3D RRAM Chip Combines Data Storage and Computing

[Guest_Jim_*]

The rate of innovation is always increasing as new creations lead to more new projects, but this progress is not uniform. In some instances the ability to produce something can come to outpace the ability to use it, and this is happening with data currently. Right now we have the ability to generate more data than many systems can efficiently handle, but there are many researchers working to change that, including some at MIT and Stanford University where a very advanced chip has been created that combines data storage and processing.

This new chip combines two technologies that can still be considered futuristic for computers; carbon nanotubes and a 3D architecture. Modern computer chips have a 2D design, though some have a 2.5D design with layers stacked and connected to each other. The benefit to a full 3D design is that the multiple parts of the chip are able to communicate with each other much more quickly and efficiently than what is currently possible. Carbon nanotubes can also take this to a new level as their small size and electrical properties allow the chips to be made denser. In this case the chip is a form of resistive random-access memory (RRAM), which is a kind of nonvolatile memory and has some one million RRAM cells and two million carbon nanotube field-effect transistors. This combination of memory and computing removes the bandwidth bottleneck between data and processing that is an issue with today's largest datasets.

To prove the capabilities of this design, the researchers also added over one million nanotube-based sensors for detecting and classifying gases. The measurements from the sensors were processed all in parallel and written directly to memory, thanks to this integrated design of emerging nanotechnologies. What makes this accomplishment even more impressive is that the chip is compatible with CMOS, so such an RRAM chip could be combined with current silicon chips and there is a fair chance there will be many more applications for this design in the future.

Source: MIT

Back to original news post