Are there still good reasons to "preconditioning" your SSD, especially before doing benchmark? And if so, why? Today we have better SSDs and almost all system-configurations support TRIM.

  • Found this note in regards to read benchmark and new disks: "..reads might won't make it all the way to the nand if the mapping table is empty" (spinics.net) – MrCalvin Oct 31 '20 at 13:14

The answer is yes. The professional way of benchmarking an SSD is described.

Before a benchmark is done, if not brand new, the SSD has to be purged to put it into a state that is as close as possible to fresh out of the box. This is usually done by a manufacturer utility that can TRIM/erase the whole disk.

The second step is to apply a workload repeatedly to the SSD, while monitoring its performance, until performance results get into stable state where the repetitions don’t fluctuate by too much.

When the SSD is in a stable condition, the benchmark should be run immediately, not giving time to the firmware to do any garbage collection that will affect future results.

  • Are you saying it's all about a full-disk-trim, which I always do and can be done by blkdiscard -v /dev/xx on linux). It was my understanding there were more to it, like writing the whole disk one or multiple times with data for some reason? – MrCalvin Oct 31 '20 at 11:36
  • I'm saying that after TRIM the state of the disk is abnormal and a benchmark would be unrealistic. The above procedure aims for a realistic state. – harrymc Oct 31 '20 at 12:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.