TORONTO – Siva Sivaram can only speak for Western Digital, but he sees 64-layer 3D NAND as crossing the threshold to being more cost effective than planar NAND.
In a recent telephone interview with EE Times, the company's executive vice president of memory technology described 64-Layer 3D NAND as a "seminal technology” that will find its way into more than 50 of Western Digital's product lines. He anticipates half of Western Digital's bit output will be 3D NAND this calendar year, more than 75 percent of which will be 64-layer 3D NAND. "Sixty-four-layer is the first point where we truly think it's cost competitive with 2D NAND.”
The company recently announced what it called the world's first client SSDs built with its 64-layer 3D NAND technology. The WD Blue 3D NAND SATA SSD is aimed at DIY enthusiasts, resellers and system builders, while the SanDisk Ultra 3D SSD is targeted at gaming and creative enthusiasts who want to improve their PCs. Both will be available in capacities are large as 2TB. Sivaram said it's the high-end consumer devices that will see the most uptake of 64-layer 3D NAND first. As it expands downstream, it will also find its way into mobile, embedded and then the enterprise. "Enterprise takes the longest time to qualify,” Sivaram said.
Western Digital sees its 64-layer 3D NAND finding its way into a wide range of products starting with high performance consumer devices such as SSDs for gaming and graphics intenstive PCs.
For Western Digital, 64-layer 3D NAND is now cheaper than its best 2D NAND. "Not everyone has got there yet, but everyone is getting there pretty quickly,” Sivaram said.Western Digital did do some limited production of 48-layer 3D NAND for the retail segment, but it wasn't a full-fledged node for the company, he said, and 32-layer was only done internally. "It was a necessary learning step," he said.
He said 15nm 2D NAND was "extremely aggressive” lithography. "But when you go to 3D NAND, the pitches are pretty relaxed," he said. "It's not quite as aggressive as the 15nm.”
However, 3D NAND is expensive in terms of the new tools that are needed as well as the manufacturing space. Sivaram said 64-layer is the crossover where they are getting more bits to justify the increased cost. "But even with 64 layer, three bits per cell is essential,” he said. "Three bits per cell when built right is cheaper than any 2D NAND.”
Sivaram said 64-layer 3D NAND will dominate over the next 18 months, noting that Western Digital has been sampling since this time last year. He wasn't in a position to discuss what the uncertain future of its relationship with Toshiba will mean for its 3D NAND production, nor would he talk specifics of what the company's roadmap is past 64 layers.
The next step for 3D NAND will likely be string stacking, said Jim Handy, principal analyst with Objective Analysis. This involves stacking individual 3D NAND devices on top of each other, whether it's 32, 48 or 64 layers to create a stack of 96 or 128 layers. "In theory, it could take multiples of 64 layers,” he said. "The industry doesn't know far that can go.”
As the number of layers goes up, so do the aspect ratios, said Handy, making it harder to etch a hole accurately. "There are other problems, too, but the aspect ratio is a real bugaboo,” Sivaram said.
String stacking allows for a series of holes with aspect ratios of 60 instead of having to make a single hole with 120 or 240. "What everybody is wondering about is when it will be economic enough to go from 3D NAND to some alternate technology.”
It's been nearly a decade since SanDisk said ReRAM would replace NAND after three generations, added Handy. "But now you look and everybody is already shipping their third generation 3D NAND and sampling fourth generation. 3D NAND is probably going to stick around longer than anyone originally expected.”
Objective Analysis' comparision of a 16nm 128Gb TLC planar NAND and the 32-layer TLC 384Gb 3D NAND that assumes perfect yield and efficient manufacturing.
He said the layer count has also increased faster than anyone would have expected. "It was fueled by the fact the low layer count parts were not able to meet their cost goals. The whole point of 3D NAND was that it was supposed to be cheaper than planar, and the lower layer count devices where not.” Handy said Western Digital's and Micron's aggressive ramping suggest both have found a path to profitability for their 3D NAND; it's not clear if Samsung is doing as well as Micron or Western Digital as it uses a different technology for its 3D NAND and doesn't reveal enough information to make an assessment.
Handy said the forecast for 3D NAND to truly meet its cost and yield goals has been the middle of 2018. "We've been saying that for a couple of years. 3D NAND is still not realizing its cost goals and it's still not yielding well enough that the fabs are producing as many gigabytes as they were designed to do,” he said. "It should be a lot more profitable than it is.”
In the meantime, the ultimate ownership of Toshiba's 3D NAND business won't change the landscape too much, aside from who Western Digital is sourcing from. "A year ago, it would have been a SanDisk / Toshiba development,” he said. "The names have changed but the players have remained the same.” His short take is that whatever happens, Western Digital will still crank out flash based on the SanDisk / Toshiba technology. "It's a question of who gets the collective profits.”
—Gary Hilson is a general contributing editor with a focus on memory and flash technologies for EE Times.