Sunday, June 2, 2024

NVIDIA scales its Spectrum-X Ethernet networking platform

At COMPUTEX 2024 in Taiwan, NVIDIA's Jensen Huang unveiled a roadmap for new semiconductors that will arrive on a one-year rhythm. 

The Rubin platform will succeed the upcoming Blackwell platform, featuring new GPUs, a new Arm-based CPU — Vera — and advanced networking with NVLink 6, CX9 SuperNIC and the X1600 converged InfiniBand/Ethernet switch.

“Our company has a one-year rhythm. Our basic philosophy is very simple: build the entire data center scale, disaggregate and sell to you parts on a one-year rhythm, and push everything to technology limits,” Huang explained.

Spectrum-X features the NVIDIA Spectrum SN5600 Ethernet switch and the NVIDIA BlueField -3 SuperNIC. The platform leverages adaptive routing and congestion control for maximum bandwidth and noise isolation. It enables advanced cloud multi-tenancy, GPU compute elasticity and zero-trust security.

NVIDIA will launch new Spectrum-X products every year, delivering increased bandwidth and ports and enhanced software feature sets and programmability.  NVIDIA claims Spectrum-X accelerates generative AI network performance by 1.6x over traditional Ethernet fabrics. Data center networks based Spectrum-X switches are currently designed connecting for tens of thousands. With 1.6 Tbps interfaces on the horizon this will soon increase to millions of GPUs.

The next generation NVLink Switch chip will feature:

  • 50B transistors and use TSMC's 4NP 
  • 72-ports of 400G SerDes
  • 4 NVLinks at 1.8 TB/s
  • 7.2 TB/s full-duplex bandwidth

Another key networking aim for NVIDIA is bringing the capabilities of Infiniband to Ethernet for hyperscale data centers, including

  • Network-level RDMA
  • Congestion control using the switch telemetry
  • Adaptive routing using the Bluefield NICs
  • Noise isolation between training models

NVIDIA cited rapid adoption of its Spectrum-X Ethernet networking platform. 

CoreWeave, GMO Internet Group, Lambda, Scaleway, STPX Global and Yotta are among the first AI cloud service providers embracing NVIDIA Spectrum-X. Additionally, several NVIDIA partners have announced Spectrum-based products, including ASRock Rack, ASUS, GIGABYTE, Ingrasys, Inventec, Pegatron, QCT, Wistron and Wiwynn, which join Dell Technologies, Hewlett Packard Enterprise, Lenovo and Supermicro in incorporating the platform into their offerings.

“Rapid advancements in groundbreaking technologies like generative AI underscore the necessity for every business to prioritize networking innovation to gain a competitive edge,” said Gilad Shainer, senior vice president of networking at NVIDIA. “NVIDIA Spectrum-X revolutionizes Ethernet networking to let businesses fully harness the power of their AI infrastructures to transform their operations and their industries.”

AMD scales its GPU accelerators for AI

During an opening keynote at COMPUTEX in Taiwan, AMD's CEO, Dr. Lisa Su, unveiled the company's next gen Instinct MI325X GPU accelerator with up to 288GB of HBM3E memory for release later this year. AMD will pursue an annual cadence for new product releases.

Following the Instinct MI325X, the next AMD Instinct MI350 series, powered by the new AMD CDNA 4 architecture, is expected to be available in 2025 bringing up to a 35x increase in AI inference performance compared to AMD Instinct MI300 Series with AMD CDNA 3 architecture. Expected to arrive in 2026, the AMD Instinct MI400 series is based on the AMD CDNA “Next” architecture.

“The AMD Instinct MI300X accelerators continue their strong adoption from numerous partners and customers including Microsoft Azure, Meta, Dell Technologies, HPE, Lenovo and others, a direct result of the AMD Instinct MI300X accelerator exceptional performance and value proposition,” said Brad McCredie, corporate vice president, Data Center Accelerated Compute, AMD. “With our updated annual cadence of products, we are relentless in our pace of innovation, providing the leadership capabilities and performance the AI industry and our customers expect to drive the next evolution of data center AI training and inference.”

Finally, AMD highlighted the demand for AMD Instinct MI300X accelerators continues to grow with numerous partners and customers using the accelerators to power their demanding AI workloads, including:

  • Microsoft Azure using the accelerators for Azure OpenAI services and the new Azure ND MI300X V5 virtual machines.
  • Dell Technologies using MI300X accelerators in the PowerEdge XE9680 for enterprise AI workloads.
  • Supermicro providing multiple solutions with AMD Instinct accelerators.
  • Lenovo powering Hybrid AI innovation with the ThinkSystem SR685a V3
  • HPE is using them to accelerate AI workloads in the HPE Cray XD675.

New subsea cable to link Singapore with Batam, Indonesia

Singtel and PT Telekomunikasi Indonesia International (Telin) agreed to develop a new submarine cable system connecting Singapore and Batam, Indonesia, under the newly formed INSICA (Indonesia Singapore Cable System) Consortium.

When operational in the fourth quarter of 2026, the 100-km INSICA cable system will support the surge in data centre telecommunications traffic between Singapore and Batam. INSICA will feature a 24-fiber pair subsea cable and two diverse terrestrial cable paths, offering a maximum capacity of up to 20 terabits per second per fiber pair. This will deliver exceptional bandwidth, seamless connectivity and robust network security and enable efficient resource sharing and scalability. The new diverse link provided by INSICA will enhance network protection and reliability, ensuring uninterrupted 24/7 operations for data centres.

Mr Ooi Seng Keat, Vice President of Digital Infrastructure & Services at Singtel, stated, “Batam is emerging as a prime location for data centres due to its close proximity to Singapore. With this cable system, we’ll be able to enhance the connectivity between the countries to support the intensive, higher power density AI workloads of enterprises and cloud companies. The development of the INSICA cable system is yet another step that we’re taking in architecting a hyper-connected, digital ecosystem to serve the long-term demands of the region’s digital future and boost the regional economy.”

STM plans high-volume 200mm silicon carbide fab in Italy

STMicroelectronics has unveiled plans to establish a new high-volume silicon carbide (SiC) manufacturing facility in Catania, Italy. This facility will be part of ST’s Silicon Carbide Campus, designed for mass production of SiC power devices and modules, including a full range of manufacturing processes from substrate to packaging.

Key points include:

  • Integration of Facilities: The new facility will work alongside an SiC substrate manufacturing site to create a fully integrated manufacturing hub for SiC on a single site.
  • Silicon Carbide Campus Features: The campus will encompass all production stages, including SiC substrate development, epitaxial growth, and front-end wafer fabrication using 200mm technology. It will also handle module assembly and have extensive R&D and design capabilities.
  • Production and Investment: Production is expected to start in 2026, with full capacity reached by 2033, producing up to 15,000 wafers weekly. The total investment is projected at about five billion euros, with around two billion euros in support from the Italian government under the EU Chips Act.
  • Sustainability Commitment: The campus design and operations will emphasize sustainable practices, particularly in resource consumption like water and power.


“The fully integrated capabilities unlocked by the Silicon Carbide Campus in Catania will contribute significantly to ST’s SiC technology leadership for automotive and industrial customers through the next decades,” said Jean-Marc Chery, President and Chief Executive Officer of STMicroelectronics. “The scale and synergies offered by this project will enable us to better innovate with high-volume manufacturing capacity, to the benefit of our European and global customers as they transition to electrification and seek more energy efficient solutions to meet their decarbonization goals.”


FCC marks the end of the Affordable Connectivity Program

Due to a lack of additional Congressional funding, the FCC officially ended the Affordable Connectivity Program (ACP).  

During the wind-down of the ACP, Chairwoman Rosenworcel sent monthly letters to Congress emphasizing the program’s importance and the need for additional funding. In a recent letter, she highlighted the nationwide necessity to support low-income families struggling to afford high-speed internet and detailed the Commission’s actions to mitigate the impact of the ACP’s conclusion on enrolled households. 

“The Affordable Connectivity Program filled an important gap that provider low-income programs, state and local affordability programs, and the Lifeline program cannot fully address,” said Chairwoman Rosenworcel. “The Commission is available to provide any assistance Congress may need to support funding the ACP in the future and stands ready to resume the program if additional funding is provided.”

Agency wind-down measures have included: (1) encouraging ACP providers, for which participation in the ACP was voluntary, to develop low-income programs of their own and to provide their ACP subscribers information on their low-income programs or low-cost plans; (2) offering training and resources for state public utility commissions and agency ACP grantees and outreach partners to raise awareness of the Commission’s Lifeline program; (3) and reminding current Lifeline providers of their requirement to publicize the Lifeline program.  

The Lifeline program offers a $9.25 monthly benefit on broadband service for eligible households. Although the Lifeline benefit may alleviate some financial pressure for certain ACP households, it is not a replacement for the ACP.  Not all ACP households will qualify for Lifeline, and by statute, many ACP providers are not eligible to participate in the Lifeline program.

Some highlights on the Affordable Connectivity Program: 

  • Over 23 million households were enrolled in the ACP at the time the program stopped accepting new enrollments. 
  • The ACP served households in every county in the United States.  
  • The participation among households in Tribal areas increased by 136 percent, with approximately 330,000 Tribal subscribers enrolled in the program when the enrollment freeze took effect.
  • It provided low-income households consistent connectivity.  In response to the FCC’s winter 2023 agency survey, 68% of ACP households reported they had inconsistent connectivity or zero connectivity at all before ACP.   
  • Roughly 15 percent of all households in the program are from rural areas. 
  • According to a national survey, more than four million households with an active or former military member are enrolled in the ACP.  
  • Nearly half of ACP households are led by someone over the age of 50. 
  • Approximately 3.4 million households seeking to enroll in the ACP indicated participation in the National School Lunch or Breakfast Programs as one of the ways they qualify for the ACP.