnostic Verification:

Open Device Manager -> Disk drives -> Right-click your NVMe drive -> Properties -> Policies tab.

Inspect the state of Enable write caching on the device and Turn off Windows write-cache buffer flushing.

Step-by-Step Fix:

1Enable Advanced Write Caching in Device Manager:

Open devmgmt.msc, expand Disk drives, right-click target NVMe drive, select Properties.

Click Policies tab.

Check the box for Enable write caching on the device.

Check the box for Turn off Windows write-cache buffer flushing on the device.

2Apply Policy Changes via PowerShell (Alternative):

Open PowerShell as Administrator:

powershell
     Set-Disk -Number 0 -IsCacheEnabled $true

3Re-run Benchmark in CrystalDiskMark:

Verify write performance improvements. Note: Disabling buffer flushing carries a risk of data loss if the system experiences an abrupt power outage without an Uninterruptible Power Supply (UPS).

Prevention & Long-Term Monitoring:

Maintain battery backups (UPS) on critical desktop workstations if write-cache buffer flushing is permanently disabled.

Intel VMD / AMD RAID Controller Abstraction Passthrough Fix

Solution:

Root Cause: Storage Controller Virtualization Overhead (Intel VMD / AMD RAID Mode)

Intel Volume Management Device (VMD) or AMD RAID drivers intercept direct PCIe register access to create software RAID arrays. When running a standalone NVMe drive through VMD/RAID abstraction abstraction layers, native NVMe command passthrough is converted through software stack wrappers, degrading burst throughput and IOPS latency.

Diagnostic Verification:

Open Device Manager (devmgmt.msc).

Expand Storage controllers.

If listed as Intel RST VMD Controller or AMD RAID Controller instead of Standard NVM Express Controller, virtualization layering is active.

Step-by-Step Fix:

1Disable Intel VMD / AMD RAID in BIOS Settings:

Reboot system and enter BIOS Setup.

Locate Intel VMD Technology or SATA/Storage System Mode.

Set parameter to [Disabled] or [AHCI/NVMe].

*Warning:* Disabling VMD on an existing Windows installation without pre-loading generic stornvme boot drivers will result in INACCESSIBLE_BOOT_DEVICE BSOD. Enable Safe Mode prior to toggling BIOS option.

2Switch Windows Driver to Native Stornvme:

Boot into Windows Safe Mode (bcdedit /set {current} safeboot minimal).

Restart into BIOS, disable VMD/RAID mode, and save changes.

Boot into Windows. Windows will automatically attach the native stornvme.sys driver.

Remove Safe Mode flag in PowerShell:

powershell
     bcdedit /deletevalue {current} safeboot

3Restart and re-test in CrystalDiskMark.

Prevention & Long-Term Monitoring:

Only enable VMD/RAID settings in BIOS when managing physical multi-drive hardware RAID arrays.

What physical partition or drive controller factors characterize the write speed limit?

The drive is nearly full (> 90% capacity used) or unallocated partition alignment is incorrect
The drive controller is DRAM-less and Host Memory Buffer (HMB) allocation is disabled or misconfigured in Windows
Drive is formatted with an incorrect File System Cluster Size (4KB vs 64KB alignment issue)
Third-party antivirus or security software actively scans temporary benchmark files in real time

Storage Capacity Saturation & Partition Alignment Fix

Solution:

Root Cause: High Drive Utilization Overhead & Partition Offset Misalignment

When NVMe flash drives exceed 85–90% capacity utilization, the SSD controller struggles to find free flash blocks for immediate write operations. The controller is forced to execute read-modify-write cycles synchronously during benchmarks, causing severe write performance collapse. Additionally, non-4K aligned partition offsets cause single I/O requests to span across multiple physical flash pages.

Diagnostic Verification:

Check partition alignment using System Information (msinfo32). Navigate to Components -> Storage -> Disks.

Locate Partition Starting Offset. Divide the byte value by 4096. If the result is not a whole integer, the partition is misaligned.

Step-by-Step Fix:

1Free Drive Capacity Below 80% Threshold:

Delete unnecessary system temp files or migrate data to secondary storage to restore clear blocks for the controller's garbage collection routines.

2Re-Align Misaligned Partitions:

Open PowerShell as Administrator and run diskpart:

cmd
     diskpart
     list disk
     select disk 0
     list partition

If offset is misaligned, backup drive data and recreate partition with default 1024KB offset alignment using standard Windows Disk Management (diskmgmt.msc).

3Force Full Drive Re-Trim:

Execute full trim command via PowerShell:

powershell
     Optimize-Volume -DriveLetter C -ReTrim -Defrag -Verbose

Prevention & Long-Term Monitoring:

Set storage alert thresholds at 80% capacity to prevent flash block allocation bottlenecks.

Host Memory Buffer (HMB) Re-Allocation Fix

Solution:

Root Cause: DRAM-less Host Memory Buffer (HMB) Disablement

DRAM-less NVMe SSDs utilize Host Memory Buffer (HMB) technology to allocate a small portion of system host RAM (typically 16MB to 64MB) via PCIe DMA to store the drive's lookup translation table (FTL). If HMB is disabled by Windows or registry policies, lookup times freeze and write speeds collapse.

Diagnostic Verification:

Open PowerShell as Administrator and query HMB status:

powershell
  Get-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Control\StorPort" -Name "HmbAllocationPolicy" -ErrorAction SilentlyContinue

Step-by-Step Fix:

1Re-Enable HMB in Windows Registry:

Launch Command Prompt as Administrator and execute:

cmd
     reg add "HKLM\SYSTEM\CurrentControlSet\Control\StorPort" /v "HmbAllocationPolicy" /t REG_DWORD /d 2 /f

Value 2 ensures dynamic allocation based on drive requests.

2Reset GPU / Motherboard DMA Security Lockouts:

Ensure IOMMU / VT-d settings in BIOS are not blocking direct memory access (DMA) allocations requested by the storage controller.

3Reboot System and Verify with Drive Utility:

Confirm HMB status using manufacturer dashboard software (e.g., Kioxia SSD Utility or WD Dashboard).

Prevention & Long-Term Monitoring:

Avoid manually disabling system virtualization components that interfere with StorPort DMA allocations.

File System Cluster Allocation Size Optimization

Solution:

Root Cause: File System NTFS Cluster Size Misconfiguration

Formatting high-performance NVMe SSDs with non-standard file system cluster sizes (e.g., 512 bytes or 2KB) creates overhead for modern SSD controller pages natively optimized for 4KB or 64KB sector operations. Small cluster sizes cause excessive write amplification.

Diagnostic Verification:

Open PowerShell as Administrator and run:

powershell
  Get-Volume -DriveLetter C | Select-Object DriveLetter, FileSystemLabel, AllocationUnitSize

Verify if AllocationUnitSize is reported as 4096 (4KB) or 65536 (64KB).

Step-by-Step Fix:

1Backup Critical Disk Data to External Media.

2Reformat Partition with 4KB or 64KB Allocation Unit Size:

Open Command Prompt as Administrator:

cmd
     format X: /FS:NTFS /A:4096 /Q

For dedicated media or game storage drives running large sequential files, format using 64KB clusters (/A:65536).

3Run Benchmark in CrystalDiskMark to confirm speed restore.

Prevention & Long-Term Monitoring:

Standardize on 4KB allocation units for OS boot drives and 64KB allocation units for secondary media scratch disks.

Antivirus Real-Time File System Filter Interference Fix

Solution:

Root Cause: Antivirus I/O Driver Interception (Filter Drivers)

Real-time antivirus software attaches file system filter drivers (e.g., WdFilter.sys for Microsoft Defender) to monitor storage I/O operations. During CrystalDiskMark execution, these filter drivers intercept every test file write command synchronously to scan buffers, creating massive artificial write bottlenecks.

Diagnostic Verification:

Open Command Prompt as Administrator and list active file system filters:

cmd
  fltmc filters

Observe if third-party antivirus filters (e.g., CrowdStrike, Bitdefender, SentinelOne) are bound to the volume.

Step-by-Step Fix:

1Exclude Benchmark Directory in Windows Defender / Antivirus:

Open Windows Security -> Virus & threat protection -> Manage settings.

Scroll to Exclusions -> Click Add or remove exclusions.

Add folder exclusion for the path where CrystalDiskMark writes test files (typically C:\Users\[Username]\AppData\Local\Temp\).

2Temporarily Pause Real-Time Protection during Benchmark Execution:

Disable real-time monitoring temporarily, run CrystalDiskMark test, and compare results.

3Verify Performance Restoration.

Prevention & Long-Term Monitoring:

Maintain proper security exclusion rules for storage diagnostic tools on enterprise workstations.

Related Diagnostic Guides

System Interrupts High CPU Usage 100% Fix Tree

Windows 11 File Explorer Slow To Load Green Bar Fix

Mechanical Keyboard Some Keys Double Typing: Debounce Diagnostic

AirPods Connected to Windows But Sound Is Low Quality: Hands-Free Profile Fix

Printer Driver Is Unavailable Windows 11: Driver Store Cleanup Fix

Subscribe to StepDiag

Embed Guide on Your Website