LONDON Ė By 2016 magnetic RAM and phase-change memory will be a billion-dollar annual business, according to market research firm Yole Developpement (Lyon, France).
Five application areas will fuel market growth for four emerging nonvolatile memory technologies MRAM, PCM, resistive RAM (ReRAM) and ferroelectric RAM (FeRAM), the firm reckons, although the emerging technologies will continue to be much smaller than the standard memory types of DRAM and NAND flash memory which had combined revenues of more than $50 billion in 2012.
Overall, the global emerging non-volatile memory market will grow from a value of $209 million in 2012 to $2 billion in 2018, equating to a compound annual growth rate of 46 percent for the period. MRAM and PCM will lead the emerging NVM market ahead of ReRAM, reaching a combined annual value of $1.6 billion in 2018, Yole forecasts
Yole asserts that enterprise storage is set to provide the largest opportunity for emerging non-volatile memory, with both spin-torque transfer (STT) MRAM and PCM being used for cache memory in front of NAND flash.
At the same time mass storage, currently served by NAND flash memory could begin to start using 3-D ReRAM in about 2017 or 2018. This is the time when the scalability of 3-D NAND with data words arranged vertically through the silicon stack is predicted to slow down. When this happens, ReRAM is expected to ramp up rapidly to replace NAND.
Mobile phones will adopt PCM as a replacement for NOR flash memory in multichip packages thanks to the 1-Gbit chips made available by Micron in 2012 and higher density PCM chips are expected to arrive in 2015, Yole said.
Meanwhile STT MRAM is set to replace SRAM in system-on-chip applications due to the technologies lower power consumption and superior scalability.
Yole also expects microcontrollers to adopt some form of MRAM and PCM as a replacement for embedded flash.
The more mature FeRAM technology will grow at a CAGR of 10 percent with a focus on industrial and transportation applications because only relatively low-capacity chips are available.
Click on image to enlarge.
Annual market for emerging nonvolatile memories market for applications including cache memory for enterprise storage, industrial & transportation, mass storage, mobile phones, smartcard MCU. Source: Yole Developpement.
They are fine technologies; the materials they use are very particular, must be processed and operated in particular ways. There is therefore a natural division between companies who have already invested in these materials (so need to continue their development and in some cases already offer products) and those who haven't (so will be reluctant).
The company I work for builds industrial equipment with a long life cycle. Our products are expected to remain in operation for 10 years or more. Our producta are also specified to operate over the industrial temperature range (-45C to +85C)
How do MRAM, PCM and ReRAM compare for replacing Flash this type of use? We've noticed that as the processes shrink, Flash is becoming less reliable. In addition, most vendors appear to be chasing the commercial market with higher densities and lower reliability.
Have there been any comparisons done on these technologies?
The growth of the non-volatile memory market was expected because the MRAM nonvolatile memory technologies provide better results and are cheaper. My nephew had the opportunity to try the new ferroelectric RAM`s and he was very pleased with them, he even called the Innovia CMC specialists to ask if it is worthy to invest in this.
It is also interesting that Yole bracketed PCM and MRAM together to indicate a market size they think the two technologies together could reach but saying nothing about each one.
By 2016 it could be MRAM 100 percent and PCM 0% or vice versa.
Almost certainly the absolute number will br wrong and the perentages somewhere between the two extremes I've outlined.
PCM was easier to sell against the field-driven MRAM that Everspin or Crocus made, since it could scale better. But now it could have a more difficult time against the new STT type MRAM. But a big difference is the MRAM resistance is becoming negligible or at least smaller compared to the transistor itself, which can be 20 K-ohm or more. Any resistance-based memory therefore cannot have a minimum resistance less than say 30 K-ohm, and equivalently, maximum current cannot exceed say 40 uA. The first condition is harder for MRAM, the second harder for PCM.