Intel and Numonyx on Wednesday announced what they categorized as a “key breakthrough” in research on phase change memory (PCM).
Their researchers demonstrated a 64Mb test chip that lets manufacturers stack multiple layers of PCM arrays within one die.
This could lead to the creation of smaller memory devices with more capacity and lower power consumption for RAM and storage.
PCM Arrays
The development is the result of joint research between Numonyx and Intel focusing on multi-layered or stacked PCM cell arrays.
“The research of our present work is based on thin film selectors that enable multi-layers of PCM to be stacked above the base CMOS, allowing for higher memory density in the future,” Al Fazio, Intel fellow and director of memory technology development, told TechNewsWorld. “The 64Mb [product] is the largest capacity memory thus far demonstrated for stacking/multi-layer.”
The researchers demonstrated a vertically integrated phase change memory and switch (PCMS) cell. This consists of one PCM element layered with an Ovonic Threshold Switch. Layering arrays of PCMS makes the chips scalable, letting chip manufacturers create denser chips that can hold more memory.
Using switches improves communications within a chip, and multicore processor vendor Tilera is using this technology to cram 100 cores into its next-generation processor.
Tilera’s processors are built on the concept of a tile, which consists of a processor core, a cache and a switch. The switch lets each tile mesh with its neighbors on all sides, creating, in effect, the Internet on a chip because each of these switches can route packets up, down, left and right, and the array is scalable.
However, PCM chips built around Intel’s and Numonyx’s development are some time away. “We used a 64Mb test chip as the vehicle to demonstrate the capability of the technology,” Numonyx spokesperson Mark Miller told TechNewsWorld. “It is not a product. Commercialization or production of this technology is some time off.”
What’s a Numonyx?
Numonyx was formed in March 2008 by Intel and STMicroelectronics, joining their flash memory businesses to create an independent flash memory company. It focuses on supplying memory solutions for consumer and industrial devices, including cellular phones, MP3 players, digital cameras, computers and other high-tech equipment.
In December, Numonyx began shipping 128 Mb samples of a PCM chip code-named “Alverston” that was jointly developed by Intel and STMicroelectronics.
The Low-Down on PCM
Phase change memory is also known as “PCM,” “PRAM,” “PCRAM,” “Ovonic Unified Memory,” “Chalcogenide RAM” and “C-RAM.” It is a type of non-volatile computer memory, which means it does not require a constant power supply to retain information.
PCM leverages the properties of chalcogenide glass, which can be switched between two states, amorphous and crystalline, with the application of heat. The crystalline and amorphous states of chalcogenide glass have very different electrical resistivity. This forms the basis for data storage.
The amorphous, high-resistance state represents a binary 0, and the crystalline, low-resistance state represents a 1. Recent versions can achieve two additional distinct states, effectively doubling the storage capacity of chips using chalcogenide glass.
While chalcogenide glass’s sensitivity to heat makes it easy to manipulate, this characteristic is also its greatest drawback, Rob Enderle, principal analyst at the Enderle group, told TechNewsWorld. “You can’t preprogram it before attaching it to a board, as the heat associated with the process would erase it,” he explained.
Chalcogenide is the same material used in re-writable optical media, such as rewritable CDs and DVDs, Enderle said. However, in those products, its optical properties are manipulated, rather than its electrical resistivity, as chalcogenide’s refractive index also changes with the state of the material, he added.
PCM vs. Flash
PCM is one of a several non-volatile memory technologies competing with flash memory. Like RAM or EEPROM, PCM is bit-alterable, Numonyx said. The bits in bit-alterable memory can be switched directly from a one to a zero or vice versa, whereas flash memory requires an extra step to make the switch.
PCM will also achieve write throughput speeds faster and with lower latency than NAND flash, which is heavily used in the iPhone. The combination of these capabilities with bit-alterable memory will provide significantly better write performance than NOR and NAND flash, Numonyx claimed.
PCM is also scalable, whereas NAND and NOR flash are not.
“PCM’s certainly faster on the write side and has about the same capacity as flash right now, but that should improve,” Enderle said. “It is also more resistant to radiation than flash, but its biggest advantage may be in its longevity.” PCM has a shelf life of about 300 years and wears out much more slowly than flash memory, so it can be reused more often, Enderle explained.
PCM technology consumes very little power, and that may spur its adoption. As manufacturers shrink chips below 90 nm, the transistors lose increasing amounts of power to leakage, which means power requirements are difficult to scaled down.
The real test for PCM will be how it works in the real world, electrical engineer Carl Howe, who’s director of anywhere consumer research at the Yankee Group, told TechNewsWorld.
“There’s always some new memory technology coming along, and you actually have to build some and try it in some new applications before you know whether it will work,” he explained. “PCM has had great capacity, but it’s very slow, and CD ROMs are a great example of where it’s used. Digital video could be the real test of this technology.”