ReRAM to replace NAND flash?
The development of alternatives to NAND flash is accelerating with players such as Weebit Nano, which can count on the resources of CEA-LETI to finalize the production of its ReRAM components intended to start on on-board systems.
The semiconductor market has always been the scene of more or less spectacular innovations. While NAND flash memory – invented in 1989 by Toshiba – has won all the votes in recent years with its MLC, SLC or QLC developments, players such as HPE (Memristor), IBM (PCM), Intel or Micron (3DxPoint), without forgetting the MRAM (magneto-resistive RAM) of Toshiba were determined to challenge this supremacy with proposals faster and more energy efficient than NAND flash memory and RAM. Another player, the Israeli start-up Weebit Nano, is in turn trying to lay the groundwork for a new computer standard in the field of memory with its ReRAM (ResistiveRAM) technology based on silicon oxide (SiOx). .
Founded in 2014 by Yossi Keret (now the head of Nanorobotics), Weebit Nano has been piloted since 2017 by Coby Hanoch who explained to us, during a virtual press point organized as part of an IT Press Tour Israel, how she intends require less expensive, faster, more reliable and more energy efficient memory than existing Flash technology. Developed from the work of Professor James Tour of the University of Rice University, in Houston, Texas, and with the expertise of CEA-LETI, a research center specializing in nanotechnologies in Grenoble, Weebit’s ReRAM solution is based on 7 worldwide patents.
One of the main problems with Weebit Nano, according to its CEO, was moving from labs to production “by finding the right fab that doesn’t leave us at the last manufacturing line.” It is necessary to bring money to mobilize the production lines. To successfully cross this delicate course, we looked at the different options and chose to work with LETI. It is a great partner for us. They have a lot of memory experience and they are very flexible, which is important for a start-up. They contribute in an important way to our success with the contribution of the memory selector ”. Weebit Nano has therefore signed an R&D agreement with LETI in order to develop prototypes of its ReRAM SIOx. The manufacture of ReRAm should begin at the end of 2020. “The idea is to arrive at any of the founders on the market and bring them all the technical specifications so that they can produce the memories. We are at the point of commercializing our ReRAM components and we are already measuring the market response on these key points. “
With its ReRAM components, Weetbit first of all intends to establish itself in the embedded systems market (IoT and automotive).
Attacking the auto market
This resistive random access memory uses nanometallic filaments which are charged or not to preserve bits of information. Unlike NAND flash, whose mechanism resembles opening or closing a door, the filaments break and are rewoven to retain information for a longer time. But today, Weebit Nano’s positioning is not to attack NAND flash head-on, but rather to break into that of persistent memory components embedded in SoC-type chips and systems embedded in IoT or the automotive industry. . ReRAm SSDs capable of really competing with NAND flash-based models will come later. “We believe that the embedded systems market is less exposed to launch our activity,” says the CEO. “It’s an easier market to start with and we can show very resistant examples, able to keep data intact for 10 years at a temperature of 150 °. Which is very important for the automotive market ”.
And to take off, Weebit Nano is counting on the Chinese market where contacts have already been made with semiconductor maker XTX Technology, formerly known as Paragon Technology. “We have demonstrated the effectiveness of our silicon oxide ReRAM memory technology at the XTX plant,” said the manager. Asked about competition from Intel with its Optane components, Yossi Derek indicates that they are not in the same market. “Intel is focusing on the data center market with Optane, they are not present on the embedded systems market. The applications are different, we are not yet ready to attack the data center market. It’s just another opportunity for us, but today our memory is a good solution for on-board systems. My main focus is to show that we can generate income in a very short period of time and the embedded systems market presents less risk ”. At the end of the year, Webit Nano should be able to assemble its first complete memory modules and discuss business with its prospects. Let us point out to conclude that Weebit Nano is not the only one interested in ReRAm, Intel has also studied the question and presented ReRAM components at the ISSCC 2019. The start-up Crossbar is also betting on this technology.