Crossbar Inc. specializes in non-volatile 3D Resistive RAM (RRAM), which promises greater endurance and higher density than NAND flash. The Santa Clara, California-based startup claims that it reached an important milestone toward commercializing terabyte-scale memory arrays that can use a chip the size of a postage stamp.
Crossbar announced today its demonstration of pre-production 1 MB arrays that use its patented 1TnR – which stands for 1 Transistor driving n Resistive memory cells – for read/write operations. Crossbar said a single transistor was able to drive over 2,000 memory cells at low power to produce exceptionally dense solid-state storage.
Yet, although Crossbar’s RRAM carries the potential to achieve higher density and greater endurance at lower power than NAND flash, it’s an open question if or when the technology could reach mass production and turn up in shipping products.
Jim Handy, chief analyst at Objective Analysis in Los Gatos, California, said RRAM is a good candidate to replace NAND flash once NAND costs stop scaling, likely after 3D NAND runs its course. He said every NAND flash maker is looking at RRAM as the heir apparent, and Crossbar has a shot at the market.
“In NAND, it’s all about cost,” Handy noted. “Even though RRAM has significantly better specs – like endurance, over-write, speed and power – in the end, the lowest-cost solution always dominates the market.”
Handy said the Crossbar technology is interesting since it doesn’t require a separate select diode, as nearly every other technology does. “Select diodes have been pretty tricky so far. What we need to find out is whether it ramps into production smoothly, and it’s too early to know that yet,” he said.
Alan Niebel, founder and CEO of WebFeet Research in Monterey, California, said Crossbar’s emerging RRAM “could fall on its face if the technology hits a hurdle or no major manufacturer assists Crossbar in further developing and manufacturing it.”
But, Niebel gave Crossbar’s RRAM an 80/20 chance of seeing the light of day in shipping products, as long as the company partners with an integrated device manufacturer (IDM) or original equipment manufacturer (OEM). He speculated that Crossbar’s RRAM could reach enterprise products around 2018.
Crossbar said today that it is finalizing agreements with several leading global semiconductor companies and plans to announce its first licensing agreements shortly. Commercial shipments in enterprise products are expected in 2017, according to Sylvain Dubois, vice president of marketing and business development at Crossbar.
In the meantime, potential use cases for Crossbar’s RRAM are an interesting topic of discussion. Robin Harris, chief analyst at StorageMojo in Sedona, Arizona, said that if Crossbar gets a terabyte on a single die the size of a thumbnail, one possible destination could be a non-volatile RAM module for an in-memory database.
“What if you could have a server that had 64 TB in non-volatile main memory? You can put a very large in-memory database in there,” said Harris. “Then the implications are: ‘Gee whiz, all of a sudden I don’t need my EMC VMAX. There’s a whole bunch of storage that I probably wouldn’t need, and of course, eventually it would be a lot cheaper than buying a VMAX or some other high-performance array.’ ”
But, Harris said it’s difficult to pinpoint the best place to leverage the Crossbar RRAM technology until we know more about the performance characteristics and price. He said it’s quite possible the initial application will be low-power mobile devices such as tablets and smart phones.
Harris expressed optimism about Crossbar’s chances of success in creating an economically viable product. He said the company’s co-founder, Wei Lu, an associate professor at the University of Michigan, is an “extremely smart guy and cutting-edge researcher,” and Crossbar has top-tier venture capital backing from firms such as Kleiner Perkins Caufield & Byers. Harris added that Crossbar’s RRAM appears to be conducive to cheap manufacture at existing semiconductor fabs.
“The real test comes when the fab starts sampling chips,” said Harris. “That’s when you discover if you really have a market.”