One of the least-understood aspects of SSDs and NAND flash memory in general is the impact of wear & tear on the life expectancy of an SSD. Now, to be clear, the projected life span of NAND flash is understood well enough by the NAND flash manufacturer to offer recommended maximum usable lifespans (in terms of program / erase cycles – or p/e cycles). But while the theoretical maximums are all well and good to have, it would be nice to have some independent verification of the validity of those limits. Thanks to the many, many terabytes of data that will soon be written in endurance tests to several SSDs by the fine folks at The Tech Report, we’re about to find out . . . soon-ish.
Their strategy is simple: write a boat-load of data to a set of SSDs from different manufacturers, and see how long they last before they give up the ghost. Why hasn’t this been done for traditional spinning magnetic hard drives, you ask? Because while there are statistically-significant MTBF’s (Mean Times Between Failures) for traditional hard drives, there is no fundamental limit to how much data can be written to a hard drive. For SSDs, there is a theoretical limit based on how NAND flash operates.
The idea behind The Tech Report’s testing scenario is to find out (generally, and with a small sample size) how close the theoretical / practical limits to NAND Flash endurance is to the actual limit. Much as clockspeed limits on processors are very conservative relative to the possible speeds that can be extracted by skilled over-clockers, the useful lifespan of an SSD may very well be much longer than what the manufacturer recommends.
To put this all into perspective, though, NAND flash endurance has never really been a huge issue for the average consumer using an SSD. The theoretical limits are only reached after a long time or a lot of data being regularly written to the drive. Most consumers will get a new computer or a bigger, better SSD upgrade before the NAND flash wears out. Even with SSDs using TLC (Triple Layer Cell) NAND with the lowest overall p/e cycle ratings, the folks at Anandtech have determined it won’t matter for the average consumer.
But for the long term trend in SSDs, endurance is a concern as smaller process nodes decrease the write endurance and require more complex and robust error-correction methods. Fortunately, there may be light at the end of the tunnel that will allow for Moore’s Law (or at least a facsimile of it) to continue unabated for SSD price, performance, and storage improvements: Samsung V-NAND. Very cool technology, if can do everything Samsung claims it can.
So, perhaps Dr. Freeman won’t have to come to the rescue after all . . .