Non-Volatile DIMMs and NVMe Spice Up The Flash Memory Summit

Sept. 10, 2012
Technology Editor Bill Wong was at the Flash Memory Summit and delivers some feedback on what is new in flash storage especially for the embedded and enterprise arenas.

Viking Technology's ArxCis-NV (Fig. 1) was just one of the flash technologies I saw at this year's Flash Memory Summit in Santa Clara, CA. You can check out my Flash Memory Summit interviews on Engineering TV. This includes technology like Skyera's flash memory-based, network storage that packs up to 44 Tbytes of consumer grade flash into a 1U rack system (watch Skyera Delivers 44 Tbytes Of Flash Via iSCSI - Flash Memory Summit 2012 ). Another hot topic was NVM Express (NVMe). More on that later.

First I want to comment on Viking's non-volatile DRAM, the ArxCis-NV. I talked about it earlier (see The Fundamentals Of Flash Memory Storage) but the latest version is more standard and higher capacity. I'll be able to talk abou the standard aspect in a bit more detail after the Intel Developer Forum 2012 where I am headed this week. There will be motherboards specifically designed for the DIMMs that can handle the off-DIMM supercap that is normally wired directly to the DIMM. This simplifies installation and replacement especially when multiple DIMMs are involved which tends to be the case for servers.

Figure 1. Viking Technology's ArxCis-NV packs DDR3 DRAM and flash storage plus a controller on a DIMM. The latter can save and restore the contents of the DRAM via the on-board flash.

Another company that is using this approach is Agiga Tech. Its Agigaram is available for SDRAM, DDR2 and DDR3 sockets.

Just in case you have not read the other articles I wrote on the topic, the way these non-volatile DIMMs work is by having an equal amount of flash a DRAM on the DIMM along with a controller. The latter can detect a power failure and copy the contenst of the DRAM to flash. The process is reversed when the system boots. It provides the functionality of core memory but using the latest technology. It also does so using available semiconductors rather than using technologies like FRAM or MRAM that have yet to match the capacities of DRAM and flash.

The approach could taken on a global basis which would be similar to hibernation that is used on laptops. The difference is that the DIMMs rely only on a supercap to provide power when a power failure is detected. Also, the DIMMs operate independently so the shutdown process is proceeds in parallel when multiple DIMMs are involved. Overall, it is a more reliable scheme. It is also one that lends itself to embedded applications, not just enterprise installations where the products are targeted. The operating system and boot system need to be modified to take advantage of this technology.

The hot topic at the show was NVMe. The University of New Hampshire InterOperability Laboratory is already set up to test products as they are released (watch University of New Hampshire InterOperability Laboratory Tackles NVMe - Flash Memory Summit 2012"). They use PCI Express test equipment from Teledyne LeCroy that has been tuned for NVMe.

IDT was showing off their NVMe controller chips (see Controllers Speed NVM Express Drive Delivery). They have 16- and 32-channel controllers that sit between the PCIe connection and the flash memory chips.

I'll be writing more about flash technology that I saw at the show. It is the 25th anniversary of NAND flash memory. It looks like flash will continue to be a central technology in all areas from consumer to the enterprise. The remains a myriad of products with a wide range of characteristics, pricing and target markets.


To join the conversation, and become an exclusive member of Electronic Design, create an account today!