photonboy :
METTEEC,
A lot of people don't understand TBW. A higher capacity drive can obviously handle more writes (assuming everything basically equal).
The article suggests that maybe the reason for the 4TB and 2TB being the same has to do with the amount of OP (Overprovisioning) but I don't agree. I think they're just being really approximate and lumping drives together.
The reason the OP amount makes no sense is because SSD's use wear leveling algorithms, so the memory ends up wearing about the same on average. I'm not sure how a bit more OP is going to double the write endurance of the memory (2x capactiy, same TBW). Without getting into details that just doesn't make much sense.
Obviously more memory whether OP or not will help with TBW but a 2TB with some extra OP exactly matching a 4TB with proportionately less OP memory for TBW seems incredibly unlikely.
Extra overprovisioning serves several purposes, and it does have a huge impact on endurance and performance. The controller can dynamically allocate free blocks to replace failed ones, which allows the drive to keep on chugging even after a significant number of expired blocks. The more blocks you have, the more that you can designate to replace failed ones, thus increasing life. More OP also reduces other intangibles, such as read-disturb, and allows more efficient operation in general.
There are two types of wear leveling - static and dynamic rotation. Dynamic wear leveling distributes incoming data among blocks to assure an even wear rate. However, if the blocks are populated with data they aren't written to, so that creates a huge mismatch, which is where static rotation steps in. Static rotation is constantly shuffling around the existing data on the SSD in the background to assure that populated blocks are not only written once, which allows it to spread that additional endurance among the free pool.
Too much static data rotation can exacerbate read-disturb wear, so the algorithms are quite sophisticated to determine when, and where, to move data. Some may even forgo static data rotation entirely to avoid additional wear on weak cells, or the cells around them.
OP certainly increases endurance. Take the Micron 9100 we recently reviewed as the perfect example. The same model, one with 2.4TB of addressable space (from a 4TB pool) has a 6.57 PB TBW rating, while the same SSD with 4TB of flash, but more addressable area (3.2TB) provides only 3.28 PB of endurance. All other factors, including DRAM allocation, NAND, controller and firmware, are all the same, but the doubled OP provides a neat doubling of endurance. However, there is a diminishing point of returns around 50 to 60 percent OP (dependent upon controller and other factors). After that you are sacrificing too much capacity for negliglble gains.
OP also increases random write performance, the 2.4TB model his 300,000 IOPS, the 3.2TB hits only 160,000. Again, almost a double in this metric due to doubled OP.
http://www.tomshardware.com/reviews/micron-9100-max-nvm...