iTnews Asia
  • Home
  • News
  • Enterprise Technology

Samsung’s future memory sticks aimed at supercomputing, AI and machine learning

Samsung’s future memory sticks aimed at supercomputing, AI and machine learning

The highly anticipated super-fast modules should eventually filter down to the consumer market and the crossover from DDR4 to DDR5 is expected to take place by 2023.

By iTnews Asia Team on Apr 28, 2021 8:42AM

Samsung will be releasing later this year a new iteration of memory chips – the first in seven years – that will keep pace with the growth of data centres and AI computing.

The 512GB DDR5 module is the first to use High-K Metal Gate (HKMG) tech, offering 7,200 Mbps speeds — over double that of DDR4, Samsung said in a press statement.

Samsung first used HKMG tech in 2018 with GDDR6 chips used in GPUs. Developed by Intel, it uses hafnium instead of silicon, with metals replacing the normal polysilicon gate electrodes.

The new module will be used in servers performing ‘the most extreme compute-hungry, high-bandwidth workloads.’ Right now, it's aimed at supercomputing, AI and machine learning functions.

A fit for data centre power needs

The high dielectric material contained in the layer reduces current leakage and therefore allows higher performance. At the same time, Samsung is also reducing power usage in the new module by 13%, making it ideal for data centres.

With 7,200 Mbps speeds, Samsung's latest module would deliver around 57.6 GB/s transfer speeds on a single channel.

The chip industry has been eagerly awaiting the launch of the new memory standard and support of it will also come with the arrival of Intel’s Xeon Scalable processors, codenamed Sapphire Reds.

Samsung is currently sampling different variations of the DDR5 memory product families to customers for verification and certification.

To reach the editorial team on your feedback, story ideas and pitches, contact them here.
© iTnews Asia
Tags:
enterprise architecture enterprise solutions enterprise technology

Related Articles

  • Strategic control of networks can provide enterprises a competitive edge
  • Grab acquires AI robotics firm Infermove for last mile delivery
  • Agentic AI solves Frankenstack problem in marketing without breaking what works
  • OpenAI expands ChatGPT lineup with two new models
Share on Twitter Share on Facebook Share on LinkedIn Share on Whatsapp Email A Friend

Most Read Articles

Samsung pips TSMC to produce chips for Tesla’s next-gen self-driving cars

Samsung pips TSMC to produce chips for Tesla’s next-gen self-driving cars

Olam's digital roadmap is "hyper everything"

Olam's digital roadmap is "hyper everything"

Grab acquires AI robotics firm Infermove for last mile delivery

Grab acquires AI robotics firm Infermove for last mile delivery

Four Japanese companies co-develop facial recognition payment system

Four Japanese companies co-develop facial recognition payment system

All rights reserved. This material may not be published, broadcast, rewritten or redistributed in any form without prior authorisation.
Your use of this website constitutes acceptance of Lighthouse Independent Media's Privacy Policy and Terms & Conditions.