[SOLVED] storage speed difference recognized by human...

Pextaxmx

Reputable
Jun 15, 2020
418
59
4,840
I handle lots of large and small files around for engineering data logistics, and wanted to describe (nothing scientific) my take of real world SSD speed, to see if you guys agree.

5400 RPM to 7200 RPM HDD: quite noticeable
7200 RPM to SATA SSD: day and night... can't go back
SATA SSD SATA2 port vs SATA3 port: surprisingly hard to notice
Old Slow planar TLC SATA SSD vs modern SATA SSD: more noticeable than SATA2 vs SATA3 for writings
SATA3 to NVMe: surprisingly hard to notice. Things feel snappier but barely noticeable... Large files obviously transfer significantly faster but not for long.

I think maxing out SATA3 bus speed is pretty much the threshold of human sensation of SSD speed. We don't seem to benefit a lot by going faster than that.
 
Solution
Things could potentially change in the future as software is optimized for faster storage though. Currently, nearly all software is still expecting that people may be running it off a platter-based drive, and avoids accessing data in ways that would result in performance being unusable on such a system. Eventually, SSD storage is likely to become a requirement for some software though.

An example might be for games designed specifically for the new consoles and higher-end PCs, where SSDs are a given, and data can be streamed directly from storage to graphics memory on an as-needed basis. It's possible that they may be able to cache the data in system memory as a fallback when fast-enough storage isn't present, but that may come with...
In general your observations are correct. For this upgrade path:

SATA-II -> SATA-III -> PCIe 3.0 NVMe -> PCIe 4.0 NVMe

you will notice a difference only at working with large sequential files on reading or writing huge chunks of data (few gigabytes at once). Basically on data backup or video editing. Or for PCIe 4.0 NVMe with RTX IO capable GPUs on huge texture array loading (still theoretically). Otherwise SATA vs NVMe "on look and feel" are nearly equal.
 
I handle lots of large and small files around for engineering data logistics, and wanted to describe (nothing scientific) my take of real world SSD speed, to see if you guys agree.

5400 RPM to 7200 RPM HDD: quite noticeable
7200 RPM to SATA SSD: day and night... can't go back
SATA SSD SATA2 port vs SATA3 port: surprisingly hard to notice
Old Slow planar TLC SATA SSD vs modern SATA SSD: more noticeable than SATA2 vs SATA3 for writings
SATA3 to NVMe: surprisingly hard to notice. Things feel snappier but barely noticeable... Large files obviously transfer significantly faster but not for long.

I think maxing out SATA3 bus speed is pretty much the threshold of human sensation of SSD speed. We don't seem to benefit a lot by going faster than that.
For Joe User the sweet spot is ssd and sata 3.
Move above that and the gains are quite small.
Benchmarks show nice big numbers but that's about it.
 
Things could potentially change in the future as software is optimized for faster storage though. Currently, nearly all software is still expecting that people may be running it off a platter-based drive, and avoids accessing data in ways that would result in performance being unusable on such a system. Eventually, SSD storage is likely to become a requirement for some software though.

An example might be for games designed specifically for the new consoles and higher-end PCs, where SSDs are a given, and data can be streamed directly from storage to graphics memory on an as-needed basis. It's possible that they may be able to cache the data in system memory as a fallback when fast-enough storage isn't present, but that may come with an additional performance penalty, or with increased pop-in of assets.

Of course, a lot of software is unlikely to see much benefit from faster SSDs anytime soon.
 
  • Like
Reactions: Krotow
Solution