Test Setup

Synthetic Testing

• Tests were run with a 30-second-long warm-up time (result recording starts at second 31).
• Between each test, the drive was left idle for 60 seconds to allow it to reorganize its internal data.
• All write requests contain random, incompressible data.
• Disk cache was flushed between all tests.
• NVMe drives were tested without any additional heatsinks or fans blowing on them unless indicated otherwise. The case does have a normal level of airflow.

Real-Life Testing

• After initial configuration and installation, a disk image was created; it is used to test every drive.
• Automated updates were disabled for all programs. This ensures that—for every review—each drive uses the same settings without any interference from previous testing.
• Our disk image consumes around 150 GB—partitions were resized to fill all available space on the drive.
• Partitions were properly aligned.
• Disk cache was flushed between all tests.
• In order to minimize random variation, each real-life performance test was run twelve times with reboots between tests to minimize the impact of disk cache. One outlier on each side was removed and the ten remaining runs were averaged for the final score.
• All application benchmarks run the actual application and do not replay any disk traces.