I've read this article, as well as a previous PDF released by LSI in December 2012:
»www.flashmemorysummit.com/Englis ··· mith.pdf
The whole thing smells/reeks of marketing-driven justification, not technical. I get the impression these jackasses are trying to try and justify, with NAND/RAM-based media, the whole "1000 bytes to a kilobyte" nonsense that MHDD vendors introduced for no justified reason.
If there is truly technical justifications for what the "over-provisioning" areas are used for (citing: 1) Garbage collection, 2) controller firmware, 3) spare blocks, 4) other methods)), then the manufacturers should actually be doing the over-provisioning themselves by including an additional NAND flash chip
dedicated to exactly those things.
I'm also utterly amazed that the drive firmware is stored on NAND; that seems like an accident waiting to happen. EEPROM or dedicated Flash (not NAND) chip would be a much better choice. Or even better: integrate that necessity into the underlying FSP itself, much like is done on PIC chips, many of which have their own on-die dedicated RAM.--
Making life hard for others since 1977.
I speak for myself and not my employer/affiliates of my employer.