My Windows 11 PC occasionally becomes extremely sluggish or freezes while the NVMe drive shows 98–100% active time. Response times have reached about 95 ms, and Steam downloads or installations can slow dramatically during these episodes.
The system has an Intel i7-14700F, RTX 5070 Ti, 32 GB DDR5, a Gigabyte B760M C V3 motherboard, and an AGI2T0G43AI818 2 TB NVMe SSD running at PCIe 4.0 x4. The drive has about 447 GB free after I cleaned it up. CrystalDiskInfo reports 100% health, 40°C, no critical warnings, no media or data integrity errors, and no error-log entries. It has approximately 21.5 TB of host writes, 65.5 TB of host reads, and 1,622 power-on hours.
CrystalDiskMark produced these results with a 1 GiB test: sequential read/write of 5280/550 MB/s at Q8T1 and 3781/2587 MB/s at Q1T1; random 4K results were 688/154 MB/s at Q32T1 and 78/253 MB/s at Q1T1. With an 8 GiB test, sequential performance fell to 1886/452 MB/s at Q8T1 and 3215/286 MB/s at Q1T1, while random 4K results were 699/334 MB/s at Q32T1 and 62/57 MB/s at Q1T1.
The sustained write speeds of roughly 300–450 MB/s seem low for a PCIe 4.0 drive. Could this simply be SLC-cache exhaustion or a limitation of the controller and NAND, or might the drive be defective? Could that explain the 100% active-time spikes and Windows freezes? What tests should I run before replacing it, and would moving Windows to a higher-end drive such as a Samsung 990 Pro likely solve the problem?
3 Answers
Those results can be normal for an inexpensive, DRAM-less NVMe drive once its pseudo-SLC cache is exhausted. The smaller benchmark may fit mostly within the cache, while the larger test forces the drive to write directly to slower NAND. Steam installs and updates can trigger the same behavior, so 300–450 MB/s sustained writes are not automatically evidence of failure.
A 100% active-time reading means the drive is busy or waiting on an operation; it does not necessarily mean it is transferring data at maximum throughput. If the system only becomes slow during large writes, the drive’s design is a likely explanation. A better drive with onboard DRAM and a larger, more consistent cache should handle those workloads much better.
The drive is not a familiar high-end model, so it is worth checking its specifications and independent reviews for sustained-write behavior. Some budget models use QLC NAND and a dynamic SLC cache, with performance dropping sharply after the cache fills. That would fit the larger-test results, especially with games and updates that write a lot of data.
Check for a firmware update, run the manufacturer’s diagnostic tool if one exists, and review Windows Event Viewer for disk, controller, or storage-reset errors. Also verify that the SSD is properly seated and has adequate motherboard heatsink contact. A temperature of 40°C at idle or light load is fine, but monitor temperature during a long write test because thermal throttling can compound the slowdown.
SFC, DISM, and CHKDSK can rule out damaged Windows files or a file-system problem, but they will not make a slow NAND or exhausted cache perform faster. Keep a current backup before running repair commands or changing storage settings. If diagnostics show errors, the drive disappears from the system, or freezes occur during ordinary activity, replace or warranty the SSD rather than relying only on benchmark scores.
If the problem is limited to heavy disk usage, moving Windows and active game installs to a quality TLC drive such as a 990 Pro would probably improve consistency. It is not a guarantee, though—confirm the current drive’s behavior and check for other processes saturating storage before buying hardware.
The benchmark’s normal health result does not rule out intermittent stalls. If the PC also freezes during light use or web browsing, rather than only during heavy writes, I would treat that as more suspicious and investigate the drive, firmware, drivers, and system logs.