Basically, this is down to the number of NAND chip packages on the SSD, and the density of these packages. Almost all modern SSDs have a controller that can use multiple channels to read and write to the NAND. NAND is rather slow on its own. SSDs get their speed from reading and writing to several NAND packages at the same time. The sweet spot will generally be SSDs with 8 channels addressing 16 NAND chip packages.
Again we must go back to MLC NAND basics, and the read, modify, NAND block write method. But if the SSD controller is smart and fast enough, why just do one of these processes at a time? In actual fact, they don’t. While the SSD is busy doing the writing process on one block, it can also use another channel to do the read and modify. This is called interleaving, but unfortunately, 16 NAND chip packages are required to get the best out of this method with an SSD controller which supports 8 channels to the NAND array.
This makes perfect sense on the larger capacity SSDs. For example for an SSD with 256GB of NAND, you can use 16GB NAND chip packages, and for a 512GB SSD, you can use 32GB packages in order to get the magic 16 NAND chip packages. Unfortunately trying to maintain these 16 NAND chip packages on small-capacity SSDs would be prohibitively expensive and would result in small-capacity SSDs being non-competitive.
Things may change in the not-too-distant future. ONFI 3 NAND will soon be available, supporting speeds of up to 400MB/s per NAND die. So it is certainly possible that only 4 or 8 NAND chip packages are required to fully saturate the SATA 6Gbps system bus. If this should happen then smaller capacity SSDs, at least for sequential reading and writing could be every bit as fast as their larger counterparts.
