SLC SSD
An SLC SSD uses NAND operated with one stored bit per cell, represented by two distinguishable threshold-voltage states.
Cell mode. Native SLC media, full-drive pseudo-SLC operation and a temporary SLC-mode cache in a multi-bit SSD are different arrangements. The presence of an SLC cache does not make the whole drive an SLC SSD. Swissbit describes pSLC as single-bit operation of supported multi-bit NAND. [5]
Historical evidence. Grupp, Davis and Swanson studied 45 raw flash chips from six manufacturers, spanning 72 to 25 nm. Figure 1 compares chip-rated P/E limits with raw bit error rates measured at those limits, not observed SSD failure lifetimes. Table 2 gives fitted raw-chip latency trends; TLC predictions had limited supporting data. [3]
Drive qualification. A two-state encoding leaves more sensing margin than denser encoding over a comparable voltage range, but cannot universally rank complete drives. [4] Check the device's TBW or DWPD conditions, capacity and workload: write amplification, error correction and over-provisioning also affect endurance. [6]
Relations
- related
- Single-level cell (SLC), Enterprise SSD, Pseudo-SLC (pSLC), SLC cache, MLC SSD, TLC SSD, Program/erase cycle, TBW, DWPD
Advantages
- Two-state encoding trades capacity per cell for wider voltage margins than denser modes over a comparable range. [4]
Limitations
- One-bit operation stores one third as many bits as TLC at the same cell count; this is not a usable-capacity ratio for arbitrary SSDs. [4]
- A chip P/E rating is not a complete-drive TBW specification or a prediction of field failure time. [3, 6]
Citations
- Grupp, Davis and Swanson. The Bleak Future of NAND Flash Memory. USENIX FAST 2012.
- KIOXIA - Multi-level cell technology: bits and threshold-voltage states
- Swissbit - Storage technology: pSLC mode
- KIOXIA - How Does Endurance Work in SSDs?, March 2021, pages 2-4
Further reading
- SLC SSD
- Cai, Ghose, Haratsch, Luo and Mutlu. Errors in Flash-Memory-Based Solid-State Drives: Analysis, Mitigation, and Recovery. arXiv 1711.11427v2, 2018.
Sources & caveats
- FAST 2012 section 2 and Figure 1 distinguish rated cycling limits from measured raw errors; Table 2 contains historical latency fits. [3]
- The FAST 2012 authors list UC San Diego and Microsoft Research affiliations. [3]
- This is a source review, not a modern-drive benchmark or a replication of the raw-chip experiments.