TORONTO – With 3D NAND unlikely to make economic sense until at least 2015, SanDisk and its flash foundry partner Toshiba both recently announced 15nm process technologies to produce NAND flash.
SanDisk’s 1Z-nm technology will be applied to both 2-bit-per-cell and 3-bit-per-cell NAND flash memory architectures with production ramp to begin in the second half of 2014. The 15nm technology scales chips along both axes, and will be used across a broad range of SanDisk offerings, from removable cards to enterprise SSDs.
Toshiba’s new process replaces its 19nm process technology, and is aimed at providing a transitional step to 3D NAND, said Scott Nelson, senior VP of Toshiba America Electronic Components’ memory business unit. Toshiba’s 15nm process works in conjunction with improved peripheral circuitry technology to create chips that achieve the same write speed as chips formed with second generation 19nm process technology, but boost the data transfer rate to 533 megabits a second -- 1.3 times faster -- by employing a high-speed interface.
Nelson said there is room to advance floating gates before moving to 3D NAND. “3D is a new technology and to really bring it to market means you have to have a cost-effective solution to the current floating-gate solution,” he said. “We felt it would take another year or so to get to that point. That’s why we chose to continue the floating-gate path.”
Most suppliers are on the same vector, Nelson added. 3D NAND won’t make sense until 2015 or 2016. Toshiba is now applying the 15nm process technology to 3-bit-per-cell chips, and aims to start mass production in June 2014. The company will develop controllers for embedded NAND flash memory in parallel and introduce 3-bit-per-cell products for smartphones and tablets, and it will subsequently extend it to notebook PCs by developing a controller compliant with SSDs.
The 19nm process technology won’t immediately disappear, however. Nelson said the enterprise market segment, which is rapidly adopting SSDs, will require legacy support, so both processes will run in parallel for the foreseeable future. Generally, he said, new process geometries are being announced every 12 to 16 months. “Floating gate could continue to shrink; the question is do the characteristics of that further reduced geometry meet the requirements of the market,” he said. “Floating gate is nearing the end of its ability to shrink effectively. The cost benefit gets less and less.”
Moving to 15nm is all about cost, said Jim Handy, principal analyst at Objective Analysis. “Toshiba is able to get the number of gigabytes they produce per wafer up by shrinking down the process,” he said. The company is charging the same amount as other suppliers, but at a lower cost per gigabyte. “There are two ways for a vendor to become profitable. One is to raise prices if the market allows that, and the other is to lower cost.”
There are compromises, however. Handy said a finer geometry tends to lead to a decrease in speed and an increase in bit errors, but controllers are getting more intelligent to compensate. “The controller technology is moving ahead faster than the quality of NAND is declining.”
As partners, Toshiba and SanDisk have a roadmap for the next few years, and they have been be very forthright that they were going to bring out 1X, 1Y and 1Z technologies before moving to 3D NAND, said Handy. “The smaller the processors get, the more challenging they are.” He noted that Samsung announced 3D NAND production last August.
Handy said the industry has visibility to hit 12nm before it hits a brick wall, and that’s when 3D should kick in, but there’s still a learning curve ahead since 3D NAND uses processes never applied in the production of semiconductors, he said. “The transition is going to take longer than anybody expected.”