Intel and Numonyx are today announcing a key breakthrough in the development of Phase Change Memory (PCM) technology.
The breakthrough has enabled the partnership to develop a 64Mb test chip that demonstrates multiple layers of PCM arrays on a single die. These findings will enable future PCM-based devices to offer lower power consumption, higher capacity and greater storage density.
The company is calling the breakthrough chip a PCMS, or Phase Change Memory and Switch. The PCMS interleaves layers of thin-film PCM arrays with controlling thin-film selectors called Ovonic Transfer Switches (OTS). The sandwiched PCM and OTS layers are arranged in a crosspoint architecture and fitted to a CMOS substrate to create high-density, high-bandwidth PCM cells.
What is phase change memory?
Phase change memory is an interesting technology that relies on the the curious properties of chalcogenide glass, which changes phases of matter with the application of voltage.
Early attempts at PCM development defined two states of matter: The first being an amorphous solid, the second a hard crystalline structure. In the case of the former, the high resistance of the amorphous state indicates a binary zero, and the low-resistance crystalline state represents a binary one. More recent developments have discovered two additional states, bringing the storage density up to four states, or in line with the density primarily offered by MLC NAND cells.
Why phase change memory?
PCM’s characteristics make it something of a Holy Grail amongst memory researchers. For starters, the current crop of NAND-based storage devices represent their data by trapping electrons. Scientists believe that it will not be possible to store electrons in NAND cells manufactured on a process node any smaller than 20nm. PCM, however, is seen as a viable technology down to the 5nm level.
PCM also ends the write endurance debate which continues to shroud NAND technology in a shadow of doubt. Because PCM uses the electrical resistance of matter states to represent binary data, it offers an unlimited number of write cycles.
“This is important as traditional flash memory technologies face certain physical limits and reliability issues, yet demand for memory continues to rise in everything from mobile phones to data centers,” said Greg Atwood, Senior Technology Fellow at Numonyx.
This differs from NAND which uses charges of up to 20V to read, write or erase data; these charges ultimately destroy a NAND cell’s ability to read or write new information.
Phase change may even unify mass and volatile storage technologies by replacing DRAM as well. Al Fazio, Intel Fellow and Director of Memory Technology Development, explains.
“You have a memory technology that looks like memory–in other words, hardware can do a load/store because it can act on a small chunk of data with a low latency, yet it’s non-volatile so that it has the non-volatile aspects of storage.”
With a write latency of 1µs, PCM is 20x faster than DRAM.
This means that memory modules based on PCM could operate at the speed of DRAM, but retain their data even through power loss. Imagine how quickly a PC might boot if the operating system’s last environment state was ready to go in memory the moment the PC was powered on.
Because PCM cells are byte-addressable, rather than grouped into 512KB blocks as with NAND, PCM-based storage solutions would offer the consistent read/write performance we see in today’s desktop memory. No more optimal write sizes, no more write amplification, and no more trading between IOPS and throughput.
Looking ahead
Phase change memory is very much an infant product. Vendors like Intel, Samsung, and Numonyx are working feverishly to further improve capacity and speed before the technology can go into production. We must also consider their efforts in the context of PCM’s glacial development, which thus far spans over forty years. It can easily be surmised, then, that a foundry configured to produce PCM cells falls on the far side of five years out.
Even so, the significance of the Intel/Numonyx PCMS should not be underestimated. A stackable, high-density, high-throughput PCM chip is an outstanding achievement in the ongoing efforts to realize a commercial product.
In the mean time, Fazio and Atwood are preparing their research for the International Electron Devices Meeting (IEDM) in December where it will be presented by Intel Senior Principal Engineer DerChang Kau.
RSS Feeds