Benchmark Results: CrystalDiskMark Streaming Performance
We recently added CrystalDiskMark to our benchmark suite after discovering an issue with h2bench2’s write testing methodology and the way it interacted with SandForce-based SSDs.
For the most part, CrystalDiskMark corroborates the sequential performance seen in Iometer. The RevoDrive X2 and IBIS are both on par, followed by the RevoDrive. The only exception is the Vertex 2 versus X25-M battle, where the SandForce-powered Vertex 2 falls behind in CrystalDiskMark due to our use of random data in that test. For a more detailed explanation, see below.
It’s important to explain differences in the way that Iometer and CrystalDiskMark measure performance here. Iometer doesn’t use completely random data in its write test, giving the SandForce-based drives a large advantage. We configure CrystalDiskMark to use completely random data, however, which is far more difficult for SandForce’s DuraClass architecture to compress, and Intel starts to look significantly better in comparison.
The RevoDrive’s performance is more than halved, While Intel’s perceived sequential write performance doubles. This is enough to put the 160 GB X25-M just behind the RevoDrive X2.
Even more interesting is the margin separating the IBIS and RevoDrive X2, though, considering both of these devices employ very similar designs. According to OCZ, the results are indicative of what happens when the drive is in a very dirty state. There is a tool available on its Web site to secure-erase the card, returning it to a new state. However, that's not going to be an option for most users, so the significantly reduced write performance is actually something to keep in mind.
