Best Way to Compare SSDs for Real-World Speed Using PassMark PerformanceTest

[Aria James]

Hi everyone,

I'm working on upgrading the storage in several workstations and want to ensure that the SSDs I select deliver noticeable improvements in typical workloads, not just synthetic results. I've run PerformanceTest and examined Seq Read/Write and IOPS, but I'm wondering how these numbers translate to daily tasks such as opening applications, transferring large files, and multitasking. Can anyone share best practices for interpreting SSD benchmark results from PassMark? For example, which specific disk tests should I focus on, are there threshold values I should look for, and how do these benchmarks compare to real-world performance? I'd also love to hear about any tweaks to test parameters or settings in PerformanceTest that help reflect actual usage. Thanks in advance for your insights!
Here are some references I've gathered during my research:
https://www.youtube.com/watch?v=3HxrIJJXYjk
https://www.tomshardware.com/reviews/best-ssds
https://www.anandtech.com/show/16127...d-roundup-2024
https://www.theengineeringprojects.c...e-in-2024.html

[David (PassMark)]

There isn't a just one "real world usage" case. There are thousands.
For example some applications open hundreds of small XML and image files when then load (3D games especially) while others are a single .exe file. (there is also the pre-fectch service & caching to consider). File transfers depend on the file size and if the source and destination are the same drive or not. Multitasking is very nebulous. SSD's can also get slower when full, plus they can also slow down dramatically under sustained writing or if they overheat. So working out a good benchmark isn't as trivial as you might expect.

The best benchmark is timing your own use case. For example if this was a database server, time how long your most common queries take to execute.
If this was a development machine, time how long a large compile takes to execute.

Otherwise PerformanceTest offers lots of different options for customizing the disk benchmark tests via the advanced options.

Most people look at the biggest numbers when evaluating drives (e.g. threaded high queue depth sequential read). But in my opinion the IOPS4KQ1 value is far more important as this is were drives are weakest and many real world cases are similar to this.