Monday, May 20, 2024

Broadcom intros 400G RoCE/RDMA Ethernet NICs

 Broadcom has announced the release of its latest 400G PCIe Gen 5.0 Ethernet adapters, aimed at enhancing the efficiency and performance of data centers, particularly those focused on AI applications. These new products are designed to address the needs for higher bandwidth and improved connectivity in environments where XPU bandwidth and cluster sizes are expanding rapidly.

Key Points:

  • Advanced Manufacturing: The adapters are the industry’s first to be built using 5nm process technology, which contributes to better power and thermal efficiency.
  • Improved Connectivity Options: The adapters support passive copper cables up to five meters and ultra-low power linear pluggable optics transceivers. They are compatible with Broadcom’s high-radix Ethernet switches, facilitating higher rack density through mainstream air-cooling technologies.
  • Designed for AI Networks: With features like a third-generation RoCE pipeline and low-latency congestion control, the adapters are suited to the demanding requirements of AI data centers.
  • Compatibility and Standards: The software for the adapters is vendor-neutral, supporting various CPUs, GPUs, PCIe, and Ethernet switches. This reflects Broadcom’s commitment to open standards, aligning with its role in the Ultra Ethernet Consortium.

The BCM57608 adapter series is now available from multiple server vendors and directly from Broadcom. Jas Tremay, Vice President and General Manager of the Data Center Solutions Group at Broadcom, highlighted the importance of these adapters in supporting the evolving requirements of AI-focused data centers. This development is part of Broadcom’s ongoing efforts to strengthen its position in Ethernet data center infrastructure.

Microsoft and Qualcomm herald the AI PC era

 In a significant leap for PC technology, Microsoft, in collaboration with global OEMs, has unveiled a new range of PCs powered by Qualcomm's cutting-edge Snapdragon X Elite and Snapdragon X Plus processors. These devices promise to transform user interactions through advanced AI-driven capabilities.

Central to this technological revolution is Microsoft's Copilot+, designed to radically change how users engage with their PCs. The innovative software leverages the powerful new silicon, capable of delivering over 40 trillion operations per second (TOPS). This immense processing power is paired with all-day battery life and support for the most sophisticated AI models currently available.

A standout feature of this new suite of technologies is "Recall," a function that utilizes a personal semantic index on the device itself to manage and organize files more intuitively.

Qualcomm has not held back in its ambition, claiming its neural processing unit (NPU), powered by Snapdragon X Elite, sets a new standard in performance efficiency for laptops. According to Qualcomm, the Snapdragon X Elite offers up to 2.6 times the performance per watt compared to Apple's M3 chip, and up to 5.4 times that of Intel’s Core Ultra 7.

This announcement marks a pivotal moment in PC computing, promising a future where PCs are not only faster and more energy-efficient but also smarter and more responsive to user needs through AI integration.        

Tech Update: Massive Bandwidth Growth to/from AI Data Centers

The surging demand for AI data centers will put a strain on the underlying infrastructure. Russ Esmacher, SVP Strategy, Marketing and Corporate Development from Infinera discusses: 

 - The proliferation of data centers supports applications such as AI, cloud, and compute, driving bandwidth and connectivity across all markets. 

- There is a huge need for solutions and tools to stay ahead of this curve of massive bandwidth growth, focusing on cost efficiency and power management.

Have a tech update that you want to brief us on? Contact!

Check out other Tech Updates on our YouTube Channel (subscribe today): and check out our latest reports at:


Dell’Oro: RAN Market Still a Disaster

Preliminary findings reveal that the Radio Access Network (RAN) market is still struggling after the peak in 2022, according to a new report from Dell’Oro Group. The results in the first quarter were exceptionally weak, underpinned by poor results across most suppliers. Preliminary findings suggest that the overall 2G-5G RAN market—including baseband plus radio hardware and software, excluding services—declined 15 to 30 percent in 1Q 2024, resulting in a third consecutive quarter of double-digit contractions.

“It’s difficult to find a silver lining in the first quarter,” said Stefan Pongratz, Vice President and analyst at the Dell’Oro Group. “We’ve been monitoring the RAN market since the year 2000, and the contraction experienced in the first quarter marked the steepest decline in our entire history of covering this market. In addition to the known coverage related challenges that the market is dealing with when comps in the advanced 5G markets are becoming more challenging, there are now serious concerns about the timing of the capacity upgrades given current network utilization levels and data traffic growth rates,” continued Pongratz.

Additional highlights from the 1Q 2024 RAN report:

  • Middle East and Africa are growing, Latin America is stable, and the remaining regions, including North America, Europe, and Asia Pacific declined sharply in 1Q 2024.
  • Vendor rankings remain stable while vendor shares are shifting. In the first quarter, Huawei’s 4QT (four quarter trailing) revenue share improved relative to 2023 while Nokia lost some ground over the same period.
  • The top 5 RAN suppliers based on worldwide revenues are Huawei, Ericsson, Nokia, ZTE, and Samsung.
  • * Growth projections have been revised slightly downward over the short term. Global RAN is now projected to decline 5 to 8 percent in 2024.

Dell’Oro: Microwave Transmission Market Dips 8% in Q1

Point-to-point Microwave Transmission equipment revenue declined 8 percent year-over-year in the first quarter of 2024, according to a new report from Dell’Oro Group. The market contraction was mostly due to a rapid deceleration in mobile radio deployments in the quarter.

“This was a tough quarter for the microwave transmission equipment market,” stated Jimmy Yu, Vice President at Dell’Oro Group. “The market slowdown occurred across multiple geographies with only two regions posting a positive year-over-year growth rate. Those regions were North America and the Middle East and Africa. As a result of the sharp slowdown in mobile networks, we lowered our full year 2024 outlook to only a one percent growth rate. We continue to believe the market environment will strengthen as the year unfolds, but the first half of 2024 is now projected to be much weaker than we initially anticipated,” added Yu.

Additional highlights from the 1Q 2024 Microwave Transmission & Mobile Backhaul Quarterly Report:

  • The Microwave Transmission market declined year-over-year in 1Q 2024, driven by less demand for Mobile Backhaul. Revenue associated with Mobile Backhaul declined 11 percent year-over-year in the quarter.
  • Among the major regions for microwave installations for wireless backhaul, India had the steepest decline, shrinking by 50 percent compared to the year-ago quarter.
  • Four vendors (Aviat, Ceragon, Huawei, and ZTE) accomplished to increase their revenue in 1Q 2024 compared to the same period last year. Aviat and Ceragon benefited from acquisitions that closed at the end of 2023.

SambaNova adds Lip-Bu Tan as Executive Chairman

SambaNova Systems, a full-stack AI platform driven by its custom silicon, appointed Lip-Bu Tan as its new Executive Chairman. 

Lip-Bu Tan is a recipient of the Robert N. Noyce Award, the Semiconductor Industry Association’s highest honor, for his outstanding contributions. In 1987, he founded Walden International, a venture capital firm that has been instrumental in driving the growth of semiconductor innovation. Lip-Bu’s transformative leadership as Executive Chairman and CEO of Cadence Design Systems resulted in him being awarded the Dr. Morris Chang Exemplary Leadership Award by the Global Semiconductor Alliance.

"It will be an honor to guide SambaNova as Executive Chairman," said Lip-Bu Tan. "The potential of this company is immense — SambaNova has differentiated itself by providing a privatized, full-stack AI platform that enterprises can own; the fastest, open models, most efficient chips, software and services. SambaNova is committed to ensuring customers can own and deploy their data and models wherever, and however, they want. I’m excited to be a part of their fast growth.”

"We are thrilled to have Lip-Bu step into the Executive Chairman role. I look forward to working with him on a daily basis to accelerate our growth," said Rodrigo Liang, Co-Founder and CEO of SambaNova Systems. "His unparalleled experience and visionary leadership will be invaluable as we enter new markets. We’re committed to building the world's leading privatized AI platform for the enterprise, with a strong focus on customer relationships, innovation, and strategic execution."


Aviz enables RoCE-based AI fabric for Tensorwave’s GPU service

 Aviz Networks is working with TensorWave, a start-up building a GPU-as-a-Service. The collaboration is focused on intelligent networks powering AI by implementing RoCE (RDMA over Converged Ethernet)-based AI fabrics to optimize GPU as a service offerings. The implementation features Aviz’s Open Network Enterprise Suite (ONES) multi-vendor SONiC solution.

By deploying Aviz Networks’ technology, TensorWave will enhance its GPU as a service offering by utilizing advanced AMD-based MI300 accelerators to efficiently meet the growing demands of the market, as companies seek to leverage GenAI, LLMs and machine learning for their businesses. 

TensorWave has quickly established itself as a leader in providing GPU as a service. The integration of Aviz technology will enable the management and operations of multi-vendor RoCE-based AI fabrics, crucial for handling diverse GPUs, DPUs, and high radix switches. ONES’s unique capabilities include RoCE orchestration, real-time visibility, and threshold-based anomaly detection, making it the only vendor-agnostic AI fabric controller on the market.

The existing feature set of ONES, which includes support for high-density platform configurations, is particularly beneficial for environments managed by TensorWave. The integration of Network Copilot with ONES extends these capabilities further, providing intelligent guidance and automated management functions, simplifying the complex tasks of network configuration and maintenance.

"Our technology is crucial in enabling TensorWave to provide state-of-the-art GPU services, representing a major advancement in our mission to simplify and elevate Networks for AI, and AI for Networks," Vishal Shukla, Founder and CEO of Aviz Networks

Astera Labs tests PCIe 6.x interoperability

 Astera Labs announced expanded PCIe 6.x testing capabilities in its Cloud-Scale Interop Lab to enable seamless interoperability between Aries 6 PCIe/CXL Smart DSP Retimers and a broad range of PCIe 6.x hosts and endpoints. This paves the way for AI platform developers to design high-bandwidth, low-latency PCIe 6.x connectivity with confidence, reduce overall development time, and deploy at scale.

Higher bandwidth PCIe 6.x technology is required to maximize utilization of GPUs, CPUs, and AI accelerators to meet the performance demands of new AI workloads in hyperscale systems; however, this creates new connectivity issues with increases in speed, complexity, and scale. These challenges emphasize the need for extensive testing to ensure robust interoperability between the wide variety of PCIe 6.x components within an AI system deployed at cloud-scale.

Astera Labs has chosen Taiwan to launch its first Cloud-Scale Interop Lab outside of Silicon Valley given the region’s rich collection of leading technology design and manufacturing companies. This new interop lab location will facilitate closer collaboration with key ODM customers to test Aries 6 in complex PCIe topologies with a broad variety of hosts and end points interconnected over varying channel insertion loss budgets in real systems.

Thad Omura, Chief Business Officer, Astera Labs, said, “As AI systems continue to advance at a rapid pace, data center operators need to deploy increasingly complex systems on an accelerated timeline. Our intense focus on standards compliance and plug-and-play interoperability is foundational to why our widely deployed, field-tested Aries Retimer portfolio sets the gold standard for PCIe/CXL connectivity. Expanding our Cloud-Scale Interop Lab test suite to support PCIe 6.x operation fast-tracks deployment for customers integrating Aries 6 – the industry’s lowest power PCIe 6.x/CXL 3.x Retimer – with solutions from our ecosystem partners.”

Broadcom expand its Global Cybersecurity Aggregator Program

 Broadcom is expanding its Global Cybersecurity Aggregator Program with a new Accelerate Program aimed at providing enhanced service levels for more Broadcom software solutions delivered to a broader range of Broadcom customers across multiple geographies.

Through the Broadcom Accelerate Program, partners will act as an extension of the Broadcom team, helping customers receive the most value from their Broadcom software investment. This includes providing deep technical support capabilities in the customer's preferred language and time zone, as well as enhancing the customer experience through accessibility, specialized knowledge, personalized assistance, training programs, and value-added services.

The Accelerate Program now supports an expanded set of Broadcom’s CA Technologies enterprise software, mainframe software, and Symantec cybersecurity solutions, for Broadcom corporate and commercial customers.

“Our goal is to deliver best-in-class experiences for customers of all sizes, and we continue to deliver on that with our partners through innovative models, like the Broadcom Accelerate Program,” said Cynthia Loyd, vice president of Global Commercial & Partner Sales, Broadcom. “These distributor partners will extend our reach with their local expertise and deep knowledge of Broadcom solutions. This program will help ensure we are delivering superior service and support at all stages of the customer journey.”