Sequential Read Performance
Crucial's 128 GB m4 almost ties the 120 GB Vertex 3 in 128 KB sequential reads. Plextor's M3 Pro also keeps pace, though it falls behind at a queue depth of two. Meanwhile, Plextor's 128 GB trails by only 20 MB/s at queue depths lower than four.
Corsair's 128 GB Performance Pro exhibits performance characteristics similar to OCZ's 128 GB Octane, but manages to pull away when moving up to four or more outstanding commands. At the lower end of the capacity scale, Crucial's 64 GB m4 clearly beats Plextor's 64 GB M3, with the Crucial SSD surging to ~370 MB/s and the Plextor drive down at 300 MB/s with a single command outstanding.
Sequential Write Performance
Examples include Application Installation, Document Backup
Aside from the standout performance of OCZ's 120 GB Vertex 3 operating on compressible data, this chart is interesting because performance is clearly grouped into two prominent clusters. The 128 GB Corsair Performance Pro and the 128 GB Plextor M3 Pro deliver 300 MB/s in sequential writes right out of the gate, rising to ~350 MB/s as queue depth increases. Both drives demonstrate the impressive performance that a Marvell-based SSD can achieve with optimized firmware and Toggle-mode NAND.
In contrast, the 64 GB M3, 128 GB M3, 128 GB Octane, 64 GB m4, and 128 GB m4 all achieve similar performance to each other, regardless of queue depth. With the exception of the 64 GB m4, all of these drives deliver about ~200 MB/s in 128 KB sequential writes. The inclusion of the 128 GB M3 in this group might seem a little odd considering it features Toggle-mode NAND, just like the M3 Pro. However, Plextor states the non-Pro variant doesn't feature the same set of optimizations, which is why performance is capped at a lower level.
- Differentiating With Marvell's SSD Controllers
- Test Setup And Firmware Notes
- Benchmark Results: 4 KB Random Performance
- Benchmark Results: 128 KB Sequential Performance
- PCMark 7 And Power Consumption
- Examining Steady-State Performance
- SSD Management: Problems With Secure Erase
- Two Marvell-Based Stand-Outs Emerge