Growing demand for more bits at lower cost is pushing flash manufacturers to build upwards
2 mins read
Technology has always been the central element of the memory business, with process technology helping to meet demand for more bits per unit area for less.
But burgeoning demand for flash memory, said by some analysts to be growing at a compound annual rate of 45%, is outstripping the ability of memory developers to use Moore's Law to meet the needs. The conclusion drawn by leading memory manufacturers is that it's time to use the Z direction.
Samsung is the first company to announce that it has a 3d product in manufacture. The 128Gbit NAND device uses Samsung's V-NAND technology, in which vertical cell structures are linked using a proprietary interconnect. Up to 24 layers can be stacked, with special etching technology connecting the layers.
Jeong-Hyuk Choi, senior vice president, flash product and technology, said: "The new 3d V-NAND technology is the result of years of effort to push beyond conventional ways of thinking." He added more 3d V-NAND products will be launched.
According to Samsung, V-NAND at least doubles reliability, while doubling write performance compared to a 10nm class floating gate NAND flash memory.
Not far behind in the race is Micron, which recently unveiled a 128Gbit multilevel NAND memory manufactured on a 16nm process. It says the 16nm node is not only the leading flash process, but also the most advanced processing node for any sampling semiconductor.
Glen Hawk, pictured, head of Micron's NAND solutions business, said even though the move to 16nm brings denser memories which cost less per bit, customers continue to demand more memory at a faster rate than the industry can scale. "We have to work out how to do this more quickly," he commented.
He sees flash being used 'everywhere'. "The early implementation was in the USB drive, but now it's commonplace and demand for flash is driving the memory industry." Two markets are leading the demand for flash: tablets; and solid state drives. "It's still early in the SSD lifecycle," Hawk continued, "and the focus is on density and cost. The market started to take off last year and the cost is already as low as $1 per Gbyte."
Like Samsung, Micron is looking to build upwards, rather than meet density demands by shrinking planar circuits. Not only does 3d enable much larger flash devices, it also brings technical and economic benefits. Memory companies are already thinning memory dice and stacking them in order to get high densities. "We may be taking eight dice and putting them in a package to offer 1Tbit of storage," said Hawk. "But stacking is twice the cost of a 3d memory."
He said it takes about 300 steps to make a NAND flash memory. "With 3d only a few of those have to be replicated, so 3d represents a small incremental cost. But the equipment required for 3d is very different to planar."
Then there's the benefit of being able to use larger memory cells. "With multilevel cells, the number of electrons that define state is around 20," Hawk pointed out. "3d takes us back to the point where we can deal with tens of thousands of electrons."
"When stacking vertically," Hawk noted, "you can relax the XY dimensions, so you don't have to use a 128Gbit stack; you could go back to a 64 or a 32Gbit layer and it wouldn't take many layers to get the same density. When you make the XY dimensions larger, the process becomes more manageable."
But Micron isn't yet ready to put 3d into production. "We're focusing on cost," he claimed. "New fabs are expensive and the cost of converting existing facilities is rising. 3d can be made with two, three or four tiers – it's not that difficult. The issue is that you need dozens of tiers to get to the cost crossover point. But it will happen," he concluded.