Benchmark Results: 128 KB Sequential Performance
Sequential Read Performance
Examples include file copying, transcoding, game level loading, some gameplay, watching video, and editing video
Except for the Agility 4, all of the SSDs approach 550 MB/s in our sequential read test at a queue depth of four and higher.
Oddly, Samsung's 830 is the fastest performer at a queue depth of one, delivering ~470 MB/s. The 840 Pro isn't too far behind at 410 MB/s, effectively tying the Vertex 4. Meanwhile, the Vertex 3 and m4 appear in the middle of the pack, pushing ~370 MB/s.
OCZ's Agility 4 and Corsair's Neutron GTX start the test lagging, although the Corsair drive quickly ramps up above 500 MB/s at higher queue depths. Unfortunately, OCZ's new mainstream drive appears to get stuck at a maximum throughput just north of 400 MB/s.
Sequential Write Performance
Examples include Application Installation, Document Backup
SandForce's second-gen controllers were the first to saturate the SATA 6Gb/s interface with sequential writes. Crucial's m4 debuted a month later, but could only offer about 60% as much performance. The caveat is that SandForce's drives rely on compression to achieve those numbers. Even in a worst-case scenario, though, the 240 GB Vertex 3 is slightly faster than the m4 when it comes to dealing with incompressible data.
Samsung's previous-generation 830 delivers up to 400 MB/s, which makes it strong competition for SandForce-based drives. However, that's only the case in environments with an even mix between compressible and incompressible data. Presented with completely compressible data, the Vertex 3 rises to the top.
Fortunately, the 840 Pro doesn't force us into all of those conditional statements. It's a notable improvement over the 830, matching what SandForce's technology relies on compressible data to do (except that Samsung can do it all of the time).
The Vertex 4 was the first drive we saw to achieve something along the same lines, and Corsair's Neutron GTX matches both competitors by shooting up to 500 MB/s or so at a queue depth of two. It's amazing what a year or evolution in this segment can do.
