Fragmentation Over Time / TRIM
Fragmentation Over TimeThe Intel drives don’t do any sort of funky SandForce-based compression, meaning we can use HDTach as a rough guide of internal fragmentation. As a baseline, here are the first two passes of a brand new drive:
You can see the Intel firmware on the Marvell does very much the same that it did even back on the X25-M G1. It treats all flash as a pool of storage, even the spare area. You can tell this because that last higher speed ‘bump’ in speed is present on the second write pass, showing ~12GB worth of spare area is present on the drive.
Now for the first pass after our test suite completes:
The above looks different from the usual punishment because TRIM was at work on the 510. Formats that took place in the middle of our testing were able to push it back to a fresh state, meaning our result is only showing the last few benches before the end of the suite. This is a good indicator that in normal use under a TRIM enabled OS, the 510 will do its best to keep performance closer to new than used.
I hammered on the drive a bit and then performed a format under Windows 7. Then I ran two more HDTach passes:
In theory these should have looked like the first two, but they don’t. The reason? The drive was heavily fragmented, and there seems to be a Windows 7 limitation or timeout on just how much of the drive it will TRIM during a given operation (in this case the format operation). I’m still looking into this, but I will say that it’s a bit of an extreme scenario where the drive goes an extended period without TRIM. Under normal or even heavy use within Windows 7, TRIM should have no issue keeping a 510 in a relatively fresh state.
Overall I was impressed with the TRIM implementation of the 510 Series.
