Phase-change memory (PCM) is an emerging type of non-volatile random access memory. PCM chips are currently under development for enterprise-class data storage and chips in production for mobile phones today. We spoke with Haris Pozidis, a leading PCM researcher, to learn more about the technology and where it may fit in storage environments. Pozidis manages the Memory and Probe Technologies group at IBM's Zurich laboratory, which focuses on non-volatile memory concepts. He holds more than 35 U.S. and European patents in the areas of solid-state memory technology, probe-based data storage, control systems technology and optical data storage.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
IBM is performing research around racetrack memory and memristor technologies. How do you see phase-change memory stacking up against those technologies? Which is the closest to reality for enterprise storage/server environments?
Haris Pozidis: Indeed, IBM is doing research on racetrack memory and memristor technologies. It is generally believed that those technologies, while potentially interesting, are further away from commercialization, and thus exploitation, in IBM servers than PCM.
Enterprise-class phase-change memory on the way, but roadblocks remain
PCM has already seen 10-12 years of intense research, and recently advanced development in the labs of most memory manufacturers, [while] these alternative technologies are mostly still in the basic research stage.
In terms of characteristics, memristors are thought to be an eventual replacement for NAND flash memory due to their potentially better scalability and performance. However, their limited capability for multi-level cell storage might limit their price competitiveness in cost per [gigabyte]. Racetrack memory, on the other hand, has the potential to serve as universal memory, as it promises both fast switching and extremely high capacity. However, fundamental issues need to be resolved to realize the high-capacity promise of racetrack memory, having to do with the physics of magnetic domain wall movement.
When does IBM expect to go to market with phase-change memory products for enterprise storage?
Pozidis: On June 10th this year, IBM and SK Hynix announced a joint development agreement for the development of PCM chips with multi-level cell capability and high performance. These chips are targeted, among others, for high performance enterprise storage/server applications.
I am not at liberty to reveal the actual timeline. In general, it is believed in the IT industry that 'storage-class memory' products, such as PCM chips, would start appearing in computing systems on the 2015-2016 timeframe.
Do you expect that phase-change memory will replace flash or just be another option?
Pozidis: We expect that PCM, with its universal characteristics, will bridge the latency and endurance gap between fast/volatile/expensive main-memory -- mostly DRAM today -- and slow/non-volatile/cheap storage -- in the form of NAND flash today.
New applications and workloads will be able to utilize and benefit from a larger, non-volatile memory technology sitting alongside today's DRAM, and other applications will benefit from a faster, higher endurance and byte-alterable memory that would act as cache for today's NAND flash, in the form of a solid-state drive (SSD).