Marvell Right Sizes AEC Connections to Meet New AI Acceleration Demands

Marvell Right Sizes AEC Connections to Meet New AI Acceleration Demands

The News: Marvell, a provider of data infrastructure semiconductor solutions, launched the Marvell Alaska A 1.6T PAM4 DSP for active electrical cables (AECs) to address emerging 200G/lane-based accelerated infrastructure architectures. Read the full press release on the Marvell website.

Marvell Right Sizes AEC Connections to Meet New AI Acceleration Demands

Analyst Take: Marvell is expanding its Alaska A family with the debut of Marvell A 1.6T PAM4 DSP built for AECs targeted at nascent 200G/lane-based accelerated architectures. The latest addition to the Alaska A family is built on a 5nm process and uses Marvell PAM4 DSP technology to enable short-reach copper interconnect solutions that fulfill the expanding bandwidth requirements of accelerated infrastructure.

I find that AI and machine learning (ML) applications are driving greater bandwidth connectivity requirements between AI accelerators, server to top-of-rack links, and switch-to-switch interconnects within data center racks. As such, the Alaska A 1.6T DSP is designed for rapidly evolving accelerated infrastructure with 200 Gbps I/O interfaces on AI accelerators, GPUs, NICs, and switches. Key features of the Alaska A 1.6T AEC DSP include:

  • Proven DSP-based 200G PAM4 SerDes
  • Advanced digital equalizer with Feed Forward (FFE), Decision Feedback (DFE), and Maximum Likelihood Sequence Detection (MLSD)
  • Greater than 3-meter reach at 200G/lane
  • 200 Gbps per lane electrical line-side interface
  • Optimized for QSFP-DD and OSFP form factors
  • Retiming and gear-boxing support
  • Cable reference design for 1.6T AECs
  • SDK support for advanced telemetry and diagnostics

The move follows on Marvell expanding its connectivity portfolio with the launch of the Alaska P peripheral component interconnect express (PCIe) retimer product line that is designed to scale the compute fabrics of accelerated infrastructure. The retimer offering is built on Marvell 5nm PAM4 technology aimed at enabling the new Alaska P PCIe retimer product line, starting with the 8- and 16-lane PCIe Gen 6 retimers, to scale connections between AI accelerators, GPUs, CPUs, and other components inside servers.

The Alaska P PCIe launch has positioned Marvell to drive PAM4 PCIe innovations throughout AI server compute fabrics. Marvell is building on more than a decade of PAM4 technology expertise exemplified by its 5nm PAM4 IP portfolio innovation prowess. This capitalizes on a major industry inflection point as compute fabrics for PCIe and CXL transition from NRZ to PAM4. Likewise, I see the Alaska A 1.6T PAM4 DSP propelling the innovation of 200G/lane-based accelerated infrastructure architectures by catalyzing AEC adoption.

Marvell Right Sizing Copper Connectivity to Meet Fast-evolving Accelerated Compute Rack Demands

Today accelerated infrastructure environments use a combination of long-reach optical interconnects (>5m) for core switch to spine switch plus spine switch to rack switch across AI clusters and short-reach copper interconnects (<5m) for AI compute fabrics, AI accelerators to rack switches, and NICs to rack switches. As such, copper interconnects are prominently used in AI server and server to rack switch implementations. As a result, higher accelerated compute demands using AI accelerators are driving the enormous demand for higher bandwidth connectivity that requires the move to using serial 200G lanes. However, 200G serial technology limits copper connection distances as demonstrated by the inability of existing direct attach copper (DAC) cables to meet 200G serial reach and cable thickness requirements. The solution to DAC limitations is using Active Electrical Cable (AEC) technology, since it enables longer reach over thinner cables. From my view, the innovative breakthrough of the Marvell Alaska A 1.6T DSP solution is assuring cloud-optimized short-reach copper connectivity. This includes extending copper reach to >3 meters (m) that fulfill the thickness and cable reach needs of 200G serial cabling throughout accelerated infrastructure connections. This is achieved primarily by AECs being 50% thinner than equivalent DACs.

AI driving bandwidth growth across the full array of connectivity tiers including data center-to-data center, cluster-to-cluster, server-to-server, and inside server environments. PCIe Gen 6 is the first PCIe standard to use PAM4 signaling, replacing the non-return-to-zero (NRZ) modulation used over the last two decades. In accord, inside AI server compute fabrics are accelerating migration to PCIe Gen 6 to attain the faster connections needed between AI accelerators, GPUs, CPUs, and other server components.

Marvell Alaska A 1.6T DSP: A Team of AEC Suppliers Ready to Ride Marvell PAM4 IP Acumen

By working with top-tier AEC suppliers, such as Amphenol, Molex, and TE Connectivity, Marvell bolsters its channel position to spur the delivery of cloud-optimized AECs that fulfill the unique requirements of demanding cloud environments. This includes supporting 3m+ reach, 28 to 32 American Wire Gauge (AWG) form factors, and gearbox cable capabilities.

From my view, the breakthrough Alaska A 1.6T DSP offering extends copper connectivity on an innovative basis by delivering a complete platform for customers to use. This includes flexible 224G/112G/50G PAM4 support, retimer and gearbox modes, comprehensive debug and diagnostics capabilities, plus 12mm X 14mm package size.

Marvell’s 5nm in-house PAM4 IP acumen, as demonstrated by multiple products already shipping in high volume, provides the foundation for competitive advances such as SerDes supporting >40dB insertion loss compensation as well as enabling copper connectivity >3 meters. Marvell PAM4 technology, including 200G, 400G (4 x 100G lanes), and 800G (8 x 100G lanes), is already deployed extensively in top-tier cloud data centers to provide the foundation for AEC-enabled 1.6T (200G per lane) innovation as well as efficiently scaling massive bandwidth expansion.

Marvell is capitalizing on its field-proven data connectivity portfolio, consisting of its data center interconnect (DCI) network (i.e., Marvell Coherent DSP, Driver, TIA), frontend and backend networks (i.e., Marvell PAM4 DSP, Drive, TIA, Teralynx), and inside server fabric (Alaska P PCIe Retimer) offerings to win more mindshare and influence across the connectivity ecosystem.

Key Takeaway: Marvell Alaska A 1.6T PAM 4 DSP Makes Sure AI Accelerators Go the Right Distance

Overall, I believe that the Alaska A 1.6T PAM4 DSP debut strengthens Marvell’s ability to ensure AI accelerators meet the demands for higher bandwidth 200G/lane interconnects underpinned by the industry-wide migration of copper interconnects from DAC to AEC. With the Alaska A debut, Marvell alongside top-tier cable providers, Amphenol, Molex, and TE Connectivity, can drive the 1.6T AEC DSP innovation critical to catalyzing short-reach AI accelerator technology and further fortify its robust presence across the data center interconnect market.

Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.

Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.

Other insights from The Futurum Group:

Marvell Sees Time Has Come for Alaska-sized Retimer Innovation

Marvell Q4 and Fiscal Year 2024: AI Takes Center Stage

OFC 2024: Marvell Displays Accelerated Infrastructure Portfolio Gems

Author Information

Ron is an experienced, customer-focused research expert and analyst, with over 20 years of experience in the digital and IT transformation markets, working with businesses to drive consistent revenue and sales growth.

He is a recognized authority at tracking the evolution of and identifying the key disruptive trends within the service enablement ecosystem, including a wide range of topics across software and services, infrastructure, 5G communications, Internet of Things (IoT), Artificial Intelligence (AI), analytics, security, cloud computing, revenue management, and regulatory issues.

Prior to his work with The Futurum Group, Ron worked with GlobalData Technology creating syndicated and custom research across a wide variety of technical fields. His work with Current Analysis focused on the broadband and service provider infrastructure markets.

Ron holds a Master of Arts in Public Policy from University of Nevada — Las Vegas and a Bachelor of Arts in political science/government from William and Mary.


Latest Insights:

The Futurum Group team assesses how the global impact of the recent CrowdStrike IT outage has underscored the critical dependency of various sectors on cybersecurity services, and how this incident highlights the vulnerabilities in digital infrastructure and emphasizes the necessity for robust cybersecurity measures and resilient deployment processes to prevent widespread disruptions in the future.
On this episode of The Six Five Webcast, hosts Patrick Moorhead and Daniel Newman discuss CrowdStrike Global meltdown, Meta won't do GAI in EU or Brazil, HP Imagine AI 2024, TSMC Q2FY24 earnings, AMD Zen 5 Tech Day, Apple using YouTube to train its models, and NVIDIA announces Mistral NeMo 12B NIM.
An Overview of Significant Advancements and Announcements in the AI Software and Services Market in June 2024
Keith Kirkpatrick, Research Director with The Futurum Group, covers the highlights in the enterprise AI market in June 2024, featuring key players such as Apple, Meta, OpenAI, Stability, Google Deepmind, Meta, and McDonald’s.
An Overview of Significant Advancements and Announcements in the Enterprise Applications Market in June 2024.
Keith Kirkpatrick, Research Director with The Futurum Group, covers the highlights in the enterprise application market in June 2024, featuring players such as Salesforce, Pegasystems, SAP, Microsoft, Adobe, Oracle Netsuite, and more.