The Changing Face Of Non-Volatile Storage

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

The Flash Memory Summit was host to a range of new technology, from LSI’s SandForce flash memory controllers utilizing a multi-level low-density parity check (LDPC) to the Hybrid Memory Cube (see “Hybrid Memory Cube Shows New Direction For High Performance Storage” at electronicdesign.com). The need for storage is insatiable. Three additional technologies look to address the issue: Samsung’s 3D flash architecture, Crossbar’s resistive RAM, and Diablo Technologies’ TerraDIMM (see “Large-Scale Flash Moves Next To The Microprocessor” at electronicdesign.com).

3D Vertical Flash

Samsung’s 3D V-NAND flash architecture increases capacity by layering 24 cells on top of each other (see the figure). The technology improves reliability by a factor of two to 10 while delivering twice the write performance of conventional 10-nm-class floating-gate NAND flash memory.

Samsung is delivering 128 Gbits in a single chip using V-NAND’s 3D charge trap flash (CTF). The CTF architecture temporarily stores an electric charge in a non-conductive layer composed of silicon nitride (SiN) instead of a more conventional floating gate like most NAND technologies. A floating gate has more interference between cells as smaller geometries bring cells closer together.

Samsung uses a special etching technology to create the multilayer system. It punches holes through from the highest to the lowest layer of the chip. This is very difficult, but Samsung can now deliver chips on a regular basis.

Resistive RAM Is Non-Volatile

Crossbar delivers non-volatile storage using a new resistive RAM (RRAM) technology that is compatible with existing CMOS technology. It beats NAND flash with performance increased by a factor of 20. It also cuts power requirements by a factor of 20. It has an efficient 4F2 cell layout as well. RRAM is comparable in performance and endurance with other non-volatile technologies like MRAM (see “Magnetic DRAM Arrives” at electronicdesign.com) and PCM (phase change memory), but with a lower power requirement.

Crossbar’s announcement was about the technology versus chip delivery. It should scale well and even supports 3D stacking, but MRAM and PCM have yet to make a dent in flash except in niche markets. All three could give NAND flash a lot more competition in the future.

DIMM Flash

Diablo Technologies takes advantage of conventional MLC NAND flash but packs it onto a standard dual-inline memory module (DIMM). The TerraDIMM also includes the company’s secret sauce in the form of a memory controller that handles the NAND flash and the DDR3 protocol. A system can have a single TerraDIMM or multiple TerraDIMMs, and that’s key to its performance edge. The TerraDIMM already has an edge by using the fastest interface to the processor, the memory controller. High-end processors sport multiple controllers, and multichip processor clusters share memory from all the controllers.

The TerraDIMM differs from Viking Technology’s non-volatile DRAM, the ArxCis-NV (see “The Fundamentals Of Flash Memory Storage” at electronicdesign.com). The ArxCis-NV mirrors conventional double-data-rate (DDR) memory in NAND flash between power cycles. The TerraDIMM works like conventional flash, so device drivers make this NAND look like a solid-state disk (SSD) to the operating system. The system can do more, but that requires operating-system changes. These changes may show up in the near future, offering interesting application advantages.

Processor improvements will have to take a backseat to storage this year. These technologies are significant advances, and they are just the tip of the non-volatile iceberg. We may have more and faster cores, but new storage architectures will change how storage is managed and utilized.

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

Please or Register to post comments.

What's alt.embedded?

Blogs focusing on embedded, software and systems

Contributors

William Wong

Bill Wong covers Digital, Embedded, Systems and Software topics at Electronic Design. He writes a number of columns, including Lab Bench and alt.embedded, plus Bill's Workbench hands-on column....
Commentaries and Blogs
Guest Blogs
Nov 11, 2014
blog

How to Outsource Your Project to Failure 3

This article will address failure to carefully vet a potential manufacturing or “turnkey” partner and/or failure to transfer sufficient information and requirements to such a partner, a very common problem I have seen again and again with my clients over the years, and have been the shoulder cried upon by several relatives and clients in the past....More
Nov 11, 2014
blog

Transition from the Academe to the Industry Unraveled 1

There have been many arguments here and there about how short-comings of universities and colleges yield engineers with skill sets that do not cater to the demands of the industry. There have been many arguments here and there about an imminent shortage of engineers lacking knowledge in the sciences. There have been many arguments here and there about how the experience and know-how of engineers in the industry may vanish due to the fact that they can’t be passed on because the academic curriculum deviates from it....More
Nov 11, 2014
blog

Small Beginnings 5

About 10 years ago I received a phone call from an acquaintance. He had found a new opportunity selling some sort of investments and he wanted to share it with me in case I was interested. Ken had done fairly well for many years as a contract software developer primarily in the financial services sector. His specialty was writing RPG code. (RPG is often referred to as a write only language.) But he was seeing the handwriting on the wall as the industry moved on to other methods, and saw himself becoming a fossil....More

Sponsored Introduction Continue on to (or wait seconds) ×