[CLSTR]
< Back to situations

Monitor this situation.

[SITUATION] · [QUIET] · [TECHNOLOGY]

2 clusters · 8 sources · 3 days · First seen · Last updated

University of Texas SOT-MRAM development for AI

Overview

Researchers at the University of Texas at Austin have developed and tested a magnetic memory technology known as SOT-MRAM (Spin-Orbit Torque Magnetic Random-Access Memory) to address the energy and computational demands of artificial intelligence.

In collaboration with Taiwan Semiconductor Manufacturing Company (TSMC), engineers tested the non-volatile technology, which retains information without a continuous power supply. This allows devices to reduce energy consumption by avoiding data reloading during sleep modes or restarts. Testing involving neural network training and inference demonstrated write operations of 2 nanoseconds, consuming only 2 picojoules of energy per write.

As AI models increase in complexity, the SOT-MRAM architecture aims to mitigate computational bottlenecks caused by data movement between memory and processors. By bringing memory and calculation closer together, the technology allows certain operations to be executed directly within the memory devices, offering faster state changes and higher efficiency than existing solutions.

Entities

SOT-MRAM · University of Texas at Austin · Science Advances · Taiwan Semiconductor Manufacturing Company

Claims

What the coverage asserts, and how many sources carry each claim.

Timeline

  1. 11 days ago

    [TECHNOLOGY] 4 sources
    University of Texas researchers develop SOT-MRAM for AI efficiency

    University of Texas researchers developed SOT-MRAM, a non-volatile magnetic memory designed to make AI faster and more energy-efficient by performing calculations directly within memory devices.

  2. 14 days ago

    [TECHNOLOGY] 3 sources
    University of Texas and TSMC test SOT-MRAM to reduce AI energy use

    Researchers from the University of Texas and TSMC have tested SOT-MRAM, a non-volatile memory technology designed to reduce the high energy consumption of artificial intelligence systems.

Sources

androidsis.com · diarioestrategia.cl · lacronicadebadajoz.elperiodicoextremadura.com · laopinioncoruna.es · noticiasdemalaga.es · ocio.diariodemallorca.es · ocio.diarioinformacion.com · startup.mandiner.hu