Technology Trends China is the world's largest market for servers, PCs and smart phones. The demand for memory chips is extremely large. In 2014, the proportion of storage chips purchased by China accounted for 20% of the world's total; in 2015 it increased to 30%, and the proportion in 2016 was estimated. In the world, two of the world’s top ten storage chip sourcing companies are from China, Lenovo and Huawei, ranking fourth and fifth respectively. Lenovo is the second largest PC company and the fourth largest server company in the world. Huawei is The world's largest telecommunications equipment provider and the third largest mobile phone company.

3-D architecture can overcome mobile data bottleneck between chips

The memory chip industry also has an extremely important significance for China. It is information security. The “Prism Gate” incident has made China highly concerned about information security. The current social informatization has reached an unprecedented height. In the future, with the Internet of Things, big data, etc. The issue of developing information security will become even more prominent. Since all data is stored in the memory chips, the memory chip is an important part of information security.

With embedded smart technologies entering more and more areas of life, a large amount of data is being generated from autonomous driving to personalized medicine. However, as the spread of data reaches a great extent, the ability of a computer chip to process it into useful information is stagnating.

According to foreign media reports: Recently, researchers at Stanford University and the Massachusetts Institute of Technology have established a new chip to overcome this obstacle. The research results were published in the journal Nature of Foreign Media. The main authors were Max Shulaker, assistant professor of electrical engineering and computer science at MIT, and Shulker and H.-S, who were PhD students at the university. The team also includes Professor Roger Howe and Professor Krishna Saraswat from Stanford University.

Today's computers are assembled together by different chips. Then the connection between the chip for calculation and the separate chip for data storage is limited. As applications analyze a growing amount of data, the limited rate at which data moves between different chips will create a critical communication "bottleneck." The space on the chip is limited and there is not enough space to put them aside, even if they are miniaturized (this is a phenomenon called Moore's Law).

What's worse, transistors made of silicon are no longer as good as they used to be.

The new prototype chip is the fundamental change in the chip today. It uses a variety of nanotechnology, as well as a new computer architecture to reverse these two trends.

Instead of relying on silicon-based devices, the chip uses carbon nanotubes, which are nano-cylindrical 2-D graphene sheets, and resistive random access memory (RRAM) cells, which are non-volatile memories. By changing the resistance of the solid dielectric material. Researchers integrated more than one million RRAM cells and two million carbon nanotube field-effect transistors, making the latest nanoelectronic systems an emerging nanotechnology.

The RRAM and carbon nanotubes are built perpendicular to each other, forming a new dense 3-D computer architecture with logical and memory interleaving layers. By inserting between these layers, this 3-D architecture is expected to solve the communication bottleneck.

However, based on the main author of this article, Max Shulaker (a key member of the MIT Microsystems Technology Laboratory), existing silicon-based technologies are unlikely to adopt this architecture. "The current circuit is 2-D, because the establishment of a traditional silicon transistor involves an extremely high temperature of over 1000 degrees Celsius," Shulker said. "If you then build a second silicon circuit on top of it, then high temperatures can damage the bottom layer of the circuit."

The key to this work is that the carbon nanotube circuit and RRAM memory can be manufactured at a low temperature of less than 200°C. "This means they can be stacked without harming the underlying circuitry."

This provides several benefits for future computing systems. "The device is better: Compared with logic made of silicon today, the logic made of carbon nanotubes can be an order of magnitude higher than energy efficiency. Similarly, RRAM can be denser, faster, and more energy-efficient than DRAM. It refers to a conventional memory called a dynamic random access memory.

"In addition to improved equipment, 3-D integration can solve another key problem in the system: the interconnection between chips and between chips.

"The new three-dimensional computer architecture provides dense and fine-grained integration of computing and data storage, which greatly overcomes the bottleneck of mobile data movement between chips." As a result, the chip is capable of storing large amounts of data and transforming large amounts of cluttered data into useful information. . ”

To demonstrate the potential of this technology, researchers can use carbon nanotubes as sensors. On the top layer of the chip, they placed more than 1 million carbon nanotube-based sensors for the detection and classification of environmental gases.

Due to the layering of sensing, data storage, and computation, the chip can measure each sensor in parallel and then write directly to its memory, creating a huge bandwidth.

Original Title New 3-D Chip Combines Computing and Data Storage