numonyx – was that PCM or PCRAM?

On March 31st the Intel/ STMicroelectronics joint venture Numonyx emerged as an independent company. While the first priority of Numonyx will be the inherited Flash businesses its phase change memory efforts did receive plenty of editorial space.

The various articles were a bit fuzzy around what phase change memory will and will not be able to do. Statements such as “more robust” and that it may yield a “billion dollar savings” as an SRAM replacement were set in print. At least the ghost of Universal Memory was generally not disturbed.

As noted in an earlier article PCRAM faces two important challenges to which considerable research is directed. First, the basic data storage mechanism consumes considerable power when toggling between phases i.e. during a write, for example. This power requirement has always been and is likely still a focus of research, evidenced by the considerable number of patents around structures, materials and processes for reducing the programming current.  In passing it is worth recounting that the memory cell must be heated to around 600C to invoke the necessary phase change.

Second, the write endurance of PCRAM is somewhere around 100 million writes.  This is a large difference from the theoretical limitless endurance of a phase change memory cell.  It is also 9 orders of magnitude below the endurance required for both SRAM and DRAM applications.  While the reason behind this lower endurance is unclear, it is likely associated with the volume change that occurs when changing from one phase to another.  At the moment, high endurance NVM or RAM applications seem out of reach.

Finally, when looking at the Numonyx website and reading the press the acronym PCRAM is not used. They simply refer to Phase Change Memory (PCM). The obvious difference is the loss of the “Random Access”. This is a subtle change in message to position PCM away from the RAM (DRAM and SRAM) replacement promises to focus on its Flash replacement opportunities, which will no doubt be large.