Steady technical progress and continued market traction were key points of emphasis coming from the PCI-SIG Developers Conference held this week in Santa Clara. According to a new ABI Research re …
The first numbers of the available bandwidth between chiplets is out – UCIe is estimating that chiplet packages could squeeze out communication speeds of 630Gbps, or 0.63Tbps, in a very tight a …
Check out our list of 108 illustrious winners across 22 different categories of HPC.
May 29, 2023
Nvidia launched a new Ethernet-based networking platform – the Nvidia Spectrum-X – that targets generative AI workloads. Based on tight coupling of the Nvi Read more…
April 19, 2023
Researchers at Meta, MIT and other institutions connected servers with a dozen Nvidia GPUs with optical switches and a robotic arm, devising a new interconnect Read more…
April 10, 2023
There are limits on the speed of how fast copper wires can move data between computers, and a transition to light speed will ultimately drive AI and high-performance computing forward. Every major chipmaker is in agreement that optical interconnects will be needed to reach zettascale computing in an energy-efficient way. That opinion was... Read more…
April 5, 2023
MLCommons today released the latest MLPerf Inferencing (v3.0) results for the datacenter and edge. While Nvidia continues to dominate the results – topping al Read more…
March 16, 2023
Sometime later this year, perhaps around July, the Department of Defense is expected to announce the sites and focus of up to nine hubs associated with the Micr Read more…
March 6, 2023
While quantum computing makes its way haphazardly towards practical realty, quantum key distribution (QKD) is moving quickly towards greater commercial use. Tod Read more…
March 1, 2023
Ayar Labs, founded in 2015, is pursuing optics as a means of driving higher interconnect speeds and efficiencies in computing. Now, the company is announcing th Read more…
October 18, 2022
Spun out from Google last March, SandboxAQ is a fascinating, well-funded start-up targeting the intersection of AI and quantum technology. “As the world enter Read more…
Making the Most of Today’s Cloud-First Approach to Running HPC and AI Workloads With Penguin Scyld Cloud Central™
Bursting to cloud has long been used to complement on-premises HPC capacity to meet variable compute demands. But in today’s age of cloud, many workloads start on the cloud with little IT or corporate oversight. What is needed is a way to operationalize the use of these cloud resources so that users get the compute power they need when they need it, but with constraints that take costs and the efficient use of existing compute power into account. Download this special report to learn more about this topic.
Data center infrastructure running AI and HPC workloads requires powerful microprocessor chips and the use of CPUs, GPUs, and acceleration chips to carry out compute intensive tasks. AI and HPC processing generate excessive heat which results in higher data center power consumption and additional data center costs.
Data centers traditionally use air cooling solutions including heatsinks and fans that may not be able to reduce energy consumption while maintaining infrastructure performance for AI and HPC workloads. Liquid cooled systems will be increasingly replacing air cooled solutions for data centers running HPC and AI workloads to meet heat and performance needs.
QCT worked with Intel to develop the QCT QoolRack, a rack-level direct-to-chip cooling solution which meets data center needs with impressive cooling power savings per rack over air cooled solutions, and reduces data centers’ carbon footprint with QCT QoolRack smart management.
© 2023 HPCwire. All Rights Reserved. A Tabor Communications Publication
HPCwire is a registered trademark of Tabor Communications, Inc. Use of this site is governed by our Terms of Use and Privacy Policy.
Reproduction in whole or in part in any form or medium without express written permission of Tabor Communications, Inc. is prohibited.