IBM’s new line of FlashSystem storage appliances will be making waves soon as they are slated to come using one of the fastest & most hard-wearing, nonvolatile memory technologies in existence; the Magnetoresistive RAM (MRAM). The new architecture which will be used for cache writing will be replacing the capacitor-backed DRAM. The flip-side is this, however, when the MRAM is compared to Intel’s 3D XPoint memory & NAND flash memory, it is seen to be seriously lacking in terms of density.
The single supplier of discrete MRAM chips, Everspin, has moved on from the in-house manufacturing on wafers fabbed at Freescale to a partnership with GlobalFoundries in other to make MRAM on their 22nm FD-SOI process. It has also been raising the memory capacities of their MRAM chips & by year’s end, we’ll be having a taste of their 256Mb & 1Gb chips. Though they still stand too small to be used as primary storage, the increase in MRAM memory capacities has made it alluring for use in systems with a high volume of data. It is a good step forward actually if we consider the fact that previous applications of the MRAM were only used in embedded systems to replace battery-backed DRAM, small flash chips & SRAM buffers.
IBM’s new SSDs for its Flashsystem are some of the highest-density TLC-based drives, as they pack a storage capacity of 19.2TB of 64L 3D TLC NAND flash memory & use a 20-channel NAND interface with a four-lane PCIe 4.0 host interface which can operate in dual-port 2+2 mode. IBM’s current FlashSystem appliances make use of the standard 2.5″ U.2 drive form factor SSDs, a system-wide power loss protection design at a per-drive level, FIPS 140 compliant encryption & FPGA-based controllers with optional 3:1 transparent compression ratio with no inclusion of bulky supercaps as MRAM’s non-volatility eliminates the need for it.
IBM’s new FlashSystem drives will be displayed this week at the Flash Memory Summit & shipping of the drives begin this very month.