OK, So I hear people talk about about manually over-provisioning to ensuer they don't slow down as they fill up. Since Crucial/Micron C300's were brought up along some point, I ran some benches. As to the particular benches I use here, it's because they are easy to run and don't take a long to time (as I hate sitting around waiting for benches to finish). As to whether these are the best benches to run or otherwise, I don't really care, interpret the numbers as you will..
This test was run with a Micron/Crucial C300 128GB, standard Win7 partitioned and formatted to max normal size (119GB). This was run on a ICH10R Intel SATA port (hence the speeds top out at SATA2 speeds. AND FYI: I have never actually tried a test of this sort in the manner I did it.
Testing methodology:
1. Partition and format, run 1st benchmark (0% full).
2. Copy over a bunch of movies and other junk to fill it to 50%, run 2nd benchmark.
3. Copy more crap to 70% full and run benchmark.
4. Finally copy even more garbage onto disk to fill it to 90% (actually 91% because I overshot a little)
I did not do anything special between data copy, or benches. I simply did each in fast order (didn't want to take all night). Bench, copy crap to fill drive, bench, copy more, etc, etc, one after another. Seems to me that the numbers aren't showing much of a slowdown. Crystaldiskmark verifies how full it thinks the SSD is at the top right of the window.
0% full
50% full
70% full
90% (actually 91%) full