Electronicdesign 9578 Promo 0

Embedded Flash Storage Gets Smaller and Faster

Jan. 14, 2017
Flash storage has changed the way embedded designers can utilize large amounts of storage in a compact system.
Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.
1. I used Virtium’s StorFly compact SATA SSD with NVidia’s Jetson TX1.

Embedded developers have never had so many options when it comes to non-volatile, industrial-grade storage. Even the number of form factors is mind-boggling, ranging from conventional disk-drive formats—e.g., the now-ubiquitous 2.5-in. drives—to chips as well as modules like M.2.

Flash memory is the dominant storage mechanism, but alternatives like MRAM, FRAM, RRAM, and so on are growing in availability, performance, and capacity. Very ambitious developers might work directly with storage chips, but most developers will prefer to deal with storage that provides standard interfaces from SPI to PCI Express (PCIe), as well as variants like PCIe-based NVMe.

Even conventional interfaces and form factors like SATA-based, 2.5-in. solid-state drives are morphing into new configurations that work nicely with motherboards such as Virtium’s StorFly SATA modules (Fig. 1). This works nicely in the vertical orientation for a development board, though a production board would use a horizontal orientation that allows the module to be bolted down.

SATA, USB, and PCI Express tend to be the common interfaces for storage modules. I2C and SPI—including its variants, such as quad SPI (QSPI)—tend to address chip-based storage, although this is also changing in concert with rising capacity and performance demands. For example, Cypress Semiconductor’s serial HyperBus family runs at 333 MB/s using 12 pins.

2. Samsung’s NVMe chip (right) provides the same high-speed storage interface available on high-performance 2.5-in. solid-state disks (left) and M.2 modules (center).

Even storage chips are no longer limited to conventional parallel and serial interfaces. Samsung has NVMe-based chips that support NVMe (Fig. 2). The 512-GB PM971-NVMe utilizes Samsung’s 48-layer V-NAND, 4 Gb of LPDDR4 mobile memory, and Samsung’s controller. These are packed into a 20- by 16- by 1.5-mm BGA that weighs only one gram. 

Developers need to take a number of factors into account—not just physical size or memory capacity.  Performance is typically on the checklist, but application and data lifetimes matter. The latter can be more or less important depending on the technology employed. It is critical for flash storage, but less of an issue with technologies like MRAM.

Available interfaces often dictate what types of storage options are available to a developer. Even the DRAM sockets on many motherboards are candidates for non-volatile memory. Diablo Technologies has technology that allows all flash-memory DIMMs to plug into these sockets. These options are also changing how operating systems and applications deal with non-volatile storage. Looking at this storage in terms of block device interfaces is now only one option. Persistent memory is something that’s becoming more common.

Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.

Sponsored Recommendations


To join the conversation, and become an exclusive member of Electronic Design, create an account today!