Volume shipments of 3D NAND flash could arrive in 2016, if Toshiba’s and SanDisk’s plans bear fruit. Fab 5 at their Yokkaichi campus is expected to install a 3D pilot line in 2H15. The project is funded jointly with SanDisk.
Samsung started sampling 128Gb, 24 layer, 3D flash last year, but their use of a 40nm process loses much of the density advantages 3D should offer over leading-edge 15nm production today. 128Gb 3LC flash is available today.
Micron has held off on sampling their 3D flash for, they say, competitive reasons. But they claim to be very satisfied with their technology.
Samsung, Toshiba and Micron all seem to be targeting the systems market, SSDs and the like, rather than consumer devices. Due to the extra manufacturing steps, 3D flash will be more expensive per bit.
Applying 3D flash in SSDs isn’t a drop-in replacement either. The extra density of dies may also require new controllers for smarter data layout and better protection against multiple die failures. Die failures seem to be a common problem, not surprising due to the 20v writes flash requires.
The StorageMojo take
Toshiba’s move reflects a caution about 3D that is warranted. Larger SSDs are attractive, but not higher costs.
Flash is the major cost component of an SSD, so how will costlier flash will make 3D SSDs more attractive? Since there’s unlikely to be a performance boost – it’s the same old flash – capacity is the key differentiator.
So the question for inquiring minds is: what applications fit the profile of a larger, not faster, and more costly SSD. Imagine a 6TB SSD in 2016. Who buys it, and why?
Courteous comments welcome, of course.
Hi Robin, I suspect you will see high capacity SSDs (4 TB) this year and 8 TB as early as Q1 2015, maybe even sooner. In a 3.5″ form factor, flash has already eclipsed HDD’s and I expect the lead may widen in the near future. There is a clear path to 16 TB SSDs in a 3.5″ form factor within the next 12 months, compared to 8-10 TB for HDD’s in that form factor.
The challenge becomes getting enough chip enables to support such a high capacity behind a single controller. Much like the HDD world, where increasing capacity was not met with a commiserate increase in performance, high capacity SSDs will be of primary interest to “warmer” use cases where density, power efficiency, and cost per GB are paramount compared to IOps/TB.
I do not see flash ceding much to the next generation NVM technologies; if anything, these capacity gains are broadening the flash appeal and narrowing the next-gen NVM technologies to a high-end niche (more of a DRAM alternative).