This nonvolatile technology is as fast as DRAM, uses same tools and processes as other memory.
The memory market is going in several different directions at once. On one front, the traditional memory types, such DRAM and flash, remain the workhorse technologies in systems despite undergoing some changes in the business. Then, several vendors are readying the next-generation memory types in the market.
As part of an ongoing series, Semiconductor Engineering will explore where the new and traditional memory technologies are heading. In this segment, Greg Schmergel, chairman and chief executive of Nantero, sat down with Semiconductor Engineering to discuss memory trends. Nantero is a developer of carbon nanotube RAMs, called NRAMs. What follows are excerpts of that conversation.
SE: What are the problems with existing memory?
Schmergel: Every customer would like to have fast nonvolatile memory. They have been living for decades with the choice between either fast DRAM or nonvolatile and slow flash with limited endurance. Ideally, they would like a memory that is as fast as DRAM and nonvolatile like flash, or at least something in between, close to DRAM performance.
SE: What do NRAMs bring to the party?
Schmergel: Our memory answers that need. It is as fast as DRAM, while being nonvolatile. We do have unlimited endurance as well. And that has led to a lot of rethinking of the memory hierarchy and how you design systems and devices. It has led to us partnering with a number of the world’s largest memory users, both on the standalone and embedded side.
SE: We’ve been hearing about the new memory types for years. Suppliers have made some bold promises, but most have failed to deliver. Thoughts?
Schmergel: There has always been skepticism. But the openness for new options has gone up dramatically. The industry has really started to think about what these new technologies could do for their products.
SE: What is a carbon nanotube RAM and how does it work?
Schmergel: NRAM is nanotube random access memory. It’s based on carbon nanotubes, which are either in contact with each other or not in contact with each other to form high resistive and low resistive states. So you have very distinct ‘0s’ and ‘1s’. This is because there is a large difference in resistance between the ‘on’ and ‘off’ state. The carbon nanotubes move from position to position in picoseconds. So it’s a very fast memory. It requires low energy to write. And it doesn’t wear out.
SE: Last year, Fujitsu Semiconductor and Mie Fujitsu Semiconductor licensed Nantero’s NRAM technology. The companies also announced a joint development deal. Mie Fujitsu, a foundry venture between Japan’s Fujitsu and Taiwan’s UMC, plan to make NRAM devices for customers. Are NRAMs shipping yet?
Schmergel: They are not shipping yet. But it’s in an advanced stage of development. We have about a dozen customers. One customer, Fujitsu, has already announced product availability in 2018 for an embedded memory product offering. They plan to make standalone memory as well. We are working with other customers on other products as well.
SE: Carbon nanotubes are carbon-based cylindrical structures. They not only have electrical properties, but they are also strong materials. Are NRAMs based on traditional carbon nanotubes?
Schmergel: It’s a standard carbon nanotube. There are some differences. When you grow carbon nanotubes, you grow them from metal nanoparticles as a catalyst. They are generally grown in dirty environments. So they come mixed in with a lot of contaminants, often a high percentage by weight. And that, of course, is not fab compatible. For example, if have you several percent of iron mixed in with your carbon, it’s not allowed in any production fab in the world. So one of the things that we’ve done is develop a way to purify the carbon nanotube material to where it is production CMOS fab compatible. It meets the contamination standards of any production fab in the world. In fact, we’ve installed and used our carbon nanotubes for multiple production fabs in the world already.
SE: For years, we’ve heard about carbon nanotube FETs for logic. Are carbon nanotube RAMs the same thing?
Schmergel: This is a different application. We have a different carbon nanotube material. For FETs, they require separating the carbon nanotubes into a semiconducting metallic, which is a major challenge. We don’t need to do that for memory.
SE: Do you need special equipment to make NRAMs in the fab?
Schmergel: We’ve installed NRAM in multiple production fabs. It not only uses the same tools and processes, but can also run on the existing production lines that currently run DRAM or flash, or in a logic foundry. With NRAM, there are no additional capital expenditures. It can be easily spin-coated on wafers using existing tools, and patterned using existing lithography and etch tools. This will enable manufacturers to quickly and cost-effectively migrate to a new generation of memory that is both high-speed and nonvolatile.
SE: What are some of the challenges to bring NRAM to the market? What about funding?
Schmergel: At the end of last year, we announced a $21 million round in funding, which included participation from some of our key customers. Now, the challenge is to complete the product design, go through the qualification process, ramp it up to production and make yield improvements. This is what all new chips have to do. But we are really past the innovation phase and moving into the engineering phase, where we still have a lot of hard engineering work to do.
SE: Where can you use NRAMs in the market?
Schmergel: It could work as a storage-class memory. It could work in enterprise systems, servers or enterprise storage. It could work in a mobile device by either replacing some of the DRAM or supplementing the DRAM. You can achieve a substantial increase in battery life. The first applications will be in both as a standalone storage-class memory and as an embedded memory.
SE: Does NRAM replace DRAM and/or NAND?
Schmergel: Initially, NRAM could be a DRAM replacement or working in-between DRAM and NAND. Eventually, it could be a NAND replacement as well. However, it takes a few more years to get the cost down to NAND levels.
SE: What are the densities and geometries for NRAMs?
Schmergel: Our focus is on 2xnm processes. Of course, we are doing research for sub-20nm for product development. Fujitsu has announced the first products with us, which will be 55nm with 40nm to follow. We are at the later stages of product designs for a standalone multi-gigabit product, which will be DDR4-compatible.
SE: What about the embedded market?
Schmergel: We have partners/customers in all of the areas, such as microcontrollers, SSDs, analog semiconductors and a variety of areas. If you compare NRAMs to embedded flash, it’s cheaper and faster with higher endurance. The problem with embedded flash is that it’s having trouble scaling down to 28nm. It works very well at higher nodes. Cost is also a big issue. If you add embedded flash, it increases the cost of the chip because of all of the mask layers. And there are limitations in terms of speed and endurance.
SE: Can you use NRAM in industrial apps?
Schmergel: For automotive and industrial, NRAM is very well suited, because it is resistant to temperature variations. It is resistant to radiation.
SE: How do you compare NRAM with the other next-generation memory types?
Schmergel: Some of them are positioned as potential competitors to flash. That requires them to achieve very low cost to compete with standalone flash. The only one out there that is as fast as DRAM is MRAM. But MRAM is substantially more expensive and also harder to scale.
SE: Do you compete with the 3D XPoint technology from Intel and Micron?
Schmergel: Not really. It’s in a different area of the market with different applications.
SE: Will the industry ever see a so-called universal memory? A universal memory is a single technology that could replace today’s memory, such as DRAM, flash and SRAM.
Schmergel: In the next five years, it’s very unlikely that we will have a universal memory. For the long term, mass storage will still need the lowest cost option, which is NAND flash right now. But for high performance, you will eventually need a different memory.