DiskSpeedTest Download
Home/Results
Interpretation

What the four numbers actually say.

A result is only as useful as the reading of it. This page maps each figure to the thing it predicts, and points out the places where a good number still hides a slow drive.

read

Sequential read

Large blocks, in order. It sets the ceiling for copying or loading big assets, and it is the number vendors print in the largest font. On its own it says little about desktop feel.

Look for: stability across passes. A drive that drops 30 percent on pass three is throttling.

write

Sequential write

Large blocks written in order, with cache flushing between passes. This is where the gap between a good controller and a cheap one shows up first.

Look for: a cliff in the endurance run. SLC caches fill and throughput lands at the true TLC rate.

4K

Random 4K

Small blocks read from unpredictable offsets, reported in IOPS. It is the closest single figure to how the drive feels during normal work: app launches, file indexing, game streaming.

Look for: IOPS collapsing when queue depth rises. That points at a weak controller or a saturated link.

latency

Access latency

The time between a request and the first byte back. NVMe drives sit near 0.04 ms, SATA near 0.08 ms, spinning disks near 8 ms. Below a tenth of a millisecond, changes are hard to feel.

Look for: latency spikes under load. Average latency hides them, which is why the report shows the 99th percentile too.

sample

A full report, annotated.

This is the standard profile on a healthy Gen4 NVMe drive. The shape of the numbers matters more than the exact values.

MetricResultPercentileReading
Sequential read7,412 MB/sp50At the interface ceiling for Gen4 x4
Sequential write6,180 MB/sp50SLC cache holds for the whole pass
Random 4K read214K IOPSp50Healthy controller, no queue collapse
Access latency0.038 msp50Typical for a good NVMe drive
Latency, 99th0.31 msp99A few outliers, still well under a millisecond
Throughput drop0%p50No thermal or cache throttling detected
watch for

Good numbers, bad drive.

Three patterns that look fine at a glance and mean trouble on a second read.

  • Write speed that halves after twenty seconds: the SLC cache has filled
  • Read IOPS that collapse as queue depth climbs: controller saturation
  • A 99th percentile latency many times the median: background retries
  • Numbers that fall every pass: heat, not the drive
endurance write4 GB
0 s6.1 GB/s
20 s5.9 GB/s
45 s2.4 GB/s
90 s2.1 GB/s

The drop at 45 seconds is the cache running dry. Sustained speed is the lower number.

compare

Benchmark the drive you have.

Run the standard profile, keep the report, and compare it against the next drive you buy.