Remember meForgot password?
    Log in with Twitter

article imageIBM Research breakthroughs: Nanosheet to Phase-Change Memory

By Tim Sandle     Dec 28, 2019 in Technology
San Fransisco - At the IEDM conference IBM Research unveiled several scientific breakthroughs, including research in nanosheet technology, Phase-Change Memory, Electro-Chemical Random-Access Memory, and new algorithms to improve the accuracy of AI predictions.
The newly announced IBM Research innovations are intended to address a critical issues with artificial intelligence advances, including making hardware systems more efficient to keep pace with the demand of artificial intelligence software and data workloads.
The announcements were made at the International Electron Devices Meeting (IEDM) conference in San Francisco. The event is an annual micro- and nanoelectronics conference, which is held each December. The conference serves as a forum for reporting technological breakthroughs in the areas of semiconductor and related device technologies, design, manufacturing, physics, modeling and circuit-device interaction.
The key highlights from IBM are:
Nanosheet technology
IBM shared its new features for high-performance computing that meet the massive data requirements of AI and 5G. Researchers discussed strategies for enabling stacked nanosheet transistors and multiple-Vt solutions, as well as a new fabrication process: a dry selective etch of silicon-germanium. While this technology continues to scale due to ever-demanding requirements in density, power and performance, innovations are not happening fast enough. Hence, the need for new innovations geared towards creating more powerful AI hardware.
Phase-Change Memory (PCM)
IBM presented how to reduce the impact of PCM conductance drift on the inference of large-scale hardware neural networks, and how a large portion of lost information can be restored by carefully characterizing the devices and amplifying the signal. It also showcased a neuro-inspired fully silicon-integrated ultra-low power prototype chip, which has the potential to exploit AI tasks in edge computing devices to learn in real time.
Electro-Chemical Random-Access Memory (ECRAM)
IBM demonstrated how it has accelerated deep learning with this building block for AI computing by creating new memory devices with materials that already exist in a current semiconductor factory. This enables the development of an analog accelerator, which has potential to beat the speed and power of current digital processors.
More about Ibm, iedm, Research, Innovation
More news from
Latest News
Top News