Serial NAND Flash for Code Storage in Artificial Intelligence Systems

12 oct. 2020

SPI NOR Flash is currently the preferred memory type for storing boot and application code in embedded systems, but it’s struggling to provide the higher memory capacities needed to accommodate the larger code bases of new AI applications. Discover the benefits, including reliability, longevity and performance, of replacing SPI NOR Flash with serial NAND Flash in this article adopted from Winbond Electronic’s Syed S. Hussain, Director of Segment Marketing.

As today’s software-rich embedded systems demand larger code bases, and require regular updates in the field, the Program/Erase performance of SPI NOR Flash is put in the spotlight. Applications such as machine learning generate complex software code, resulting in required code storage of 1Gbit or more — a size at which the cost-per-bit of SPI NOR Flash becomes unattractive when compared to serial NAND Flash.

To provide a migration path from SPI NOR to serial NAND Flash for code storage in AI applications, Winbond has refined its serial NAND fabrication process and the serial interface to offer:

- Fast Read performance

- Fast Program/Erase performance

- Like-for-like compatibility with SPI NOR

- High signal integrity and long data retention

The Winbond QspiNAND (Quad SPI NAND) Flash is fabricated with Single-Level Cell (SLC) Flash produced on a 46nm process. The reliability and quality of this older process has been proven over many years of operation in the field. The larger circuit features of the 46nm process also provide ample headroom to allow for electron leakage without compromising data retention: Winbond QspiNAND Flash is specified for data retention of 100,000 hours over the rated operating-temperature range.

Learn more about this innovative storage solution in the full article by Winbond.

DOWNLOAD ARTICLE

Related Products

Shop All 2Gb Qspi NAND W25N02JW

Shop All 1Gb Qspi NAND W25N01JW