Fragmentation over time and GC
Upon completion of our test suite, we run multiple write passes to each drive to evaluate the impact of the random writes that took place.  One of the early samples I tested had Garbage Collection intentionally disabled.  This sample quickly showed itself to be a decidedly bad idea:

OCZ Colossus 3.5-in Solid State Drive Review - Storage 36

Luckily OCZ decided against shipping Colossus without the GC firmware.  Recall that GC is an OCZ-exclusive secret weapon that fits nicely when placing their controllers in a RAID configuration as they are in the Colossus.  We reviewed this firmware in standalone form back in August.  After running a GC equipped unit through our gauntlet I let the drive sit for a few minutes and checked on its progress:

OCZ Colossus 3.5-in Solid State Drive Review - Storage 37

Letting the drive idle for another 30 minutes showed further defragmentation was taking place at a fairly slow rate:

OCZ Colossus 3.5-in Solid State Drive Review - Storage 38

Something should be noted about the way this GC works as related to various benchmarking tools.  The pics above are a result of the licensed version of HDTach.  This version has an additional option that evaluates *every sector* of the drive.  Such a mode takes longer but is necessary to properly evaluate SSD sequential performance.  This was shown very well by the Colossus with GC firmware.  It seems that within an hour of heavy fragmentation, the GC process is able to rapidly correct enough of the flash blocks in such a way that the 32MB zone setting of HDTach shows the following result for a drive in the same state shown above:

OCZ Colossus 3.5-in Solid State Drive Review - Storage 39

HDTach in ‘Long’ mode showed artificially inflated results.

Given enough idle time, a Colossus should eventually rework itself back to full speed writes:

OCZ Colossus 3.5-in Solid State Drive Review - Storage 40

We saw the above only when our sample was in new condition.  It has not yet recovered on its own as it appears to take days to weeks for it to fully recover after heave fragmentation.  We’re OK with it taking a while to complete so long as it is doing *something* to actively bring writes back to full speed over time.  Remember, the GC-modified firmware is OCZ’s baby – no other Indilinx drives out there have such a capability and rely on TRIM only to restore write speeds to full speed.
« PreviousNext »