What's New - 2015

 

New Efficient Cache Memory Directory for Multi-Core Processors - Current trends are increasing the number N of processor cores, but overhead associated with shared memories increases sharply as N increases. A group at MIT has developed a more efficient directory design to relieve this problem.

 

IBM Research Topics - Eight key research areas are identified: 7 nm fab, quantum computing, neurosynaptic computing, Si photonics, III-V integration, C nanotubes, graphene, and tunnel FETs.

 

90 GHz photonic emitters - Duke University researchers demonstrate nanopatch plasmonic antennas and quantum dots that can transmit on-chip signals at 90 GHz and above with high efficiency.

 

Neuromorphic Computer Compares to Mouse Brain - IBM research demonstrates prototype neuromorphic computing system with 48 million electronic neurons, based on True North low-power chips, for running deep learning algorithms.

 

President Obama establishes National Strategic Computing Initiative - New inter-agency US government program to promote future high performance computing.

 

Biggest Neural Network Ever Pushes AI Deep Learning - World's largest neural net computer demonstrates improved machine learning of natural language processing.

 

IBM and SUNY Demonstrate 7nm transistors - Researchers at Albany Nanotech Complex use extreme UV lithography and SiGe channels.

 

UCSD Group Demonstrates Memcomputer - A memcomputer is a novel computer architecture whereby logic and memory are integrated together in the same devices. Prof. Di Ventra and colleagues demonstrate a proof-of-principle prototype.

 

Flash Memory Can Improve Efficiency of Big Data Computations - MIT researchers show that servers with large amounts of flash memory can match the performance of systems with faster DRAM, at lower power and cost, for certain database calculations.

 

Japan Preps for HPC-Big-Data Convergence - Prof. Satoshi Matsuoka of Tokyo Institute of Technology focuses on key role of data-centric computing in future supercomputers.

 

AMD Focuses on Energy Efficiency in New Processor Chips - Projects 25x improvement in efficiency by 2020 by reducing idle power and intelligent power management.

 

Google Announces New I/O Projects - New Google projects on futuristic input/output devices, codenamed Vault, Abacus, Soli, and Jacquard

 

Laser Links for Supercomputers - High-bandwidth free-space optical links may enable improved performance for existing supercomputers.

 

HP Machine will not include Memristive Memories - 1st generation prototype will use conventional DRAM instead of promised new technology.

 

Approximate Computing for Modeling Climate - Supercomputer simulations for climate change may be redesigned for greater efficiency and speed by appropriate use of imprecision.

 

NASA Roadmap for Future Computing - Challenges include supercomputing, and cognitive, quantum, and cloud computing approaches.

 

Rebooting Computing invited lecture in Japan - RC Co-Chair and IEEE Computer Society President Tom Conte spoke on RC at meeting of Information Processing Society of Japan.

 

New Supercomputer for Argonne National Lab - Energy-efficient supercomputer with 180 petaflops performance to operate in 2018.

 

RPI Research into Future Neuromorphic Supercomputer - Rensselaer Polytechnic Institute is conducting simulations of a novel hybrid supercomputer that would integrate both conventional and neuromorphic processors.

 

Energy Harvesting for Electronic Devices - Three Technologies for Harvesting Ambient Energy for small sensors and consumer devices (such as nodes in the Internet of Things) are described in IEEE Spectrum.

 

Designing Nanoscale Transistors - New "Computational Nanotechnology" software tools have been developed at Purdue University.

 

Petaflops on Desktops? - DARPA is interested in developing alternative hybrid analog-digital computing paradigms that may permit much more efficient simulations of certain scientific problems.

 

Neuromorphic Computing "Roadmap" Envisions Analog Path to Simulating Human Brain - Georgia Tech researchers are developing a silicon-based field-programmable analog array with distributed logic and memory, and apply it to simulating neural systems.

 

Computer plans a December 2015 special issue on Rebooting Computing - View the Call for Articles

 

Prototype nanoscale high-density 3D Chip - Research at Stanford integrates memory and logic

 

When will we have an Exascale Supercomputer? - High cost and high electrical power are reaching limits.

 

Will Tomorrow's Supercomputers Be Superconducting? - New IARPA program on Josephson logic and cryogenic memory.

 

What's Next for Computers After Moore's Law - Experts ponder this question at latest IEEE Rebooting Computing Summit

 

Faster Random Algorithm for Analyzing Big Data - Recent research at Georgia Tech has proven a new more efficient random algorithm