RRAM technology has reached the stage of commercialization
At a time when the first 3D NAND chips are only looking for their way to the market and many NAND manufacturers are still improving their designs, several promising technologies are being developed to create the next generation of memory that plan to replace NAND flash memory in the next 10 years.
One of the most promising technologies - RRAM (Resistive Random Access Memory) - is similar to NAND, is non-volatile, that is, stores data without constant power supply. This is its fundamental difference from DRAM, which requires a constant source of energy. Several companies are engaged in development in the field of RRAM, including the giants of the semiconductor industry like Samsung and SanDisk, but now it is the American startup Crossbar that has the most advanced design.
Crossbar , headquartered in Santa Clara, California, was founded in 2010 and received more than $ 50 million in funding. The roots of the company grow their University of Michigan, because the main researcher and co-founder of prof. Wei Lu is currently an assistant professor at this university. At the moment, the Crossbar team consists of 40-45 people, most of whom have extensive knowledge in the areas of semiconductor research and development.
')
The main advantages of RRAM over NAND are performance and durability. Typically, the NAND delay in reading is on the order of hundreds of microseconds, while the Crossbar achieved in their RRAM design from 50 nanoseconds. The cell durability can reach millions of rewrite / erase cycles, although in early prototypes the Crossbar focuses on more conservative ~ 100 thousand cycles.
Last week at IEDM 2014 Crossbar announced that they are moving to the stage of commercialization of RRAM. In other words, they have already demonstrated a working prototype in silicon, and also proved that the design can be transferred to commercial factories for mass production, so the company is now working with manufacturers to create the final product.
Initially, Crossbar focuses on the embedded market and licenses its technology to developers of ASIC, FPGA and SoC. The appearance of the first samples is expected in early 2015, and mass production by the end of the year or early 2016. In addition to licensing, Crossbar is also engaged in the development of individual high-capacity and density chips, which should enter the market about a year after RRAM embedded solutions (approximately in 2017).
The attractiveness of RRAM is that this kind of memory can be produced through the normal CMOS process with only a few modifications. NAND and especially 3D NAND require specialized expensive tools, which is why only a few companies can produce 3D NAND. RRAM, in turn, can be produced in almost any semiconductor factory after inexpensive refurbishment, which will result in lower prices and greater market competition.
In addition, the lithographic problems of NAND should not affect RRAM. As you know, the main reason for the invention of 3D NAND was that flat NAND does not scale very well to processes below 15 nm, which leads to a decrease in performance and cell life. RRAM can be reduced to 4-5 nm without the above problems, moreover, Crossbar has already shown a prototype 8-nm chip, which they created in their R & D laboratory. In addition, RRAM can be stacked vertically up to three layers to increase the density; by the time of the commercial release of individual chips, the company plans to increase the number of layers to 16, and the density of chips to 1 Tbit.
Of course, before RRAM technology is ready to fully compete with NAND, several obstacles need to be overcome, but it is very pleasant to hear that there has been considerable progress in development and that production companies are interested in technology.
Faster, more durable and affordable SSDs and other storage devices benefit everyone. It should be recognized that 3D NAND is only a temporary solution until the arrival of a worthy replacement, which RRAM is likely to be.
One of the most promising technologies - RRAM (Resistive Random Access Memory) - is similar to NAND, is non-volatile, that is, stores data without constant power supply. This is its fundamental difference from DRAM, which requires a constant source of energy. Several companies are engaged in development in the field of RRAM, including the giants of the semiconductor industry like Samsung and SanDisk, but now it is the American startup Crossbar that has the most advanced design.
Crossbar , headquartered in Santa Clara, California, was founded in 2010 and received more than $ 50 million in funding. The roots of the company grow their University of Michigan, because the main researcher and co-founder of prof. Wei Lu is currently an assistant professor at this university. At the moment, the Crossbar team consists of 40-45 people, most of whom have extensive knowledge in the areas of semiconductor research and development.')
The main advantages of RRAM over NAND are performance and durability. Typically, the NAND delay in reading is on the order of hundreds of microseconds, while the Crossbar achieved in their RRAM design from 50 nanoseconds. The cell durability can reach millions of rewrite / erase cycles, although in early prototypes the Crossbar focuses on more conservative ~ 100 thousand cycles.
Last week at IEDM 2014 Crossbar announced that they are moving to the stage of commercialization of RRAM. In other words, they have already demonstrated a working prototype in silicon, and also proved that the design can be transferred to commercial factories for mass production, so the company is now working with manufacturers to create the final product.
Initially, Crossbar focuses on the embedded market and licenses its technology to developers of ASIC, FPGA and SoC. The appearance of the first samples is expected in early 2015, and mass production by the end of the year or early 2016. In addition to licensing, Crossbar is also engaged in the development of individual high-capacity and density chips, which should enter the market about a year after RRAM embedded solutions (approximately in 2017).
The attractiveness of RRAM is that this kind of memory can be produced through the normal CMOS process with only a few modifications. NAND and especially 3D NAND require specialized expensive tools, which is why only a few companies can produce 3D NAND. RRAM, in turn, can be produced in almost any semiconductor factory after inexpensive refurbishment, which will result in lower prices and greater market competition.
In addition, the lithographic problems of NAND should not affect RRAM. As you know, the main reason for the invention of 3D NAND was that flat NAND does not scale very well to processes below 15 nm, which leads to a decrease in performance and cell life. RRAM can be reduced to 4-5 nm without the above problems, moreover, Crossbar has already shown a prototype 8-nm chip, which they created in their R & D laboratory. In addition, RRAM can be stacked vertically up to three layers to increase the density; by the time of the commercial release of individual chips, the company plans to increase the number of layers to 16, and the density of chips to 1 Tbit.
Of course, before RRAM technology is ready to fully compete with NAND, several obstacles need to be overcome, but it is very pleasant to hear that there has been considerable progress in development and that production companies are interested in technology.
Faster, more durable and affordable SSDs and other storage devices benefit everyone. It should be recognized that 3D NAND is only a temporary solution until the arrival of a worthy replacement, which RRAM is likely to be.
Source: https://habr.com/ru/post/364787/
All Articles