Monday, May 13, 2024

Aurora links 63,744 GPUs with Cray Slingshot Interconnects

The Aurora supercomputer at the U.S. Department of Energy’s Argonne National Laboratory has officially surpassed the exascale threshold, achieving over a quintillion calculations per second, as announced today at the ISC High Performance 2024 conference in Hamburg, Germany.

Built by Intel and Hewlett Packard Enterprise (HPE), Aurora features a groundbreaking architecture, inclusing 63,744 graphics processing units (GPUs), making it  the world's largest GPU-powered system with more interconnect endpoints than any other system to date.



Aurora Architecture Highlights

Processing Units

  • Intel CPUs: Aurora is equipped with next-generation Intel Xeon Scalable processors.
  • Intel GPUs: The system includes Intel's Ponte Vecchio GPUs, which are designed for high-performance computing (HPC) and artificial intelligence (AI) workloads.

Performance

Exascale Performance: Aurora is expected to deliver performance exceeding one exaFLOP (10^18 floating-point operations per second). This places it among the first exascale systems in the world, capable of performing a quintillion calculations per second.

Memory

  • High-Bandwidth Memory: Aurora incorporates high-bandwidth memory (HBM) for both its CPUs and GPUs, which enhances data transfer rates and overall computational efficiency.
  • Unified Memory Architecture: The system uses a unified memory architecture that allows for seamless data sharing between CPUs and GPUs, reducing latency and improving performance.

Interconnect

  • Cray Slingshot: Aurora uses the Cray Slingshot high-speed interconnect, which offers advanced network capabilities, low latency, and high bandwidth. The Cray Slingshot interconnect is based on Ethernet technology, rather than Infiniband. 
  • Per-Link Throughput: Each link in the Slingshot network provides up to 200 gigabits per second (Gbps) of bandwidth. This high per-link throughput ensures rapid data transfer rates, crucial for the vast data sets and intensive computations typical in HPC workloads.
  • Network Scalability: Slingshot's architecture allows for scaling up to very large node counts, providing high aggregate bandwidth that can support thousands of nodes in an exascale system.
  • Adaptive Routing: Dynamic selection of optimal paths to avoid congestion and improve efficiency.
  • Quality of Service (QoS): Multiple QoS levels to prioritize critical traffic.
  • Scalability: Supports large-scale deployments with thousands of nodes, making it suitable for exascale systems.

Storage

  • Lustre File System: Aurora is expected to use the Lustre parallel file system, providing fast and scalable storage solutions that can handle the immense data throughput generated by exascale computing workloads.

The installation team, comprising staff from Argonne, Intel, and HPE, is focused on system validation, verification, and scaling up. They are addressing various hardware and software issues as the system approaches full-scale operations.

“Aurora is fundamentally transforming how we do science for our country,” Argonne Laboratory Director Paul Kearns said. ​“It will accelerate scientific discovery by combining high performance computing and AI to fight climate change, develop life-saving medical treatments, create new materials, understand the universe and so much more.”

NVIDIA Grace Hopper powers AI Supercomputers

 Nine new supercomputers worldwide are using NVIDIA Grace Hopper Superchips to speed scientific research and discovery, including:

  • EXA1-HE, in France, from CEA and Eviden; 
  • Helios at Academic Computer Centre Cyfronet, in Poland, from Hewlett Packard Enterprise (HPE); 
  • Alps at the Swiss National Supercomputing Centre, from HPE; 
  • JUPITER at the J├╝lich Supercomputing Centre, in Germany; 
  • DeltaAI at the National Center for Supercomputing Applications at the University of Illinois Urbana-Champaign;
  • and Miyabi at Japan’s Joint Center for Advanced High Performance Computing — established between the Center for Computational Sciences at the University of Tsukuba and the Information Technology Center at the University of Tokyo.

Combined, the systems deliver 200 exaflops, or 200 quintillion calculations per second, of energy-efficient AI processing power.

NVIDIA’s accelerated computing platform comprises NVIDIA Hopper architecture-based GPUs, NVIDIA Grace CPU Superchips, NVIDIA Grace Hopper Superchips, NVIDIA Quantum-2 InfiniBand networking and a full suite of NVIDIA AI and HPC software.



Google and HP to offer Project Starline video conferencing

HP and Google disclosed plans to start commercialization of their Project Starline video conferencing service in 2025.

Project Starline leverages AI, 3D imaging, spatial audio, real-time compression and other technologies to create a “magic window” effect with the video service. 

Google says HP's expertise in computing, combined with their investment in Poly’s collaboration solutions, makes them uniquely positioned to deliver new and innovative experiences worldwide. 

As Alex Cho, President of Personal Systems at HP says, “With more than half of meaning and intent communicated through body language versus words alone, an immersive collaboration experience plays an important role in creating authentic human connections in hybrid environments. We are proud to partner with Google to bring this technology to market, harnessing the power of AI to shape the future of collaboration."

https://blog.google/technology/research/project-starline/

Windstream's Beach Route Alliance: Jacksonville to Ashburn

Windstream Wholesale recently activated its dark fiber Beach Route, providing fiber links from Jacksonville, Florida, to Myrtle Beach, South Carolina, and extending to Raleigh, North Carolina. 

In a concurrent move, Windstream Wholesale has partnered with Mid-Atlantic Broadband, Tilson Infrastructure, and SummitIG to further expand the dark fiber network. This collaboration, known as the Beach Route Dark Fiber Alliance, extends the route an additional 315 miles from Benson, North Carolina, to Richmond and Ashburn, Virginia. This partnership aims to streamline network connectivity by eliminating the need for multi-vendor solutions, thus offering customers a unified service experience.

Highlights:

  • Beach Route Availability: The 690-mile route connects significant cable landing stations and offers reduced latency, benefiting cloud services and content delivery networks.
  • Strategic Partnerships: The formation of the Beach Route Dark Fiber Alliance extends the fiber network and simplifies customer engagement by consolidating service management under Windstream Wholesale.
  • Enhanced Connectivity: The expansion connects key aerospace hubs and research areas, improving infrastructure for technological and industrial growth.
  • Simplified Service Model: Customers get a streamlined service with a single contract and a dedicated account team, making network management more efficient.


Polar Semi lands funding to become US-owned merchant fab

Polar Semiconductor will receive $120 million in proposed direct funding as part of the U.S. CHIPS and Science Act and a $75 million investment from the State of Minnesota expects to expand its manufacturing in Minnesota. In addition, Polar has entered into a definitive agreement pursuant to which Niobrara Capital and Prysm Capital are leading an equity investment of $175 million

Overall, the company said it plans to invest approximately $525 million, including these funds, in the expansion of the facility over the next two years, enabling it to become a U.S.-owned merchant foundry.

Through these investments, Polar expects to:

  • Double its current U.S. production capacity of 200mm semiconductor wafers, increasing production from approximately 20,000 wafers per month to nearly 40,000 wafers per month;
  • Expand and modernize its facility with new automation and AI capabilities to become globally competitive through economies of scale;
  • Better serve customers in automotive, aerospace and defense, optoelectronics, MEMS, and medical devices in the U.S. with cutting-edge semiconductor products; and
  • Create over 160 new jobs to further strengthen Polar’s support of its community.

Surya Iyer, President and COO of Polar Semiconductor, said, “We are very pleased to announce this historic investment in Minnesota semiconductor manufacturing. Our expanded manufacturing facility will allow us to increase capacity and branch into innovative technologies to serve new customers and markets. Polar and its employees are grateful to the U.S. Department of Commerce and the State of Minnesota for their commitment to the future of American semiconductor manufacturing and appreciate the strong collaboration with the CHIPS Program Office, Minnesota Department of Employment and Economic Development (“DEED”), and the City of Bloomington, Minnesota, throughout this process. Polar is also pleased to welcome a significant equity investment from Niobrara Capital and Prysm Capital, which will allow the Company to become U.S.-owned, and for the continued support of our long-term partners, Sanken Electric and Allegro MicroSystems.”

https://polarsemi.com/news/polar-semiconductor-announces-plans-to-expand-semiconductor-manufacturing-facility-in-minnesota/

Semiconductor Industry Assoc projects expansion in U.S. capacity

 The United States will triple its domestic semiconductor manufacturing capacity from 2022—when the CHIPS and Science Act (CHIPS) was enacted—to 2032, according Boston Consulting Group (BCG) and the Semiconductor Industry Association (SIA).

The study, titled “Emerging Resilience in the Semiconductor Supply Chain,” also projects the U.S. will grow its share of advanced logic (below 10nm) manufacturing to 28% of global capacity by 2032, up from 0% in 2022. Additionally, America is projected to capture over one-quarter (28%) of total global capital expenditures (capex) from 2024-2032, ranking second only to Taiwan (31%). In the absence of the CHIPS Act, the U.S. would have captured only 9% of global capex by 2032, according to the report.

Other key report findings:

  • America’s world-leading 203% projected increase in fab capacity from 2022 to 2032 stands in stark contrast to its modest 11% increase from the previous decade (2012-2022), which ranked last among all major chip-producing regions, according to the SIA/BCG report.
  • The U.S. share of the world’s chip manufacturing capacity will increase from 10% in 2022—when the CHIPS and Science Act was enacted—to 14% by 2032, marking the first time in decades the U.S. has grown its domestic chip manufacturing footprint relative to the rest of the world. In the absence of CHIPS enactment, the U.S. share would have slipped further to 8% by 2032, according to the report.
  • The U.S. continues to lead the world in its overall contribution to the global value chain, with strong leadership positions in high value-added areas of semiconductor technology, including chip design, electronic design automation (EDA), and semiconductor manufacturing equipment.

“Effective policies, such as the CHIPS and Science Act, are spurring more investments in the U.S. semiconductor industry. These investments will help America grow its share of global semiconductor production and innovation, furthering economic growth and technological competitiveness,” said Rich Templeton, Chairman of the Board at Texas Instruments and SIA board chair. “Continued and expanded government-industry collaboration will help ensure we build on this momentum and continue our next steps forward.”

https://www.semiconductors.org/america-projected-to-triple-semiconductor-manufacturing-capacity-by-2032-the-largest-rate-of-growth-in-the-world/

Harmonic unveils Pearl R-OLT module for fiber broadband

Harmonic introduced its Pearl Remote Optical Line Termination (R-OLT) module designed for fiber broadband service delivery, compatible with Harmonic's Ripple and Oyster node enclosures. The Pearl R-OLT module is the latest addition to Harmonic's ecosystem of fiber and DOCSIS edge devices, which are supported by the company's cOS virtualized core.

The high-density Pearl R-OLT module is designed to assist operators in streamlining their fiber upgrades, thus speeding up the deployment of high-speed fiber broadband services. It supports multiple fiber technologies such as GPON, ComboPON, and XGS-PON. The integration of various functionalities into a single module allows operators to provide high-speed broadband services efficiently, regardless of the back office infrastructure.

Key Features of Pearl R-OLT:

  • Compatible with Harmonic's Ripple and Oyster node enclosures
  • Supports GPON, ComboPON, and XGS-PON technologies
  • Can be installed in various configurations including underground vaults, compact pedestals, or aerial setups
  • Simplifies network expansion and reduces associated costs
  • Enhances subscriber experience and retention in both greenfield and existing service areas
  • Includes long-range OLT optics for cost-effective edge-out strategies
  • Supports up to 8 XGS-PON for delivering 10G fiber services
  • Powered by Harmonic's cOS virtualized broadband platform, serving over 28 million homes globally

https://www.harmonicinc.com/broadband/pearl-olt/

Intel appoints Kevin O’Buckley to head Foundry Services

Intel appointed Kevin O’Buckley as senior vice president and general manager of Foundry Services, the customer service and ecosystem operations division of Intel Foundry. 

O’Buckley most recently was senior vice president of hardware engineering for the Custom, Compute and Storage Group at Marvell Technologies. He joined Marvell as part of its 2019 acquisition of Avera Semiconductor, where he served as chief executive. He also served as vice president of Product Development at Global Foundries, and before that spent more than 17 years at IBM leading technology development and manufacturing organizations

O’Buckley starts today and becomes a member of Intel’s executive leadership team reporting to CEO Pat Gelsinger. He succeeds Stuart Pann, who will retire from Intel after 35 years of service at the end of May and remain an adviser to support a seamless transition.