Fragmentation over time
Upon completion of my round of testing I normally run a single pass of HDTach to get a feel for any fragmentation effects caused by some fairly heavy benchmarking. Here is the result:
Post-bench HDTach: 240.7 MB/sec read | 77.9 MB/sec write
After the above pass, performance returned to 82.5 MB/sec writes. I immediately followed this test with a 30 minute pass of the very same tool used to rapidly fragment the early version of the X25-M:
Reads remain at 240 MB/sec while the write average dropped to 66 MB/sec.
My tool performs full spectrum random writes across the entire drive, leaning heavily on single sector. The tool was meant to prompt the early X25’s into their ‘point of no return’ scenario for demonstration to Intel. While it is possible that extremely heavy VM use, multiple OS installs, and other heavy random write patterns could get the original X25-M to fall into a black hole, that was corrected with a firmware update. All of that considered, it would be extremely difficult for normal use to get one of these G2 drives to the point demonstrated above. As you can see from the graph, the drive is correcting itself as the writes are taking place, meaning more typical usage would cause the drive to stick much closer to 80 MB/sec. An additional pass should verify a completely defragmented unit:
The X25-M G2 back in its happy place. Sequential write speeds returned to 82.5 MB/sec.
It seems Intel has tweaked the write combining algorithms a bit in an attempt to increase random write IOPS. The down side being some slight hints of G1 pre-8820 firmware. That said, the fragmentation is very minor and is unlikely to be seen by the vast majority of end users. Sure it’s there, but not enough to be a cause of concern.
