SAN FRANCISCO—Engineers from Samsung Electronics Co. Ltd., Toshiba Corp. and SanDisk Corp. took to the podium to provide details of their respective 19-nm NAND flash chips in presentations at the International Solid-State Circuit Conference (ISSCC) here Wednesday (Feb. 22).
Samsung's Daeyeal Lee delivered a paper describing the company's 64-Gbit multi-level cell (MLC) NAND device implemented in sub-20-nm technology. The device features a 533-megabit-per-second DDR interface achieved by implementing a wave-pipeline architecture, Lee said. The chip also makes use of new techniques to overcome floating-gate coupling interference and mitigate program disturbance, Lee said.
According to Lee, the use of the new schemes, known as correction-before-recoupling reprogram and P3 pattern pre-pulse scheme, results in a 21 percent lower bit error rate compared with conventional techniques. Lee described another technique, inhibit-channel-coupling-reduction, for mitigating program disturbance in the chip.
A bit of controversy erupted at the end of Lee's presentation, when one audience member, who identified himself as a member of conference's memory subcommittee, chastised Lee for declining to describe the size of memory cell size of the chip, suggesting that Samsung was not providing enough information for an ISSCC paper.
Earlier Wednesday, Toshiba's Noboru Shibata described his company's 19-nm multi-level 64-Gbit NAND device. Shibata said the chip's die size is 112.8 square millimeters, the smallest ever reported for a NAND flash device.
According to Shibata's presentation, the Toshiba chip, which was developed under its long-standing collaboration with SanDisk, achieves an industry first with 15-megabyte-per-second programming throughput. The device employs a one-sided, all-bitline architecture and single-array configuration, he reported. The chip also employes a high-speed toggle-mode interface, he reported.
Asked to comment on the write cycle endurance of the device—an increasing concern as chip makers shrink the feature sizes of NAND—Shibata said it was equivalent to Toshiba's 24-nm NAND. Shibata's presentation described the implementation of a new memory cell programming algorithm meant to mitigate program disturbances.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.