Hewlett Packard Enterprise (HPE) announced the delivery of a new-generation supercomputer for the Academic Computer Centre Cyfronet at AGH University of Krakow, which places Poland at the forefront of this system. Named ‘Helios‘, it is Cyfronet’s fifth-generation supercomputer, created to drive AI-driven scientific research and establish new commercial companies.
Helios System Overview
The three partitions of the Helios, based on the HPE Cray EX supercomputer, are distinctly designed to perform specific computational tasks. However, the first partition is dedicated to traditional computing science research. The second partition has NVIDIA GPUs to supercharge AI applications and research. The third partition is devoted to workloads undertaking extensive big data processing in an interactive setting.
This supercomputer achieves a theoretical peak performance of 35 petaflops, more than four times that of the previous Cyfronet flagship system. It also boasts an impressive peak AI performance of 1.8 exaflops, allowing users to apply AI capabilities and drive scientific innovations at unparalleled computing power.
Also Read: Top IT, Cloud, Cybersecurity News Updates: Weekly Highlight
The Significance of Helios
With the value of much-needed compute resources increasing for all scientific fields—including chemistry, medicine, materials technology, astronomy, and environmental protection—it was necessary that Cyfronet be able to accommodate the increasing demand. This new system will be accessible to research communities in Poland’s public and private sectors, including research institutions, municipalities, and hospitals.
While energy efficiency became a priority in the face of rapidly increasing energy prices, Cyfronet required a system whose efficiency exceeded that of the previous models. In response to this, HPE Cray EX supercomputer uses liquid cooling. HPE is the world leader in energy-efficient computing, with its share of the world’s top 10 most energy-efficient supercomputers being the majority. Because of its direct liquid cooling technology, a key feature of the supercomputing solution, Helios efficiently dissipates heat. It reduces energy consumption, especially when applied to compute-intensive cases such as generative AI.
Helios: Customized Partitions for Diverse Workloads
The Helios supercomputer is rigorously designed with three specialized partitions, tailored to serve different types of workloads, and these are tailored for scientific communities and commercial entities:
- HPE Cray Supercomputing EX with CPU Partition: This part is comprised of 75,264 “Zen 4” cores obtained from 4th Gen AMD EPYC processors and is paired with 200 terabytes of DDR5 memory. It is primarily used for modelling and simulations, such as computational fluid dynamics for wind turbines and propellers, simulations of car crash events, and medical applications such as proton therapy beam simulations for cancer treatment.
- HPE Cray Supercomputing EX with GPU Partition: Fitted with 440 NVIDIA GH200 Grace Hopper Superchips, this partition is optimized for image-intensive computer simulations in materials engineering, solid-state physics, drug discovery, and large-scale AI training, including generative AI applications.
- HPE Cray Supercomputing XD665 with “INT” Partition: This partition is dedicated to interactive workloads that require very significant data processing, AI model optimization, inference, and everything similar—24
- NVIDIA H100 Tensor Cores and high-speed NVMe local storage make this a very powerful platform for tuning AI models and efficiently running other AI-driven applications.
Also Read: Innovating Security Operations: The Role of Generative AI
Conclusion
HPE Slingshot connects all of Helios seamlessly, enabling smooth operation and speedy performance. HPE Slingshot is purpose-built, high-performance networking that addresses the increasing requirements for speed and congestion control inherent to workloads. This tunable interconnection mechanism dramatically improves overall system performance by offering ultra-fast networking capabilities, leveraging HPE Cray Technology.
Helios also incorporates a strong storage infrastructure, featuring 17.5 petabytes of Lustre file system storage, which has been thoughtfully engineered on a strong HPE Cray ClusterStor E1000 Storage System. This large storage capacity translates to ease of data management and access, thus providing researchers and innovators with a means to put the vast capabilities of computational science into the service of their missions with unprecedented agility and precision.
FAQs
1. What is Helios, and why is it significant for Cyfronet AGH?
Helios is the latest-generation supercomputer Hewlett Packard Enterprise (HPE) deployed at the Academic Computer Centre Cyfronet, AGH University of Krakow. It represents Cyfronet’s fifth-generation supercomputer and is designed to drive AI-driven scientific research while supporting the establishment of new commercial endeavors.
2. What are Helios’ key specifications and performance metrics?
Helios boasts a theoretical peak performance of 35 petaflops and a peak AI performance of 1.8 exaflops. It comprises three specialized partitions, each tailored for specific computational tasks, with HPE Cray EX supercomputing technology at its core.
3. How does Helios accommodate diverse computational workloads?
The supercomputer features three partitions: one dedicated to traditional computational science research, another equipped with NVIDIA GPUs for AI applications, and a third tailored for interactive workloads involving extensive big data processing.
4. How does Helios contribute to scientific advancements in various fields?
Helios addresses the escalating demand for computing resources across scientific domains such as chemistry, medicine, materials technology, astronomy, and environmental protection. Its accessibility to research communities in Poland’s public and private sectors, including research institutions, municipalities, and hospitals, fosters collaborative innovation.
5. What measures have been taken to ensure energy efficiency in Helios?
HPE Cray EX supercomputer employs liquid cooling technology to enhance energy efficiency, effectively dissipating heat and reducing energy consumption. This feature is crucial, especially in compute-intensive scenarios such as generative AI applications.
6. How does Helios facilitate connectivity and data storage?
Helios integrates HPE Slingshot for seamless connectivity, offering purpose-built, high-performance networking with a bandwidth of 200 gigabits per second. Additionally, it incorporates a robust storage infrastructure comprising 17.5 petabytes of Lustre file system storage, ensuring efficient data management and accessibility.
[To share your insights with us as part of editorial or sponsored content, please write to sghosh@martechseries.com]
