Mass 3D NAND Production May Not Mean Mass Adoption

Development and production of 3D NAND is ramping up, but 2014 will clearly be a shakeout year.

Samsung Electronics Co. Ltd. announced in August it had begun mass production of a 128 Gbit NAND flash memory that is integrated in multiple layers, while Imec, a nanoelectronics research group, recently announced it had developed a new approach to layered flash using a technique known as laser annealing. Toshiba plans to begin 3D NAND production early next year, having announced plans to expand NAND production facilities in the summer.

Now that 2D NAND flash is reaching reliability limit, says Brady Wang, principal analyst with Gartner, memory vendors across the board are looking at how to respond to ever-growing demand for non-volatile memory, even if some aren’t going public with their plans and timelines. It won’t be until next year that Samsung’s 3D NAND, which is produced at 32 nanometers with 32 layers, will be widely available, he says. “Its cost is still higher than the fully loaded cost of conventional NAND at 16 nm process geometry.”

Developing a production process for 3D NAND that will allow costs for NAND flash to continue to fall is one of the key hurdles for 3D NAND if it is to become a viable successor to 2D NAND.

As Jim Handy, principal analyst at Objective Analysis, wrote in his ongoing blog series covering the why and how of 3D NAND, it’s not stacks of chips, although many have that misconception. A stack of chips doesn’t save money.

Ever since the integrated circuit was invented, the whole race has been to see how many transistors or how many bits of memory you can squeeze onto the same size piece of silicon,” Handy said in an interview with EE Times. “That’s how semiconductor costs have come down.

Storage of multimedia, particularly video, has created a ballooning demand for NAND flash, particularly on consumer devices, says Handy. Device makers such as Apple like to maintain price points for their products while enriching the feature set, which includes better storage capacity. “What we’ve run into with regular NAND is that it’s getting far too difficult to shrink on the normal two directions you have on silicon,” he says. “The idea of the 3D NAND is if you can’t shrink anything anymore, how do you go vertically?”

3D NAND, in the short term, would push the cost of conventional NAND downward but ultimately replace it. However, because 3D NAND is not in mass production yet, there are still unanswered questions, says Handy. “Everybody is expecting new error modes to pop up, they just don’t know what they are.”

Ultimately, he says, price is going to dictate adoption of 3D NAND; it needs to be cheaper to make than regular NAND. “The whole point in going to 3D is to lower the price.”

Wang says 3D NAND is not the only technology that might meet demand for non-volatile memory. Other emerging memory types include Memristor, which was first conceived in the early 1970s. “Memristor is a promising next-generation memory that has the potential to be faster and have lower power-consumption than current NAND flash memory,” he says. However, it will be several years before mass production is possible.

Another contender is magnetoresistive random-access memory (MRAM), which has been in development since the 1990s. Unlike conventional RAM, data in MRAM is not stored as electric charge or current flows, but by magnetic storage elements.

For now, conventional NAND components will maintain their cost advantage at the chip level through at least 2015, says Wang, as it will take time for customer understanding and controller optimization to catch up. “3D NAND will have limited application traction, with select segments within SSDs representing the greatest near-term opportunity.”