New Circuit Integration Technology
“Mobile Supercomputer Fostering Innovation in the Information Society”
The end of Moore's Law
Integrated Circuit (IC) was invented to solve the interconnection problem of a large computer system. As an example, approximately 5 million solder connections were required in building the ENIAC computer. What integrated circuits do is to connect tens of billions of circuit elements with interconnects totaling several kilometers in length. However, with Moore's Law approaching its limits, we can no longer solely rely on integrating increasingly more functions onto a chip for performance improvement. We need to find a revolutionary solution to the interconnection problem.
An integration technology using near-field coupling
Our proposal is to replace mechanical connections between chips and between circuit boards, such as those using solder joints and connectors, with contactless connections using near-field coupling. In the preceding CREST project, we independently developed the ThruChip Interface (TCI) technology, which is a wireless interchip communication technology based on magnetic coupling, as well as the Transmission Line Coupler (TLC) contactless interconnect technology based on electromagnetic coupling, and successfully reduced interchip data transmission power by a factor of 1000, and demonstrated wireless connection between circuit boards.
Mobile supercomputer with world's best power efficiency of 100GFLOPS/W
The goal of our research is to demonstrate the feasibility of building a mobile supercomputer using TCI for 3D integration of system LSI and memory chips, and TLC to wirelessly connect components such as storage and sensors to the compute module, to achieve a power efficiency of 100GFLOPS/W which is 60 times better than the K supercomputer. To that end, we will implement a stacked DRAM interface with a data rate of 512GB/s, which is 30 times that of a conventional interface.
Fostering innovation in the information society
It is also our goal to reveal how applying our mobile supercomputer to emerging applications such as automated driving and big data analysis will change people's life, and the economic impact that can be achieved. Furthermore, we plan to build a new kind of artificial intelligence computer equipped with both a left brain and a right brain to enable us to explore a new paradigm of information processing.
Professor, Electrical Engineering
Instruction video(about 1m30s)
Message from the Program Manager
Greetings from the Program Manager of the Japan Science and Technology Agency's (JST's) ACCEL Kuroda project!
The opportunity of this project came when I met Professor Kuroda accidentally in the gloomy meeting room during a panel discussion at IEDM held in San Francisco in December, 2014. I told him in passing that JST has recently started a new “Shakai-jisso” project named ACCEL (Shakai-jisso is a Japanese term referring to the implementation and application of technology to solve societal problems). I did not fully comprehend what kind of role PM (Program Manager) plays sufficiently at that time, nor did I anticipate that Professor Kuroda's research proposal would be nominated for ACCEL. Eight months later in August, 2015 research activities of ACCEL formally began, and I could not help but feel that it was my destiny to be involved, given how incredible it was that a few trivial words would lead to such a big thing or a grand project.
I have so far followed a lot of pioneering research project including CREST and PRESTO which had a long way until practical implementation despite having created great technology. Nevertheless, with Professor Kuroda’s research proposal, there is great expectation that we may go to all the way to “Shakai-jisso”. Indeed, through my conversation with people from the industry, I have become fully aware of how highly regarded this technology should be. Here in this project, I would like to carry out the business matching by constantly collecting input from the industry as well as grasping industry needs accurately.
However, research and development always come together with the risk. As Program Manager, it should also be my responsibility to minimize the overall risk of the project. On the other hand, the feature of this field is that speed of the development competition is very fast. Furthermore, we will need to formulate a well-thought-out IP strategy for the core technology as well as create portfolios for our patents.
I recall Professor Nishi from Stanford University telling me that, while a lot of professors in Japan watch out only for Japanese Government in order to get a government-sponsored, Professor Kuroda always pays more attention to the industry when he does his work. This left a strong impression on me, and I feel that I have been given a great opportunity to make this apt remark something visible.
For a project in a long stretch in nearly five years, I would like to support it mainly on the program management to help accelerate Professor Kuroda's research to develop a high efficiency computer to fuel the advance in applications such as automated driving and robotics as a foundation of future smart society.
I appreciate your support and cooperation for this project.