Life after flash memory
Non-volatile flash memory has played a large role in enabling the performance we see in today’s microprocessors and computers. To store data, flash memory relies on controlling electrons stored in a transistor’s gate circuit. Flash provides attractive read-write speeds with reasonable power consumption.
So far, flash memory has kept pace with continual shrinkage of chip fabrication processes characterized by Moore’s Law that predicts doubling of chip performance roughly every 18 months. While Moore’s Law appears intact for the foreseeable future, experts see flash and random-access memory (RAM) technologies reaching scale limitations in a similar timeframe. Forward-thinking companies and developers are anticipating this eventuality.
A new non-volatile technology—phase-change memory (PCM), also known as PRAM (phase-change RAM)—is viewed as the most promising among alternatives to flash. PCM stores data by altering the chip material’s atomic structure, obtaining improved data density and other benefits over standard flash.
Under R&D scrutiny for years, PCM has taken a major step forward. In early February 2008, Intel Corp. and Geneva, Switzerland-based STMicroelectronics announced the start of prototype silicon shipments using PCM technology to customers for evaluation. Codenamed “Alverstone,” these PCM prototypes are 128 Mb devices, fabricated on a 90-nanometer (nm) process. Intel refers to this product sampling milestone as “bringing [PCM] technology one step closer to adoption.”
PCM uses an electric pulse to alter the device material’s physical state at the microscopic level. Until recently, PCM designs implemented only two phase states — amorphous , with atoms loosely arranged and the chip material at high electric resistance, and crystalline , with atoms rigidly arranged and at lower material resistance. The resistance differential and ability to switch quickly between phases translates to data bit values 0 and 1.
Recent R&D work has found two more PCM material states able to store information: semi-amorphous and semi-crystalline . This essentially doubles memory capacity.
Other benefits claimed for PCM include dramatically faster read-write speed than standard flash memory and 50% power savings. A further advantage is PCM’s ability to change data at a single-bit level, much like DRAM. To change one data bit with flash can mean erasing a data block with thousands of bits, slowing data write cycles and increasing device wear.
Phase-change memory developments are accelerating. Only a few years ago, Intel and STMicroelectronics demonstrated 8 Mb memory arrays at 180 nm process node—as part of a joint development program for PCM technology begun in 2003. Further compelling research announced in February 2008 cited the two companies as developers of a new high-density, multi-level cell (MLC), large memory device using PCM technology. This advance from single-bit per cell to higher density MLC capability has significant impact for lowering cost of memory. Others, such as Hitachi, IBM, and Samsung, also are active in PCM development.
To implement Alverstone and other new memory products, STMicroelectronics, Intel, and investment firm Francisco Partners, have agreed to establish an independent semiconductor company named Numonyx. As of this writing, the transaction is slated to close in 1Q08.
Samples now available
Intel and STMicroelectronics haven’t set an introduction date for PCM products. Flash memory developments and improvements continue — by Intel, for one — and both technologies are bound to coexist for a long time. However, the Alverstone device now offered for sampling represents a learning opportunity for customers about an emerging technology. It provides designers time to evaluate and plan for PCM in future systems.
It’s not a question of if, but how soon, phase-change memory will enhance a gamut of new consumer and automation products, especially embedded systems.
|Frank J. Bartos, P.E., is a Control Engineering consulting editor. Reach him at email@example.com .|