I found out the combination SB controller / SSD does a lot to NAND writting.
Intel chipset are winners but some Nvidias (old Nforce) ■■■■ badly. AMD are fine too, although they tend to be slower than Intel’s in benchmarks, for the 2008-2010 chipsets. Later on, they probably all went better, but as far as I know P67 Z87 Z97 rock the boat !
I asked for a RMA replacement, as suggested here, and the service was really good. UPS paid return, and fast delivery for a disk bigger in size. Frankly, even though that can be annoying, I am not sure whether other manufacturers would have done the same, and nothing tells whether other SSD drive may not show the same problem on your rig.
2246XT is a cacheless drive(there is little bit cache in controller but only enough for block erase and firmware) so you have to be careful.
You can do
-> Add user level provision ( just shrink your partitions and create minimum 8GB unallocated space)
-> use caching programs like Primocache. It can keep waf around 1,4 - 1,7 if you have enough free space.
My options are :
->Defrag disk !!!WHILE PRIMOCACHE IS RUNNING(if you dont want to kill your disk)!!! every 3 week.
Defragging will cause small files in several big block to store in one single block. (ex : you have 5 files in 5 block , while defragging process they will get into only 1 block and 4 block will be free for wear leveling)
It can consume 1-2 write cycle but it can save pretty much write cycle.
This is log of my 7months(dont look at power on count , it only counts time when disk is active) old of ssd plus 120GB. beginning of graph jumps really much but graph erased after reinstalling windows . somewhere of the middle and end of graph it just jumped so fast because i filled drive %95.
