Benchmark Results: 4 KB Random Performance
Random Read Performance (background info)
Examples include antivirus scans and typing in Word
SandForce-based SSDs continue to lead the pack at lower queue depths, explaining why OCZ's 120 GB Vertex 3 excels in situations with fewer than eight I/O access. However, this does not mean that Marvell-based SSDs should be avoided; their performance is quite respectable especially at higher queue depths. With the exception of OCZ's Octane, every SSD is quite capable of punching through the 300 MB/s barrier at a queue depth higher than eight.
Corsair's 128 GB Performance Pro and Plextor's 128 GB M3 Pro both exhibit the same performance characteristics in 4 KB random reads, perhaps a result of their Toggle-mode NAND. In comparison, Crucial's mainstream-focused m4s actually lead the pack of Marvell-based SSDs. Plextor's M3 series is also a mainstream-oriented SSD product. Yet, surprisingly it outperforms the Performance Pro and the M3 Pro.
These performance differences only become apparent at higher queue depths. Scaling down, the 120 GB Vertex 3 and m4s share the spotlight, achieving ~60 MB/s with only one I/O access. Meanwhile, the other SSDs all offer performance varying between 40-50 MB/s.
Random Write Performance, Compressible
Examples include email, file compression, and Web browsing
SandForce's compression technology provides chart-topping performance benefits for the 120 GB OCZ Vertex 3—until you throw incompressible data in the mix, causing performance to drop drastically. In that context, the 128 GB Performance Pro and M3 Pro both lead the pack at queue depths higher than eight.
At the lower end of the scale, the 128 GB M4 delivers ~70 MB/s in 4 KB random writes, matching the 120 GB Vertex 3. Meanwhile, the rest of the pack falls slightly behind with speeds between 50 and 60 MB/s with only a single outstanding I/O operation. Interestingly, OCZ's 128 GB Octane falls behind the 64 GB m4, an exception to the general rule that higher capacity corresponds to better performance.
