Revolutionizing Computing Architecture with MemComputing

An Introduction to a New Paradigm in Scientific Computing

Since the 1940s, many modern computers have followed the standard model of the Von Neuman architecture, involving a separate processor (CPU) and memory (RAM). A major issue with this structure is a problem called the Von Neumann bottleneck, in which data exchange between processors is limited due to the separation of those two components, thus affecting overall efficiency and performance of the system.


A new sort of computing architecture opts to solve this problem; MemComputing, spearheaded by Ph.D. physicists Fabio Traversa and Max Di Ventra at the University of California San Diego, combines the workings of storing memory and processing information to determine a solution in a single computing cycle. John Beane, Chief Executive Officer and co-founder of MemComputing, Inc., gave us an insight into the technology and its applications in a presentation to the Society of HPC Professionals.


How exactly does MemComputing interchange processing and memory to optimize performance? Beane’s presentation explained the technology’s patented self-organizing logic gates (SOLGs). Unlike traditional logic gates, SOLGs allow input and output at all terminals. Current Logic Gates, have specific input terminals and an output terminal where the logical result is output. Instead, SOLGs evaluate the values at the traditional input and output terminals. If the values do not satisfy the logical expression, then the SOLGs negotiate with the other SOLGs around them to satisfy the logical solution. As the SOLGs satisfy the logical expectations, they stop negotiating. Equilibrium, or a solution, is reached once all SOLGs are logically satisfied, e.g., equilibrium is reached and the output or solution will then be put into main memory. Essentially, the solution is reached in a single “MemComputing” cycle (microseconds).


With the success of this computing technology, MemComputing has also released a cloud-based application builder named the Virtual MemComputing Machine (VMM). By using the new MemCPU architecture, the VMM is able to solve combinatorial optimization problems and computations in a timely and cost-efficient manner never seen before.


The MemComputing technology has seen its use in several commercial applications, with operations involving transportation logistics, as well as oil, gas, and energy markets associated with various Fortune 100 companies. Beane introduced the Society of HPC Professionals to a few case studies, in which the VMM has constructed ground-breaking solutions towards helicopter fleet scheduling, ship-borne cargo optimization and a satellite design tool used to identify airborne targets for the U.S. Department of Defense. They also shared recent work on something known as multi-agent path finding.  Multi-agent path finding is used for optimizing everything from agents in multi-player games, warehouse robots, and drone swarms. MemComputing, Inc.’s technology clearly represents a new computing paradigm that has many applications and has the potential to revolutionize computing technology,





Isabelle Sitchon

Source Name

HPE Data Science

News Category
UH Data Science News
Research Topics