In a paper published in the journal Nature Electronics, researchers from EPFLs School of Engineering in the Laboratory of Nanoscale Electronics and Structures (LANES) present a brand-new processor that tackles this ineffectiveness by integrating information processing and storage onto a single gadget, a so-called in-memory processor. They broke brand-new ground by developing the first in-memory processor based on a two-dimensional semiconductor product to comprise more than 1000 transistors, an essential milestone on the path to industrial production. “Today, there are ongoing efforts to combine storage and processing into a more universal in-memory processors that contain elements which work both as a memory and as a transistor,” Kis explains. “By setting the conductivity of each transistor, we can perform analog vector-matrix reproduction in a single step by applying voltages to our processor and determining the output,” describes Kis.
Unlike silicon, the most widely utilized semiconductor in todays computer processors, MoS2 forms a steady monolayer, simply 3 atoms thick, that only communicates weakly with its environments.
EPFL researchers have actually developed an energy-efficient in-memory processor using MoS2, combining over 1000 transistors. This processor, which effectively carries out vector-matrix reproduction, represents a shift away from standard von Neumann architecture and could increase the European semiconductor industry.
Developed by EPFL scientists, the very first massive in-memory processor using 2D semiconductor products might significantly cut the ICT sectors energy footprint.
As information and communication technologies (ICT) process information, they convert electricity into heat. Currently today, the global ICT ecosystems CO2 footprint rivals that of aviation. It ends up, nevertheless, that a big part of the energy taken in by computer processors does not enter into carrying out estimations. Instead, the bulk of the energy used to process data is spent shuttling bytes in between the memory to the processor.
In a paper released on November 13 in the journal Nature Electronics, scientists from EPFLs School of Engineering in the Laboratory of Nanoscale Electronics and Structures (LANES) present a brand-new processor that tackles this inefficiency by incorporating information processing and storage onto a single gadget, a so-called in-memory processor. They broke brand-new ground by developing the very first in-memory processor based on a two-dimensional semiconductor material to consist of more than 1000 transistors, a key turning point on the path to commercial production.
In a paper published in the journal Nature Electronics, scientists from EPFLs School of Engineering in the Laboratory of Nanoscale Electronics and Structures (LANES) present a brand-new processor that tackles this ineffectiveness by integrating information processing and storage onto a single device, a so-called in-memory processor. They broke brand-new ground by creating the very first in-memory processor based on a two-dimensional semiconductor product to comprise more than 1000 transistors, an essential turning point on the path to industrial production. Credit: 2023 EPFL/ Alan Herzog
Von Neumans Legacy
According to Andras Kis, who led the research study, the primary perpetrator behind the inefficiency these dayss CPUs is the universally adopted von Neumann architecture. Particularly, the physical separation of the parts utilized to carry out calculations and to store data. Since of this separation, processors need to obtain data from the memory to perform computations, which involves moving electrical charges, charging and discharging capacitors, and transmitting currents along lines– all of which dissipate energy.
Till around 20 years ago, this architecture made good sense, as different types of gadgets were required for information storage and processing. The von Neumann architecture is progressively being challenged by more efficient options. “Today, there are continuous efforts to merge storage and processing into a more universal in-memory processors which contain aspects which work both as a memory and as a transistor,” Kis explains. His laboratory has been checking out methods to accomplish this goal utilizing molybdenum disulfide (MoS2), a semiconductor material.
A New Two-Dimensional Processor Architecture
In their Nature Electronics paper, Guilherme Migliato Marega, doctoral assistant at LANES, and his co-authors provide an MoS2-based in-memory processor committed to among the essential operations in data processing: vector-matrix reproduction. This operation is ubiquitous in digital signal processing and the implementation of expert system models. Improvements in its effectiveness could yield significant energy cost savings throughout the entire ICT sector.
Their processor combines 1024 components onto a one-by-one-centimeter chip. Each aspect consists of a 2D MoS2 transistor as well as a drifting gate, utilized to keep a charge in its memory that controls the conductivity of each transistor. Coupling processing and memory in this way essentially alters how the processor carries out the calculation. “By setting the conductivity of each transistor, we can perform analog vector-matrix reproduction in a single step by applying voltages to our processor and measuring the output,” explains Kis.
A Big Step Closer to Practical Applications
The option of product– MoS2– played an essential role in the advancement of their in-memory processor. Unlike silicon, the most widely utilized semiconductor in todays computer processors, MoS2 forms a steady monolayer, just 3 atoms thick, that just connects weakly with its environments. In 2010, they created their very first single MoS2 transistor using a monolayer of the product peeled off a crystal using Scotch tape.
Over the previous 13 years, their procedures have grown substantially, with Migliato Maregas contributions playing an essential role. “The key advance in going from a single transistor to over 1000 was the quality of the product that we can deposit. After a lot of process optimization, we can now produce entire wafers covered with a homogenous layer of consistent MoS2. This lets us adopt industry standard tools to create integrated circuits on a computer system and equate these styles into physical circuits, unlocking to mass production,” states Kis.
Renewing European Chip Manufacturing
Aside from its purely clinical value, Kis sees this outcome as a testimony to the importance of close clinical partnership in between Switzerland and the EU, in particular in the context of the European Chips Act, which aims to boost Europes competitiveness and durability in semiconductor technologies and applications. “EU financing was crucial for both this project and those that preceded it, including the one that funded the deal with the very first MoS2 transistor, revealing just how essential it is for Switzerland,” states Kis.
” At the exact same time, it shows how work performed in Switzerland can benefit the EU as it looks for to revitalize electronics fabrication. Rather than running the same race as everyone else, the EU could, for example, focus on establishing non-von Neumann processing architectures for AI accelerators and other emerging applications. By defining its own race, the continent could get a running start to secure a strong position in the future,” he concludes.
Referral: “A massive integrated vector– matrix reproduction processor based upon monolayer molybdenum disulfide memories” by Guilherme Migliato Marega, Hyun Goo Ji, Zhenyu Wang, Gabriele Pasquale, Mukesh Tripathi, Aleksandra Radenovic and Andras Kis, 13 November 2023, Nature Electronics.DOI: 10.1038/ s41928-023-01064-1.