SSDs are seemingly doomed. Why? Because as circuitry of NAND flash-based SSDs shrinks, densities increase. But that also means issues relating to read and write latency and data errors will increase as well.
"This makes the future of SSDs cloudy," states Laura Grupp, a graduate student at the University of California, San Diego. "While the growing capacity of SSDs and high IOP rates will make them attractive for many applications, the reduction in performance that is necessary to increase capacity while keeping costs in check may make it difficult for SSDs to scale as a viable technology for some applications."
To prove this theory, Grupp teamed up with Steven Swanson, director of UCSD's Non-Volatile Systems Laboratory, and John Davis of Microsoft Research. Using PCIe-based flash cards with a channel speed of 400 MBps based on the Open NAND Flash Interface (ONFI) specification and a standard 96 NAND flash dies, they tested 45 different NAND flash chips from six different vendors that ranged in size from 72-nm to 25-nm.
The group discovered that write speed for pages in a flash block suffered "dramatic and predictable variations" in latency. Even more, the tests showed that as the NAND flash wore out, error rates varied widely between devices. Single-level cell NAND produced the best test results whereas multi-level cell and triple-level cell NAND produced less than spectacular results.
With the testing information at hand, the group fast-forwarded to the year 2024 when NAND flash circuitry is expected to be only 6.5-nm in size. They predicted that read/write latency will double in MLC flash and increase more than 2.5 times for TLC flash. Yet SSDs at that time are expected to feature capacities of 4 TB when using MLC flash, and 16 TB when using TLC flash.
"It's not going to be viable to go past 6.5-nm," Grupp said. "2024 is the end. [People] are used to working with technology that continues to just get better, but with NAND flash we're going to be facing trade-offs as it evolves."
To read the full report, head here.