Toshiba recently announced that it will begin demolishing its No. 2 semiconductor fabrication facility, the company’s NAND Flash memory plant, and replace it with a new fab facility on the same site. Why? According to Toshiba: “The primary purpose of the new wafer fab is to secure space for converting existing Toshiba and SanDisk 2D NAND capacity to 3D NAND beginning in 2016.” Toshiba also announced a deal with SanDisk to jointly invest in the new facility.
To help understand the implications of this announcement for the storage industry and data center storage users, a quick review is in order. Modern data center storage systems now offer the use of solid state storage devices, either in some combination with spinning disk or as “all Flash” arrays. The solid state storage modules that go into them are commonly based on what is referred to as 2D flash technology where memory cells within the chip are spread-out on a two dimensional grid. With 3D NAND flash, memory cells stacked on top of each other.
3D NAND promises to offer significantly better performance, a 50% increase in storage capacity on a per chip basis, and greater device longevity. Engineers and manufacturers now see an end to the 2D roadmap. But while it appears obvious that the transition to 3D NAND in the data center is bound to happen over the next two-three years, getting there presents significant challenges to manufacturers.
The move to volume 3D NAND Flash storage module production will be anything but straight forward. Currently, four manufacturers have announced their intent to build 3D NAND. But each is building them in different ways using different combinations of technologies. Samsung appears to have achieved first mover status with the announcement of vertical NAND (V-NAND) production in August of last year using existing fab facilities. However, V-NAND samples apparently have yet to be delivered to vendors that want to test them.
Now in partnership with fencing announces the construction of a new “supplementary facility for processes mainly dedicated to people using a different manufacturing technology. And they’ve determined that a current production fab isn’t going to get them there requiring a major investment in a new facility. Other 3D NAND players who have yet to make their production and availability intentions clear are Micron in partnership with Intel, and SK Hynix.
Aside from the ability to produce quality 3D NAND technology in volume, perhaps the bigger issue is price. Samsung bet that they could get to the 3D market with cost competitive product using existing facilities. Toshiba/SanDisk have now come to the opposite conclusion. Only time will tell who’s on the right and most cost-competitive road map.
Regardless of underlying technology and road map issues, it is now clear that 3D NAND Flash will begin to replace its 2D NAND Flash predecessors starting next year with volume and increasingly attractive pricing in 2016. The rush to all-Flash storage arrays in the enterprise data center has already begun. As one data center administrator commented: “The all-Flash array makes even badly written apps look good.” They foresee a day in the not too distant future when all primary data center storage is solid state. Data center administrators who are now installing current generation all-Flash storage can now at least begin to plan for the transition to 3D. Serious investments in it, like the one just announced by SanDisk/Toshiba, are now being made by producers.