Marvell Unveils Industry's First 3nm 1.6 Tbps PAM4 Interconnect Platform to Scale Accelerated Infrastructure
New Ara PAM4 DSP Reduces Optical Module Power by 20%, Enabling Mass Adoption of 200 Gbps per Lane and 1.6 Tbps Network Infrastructure to Meet Rising AI Bandwidth Demands
SANTA CLARA, Calif. — Dec. 3, 2024 – Marvell Technology, Inc. (NASDAQ: MRVL), a leader in data infrastructure semiconductor solutions, today introduced Marvell® Ara, the industry’s first 3nm 1.6 Tbps PAM4 interconnects platform featuring 200 Gbps electrical and optical interfaces. Building on the success of the Nova 2 DSP, the industry’s first 5nm 1.6 Tbps PAM4 DSP with 200 Gbps electrical and optical interfaces, Ara leverages the comprehensive Marvell 3nm platform with industry-leading 200 Gbps SerDes and integrated optical modulator drivers, to reduce 1.6 Tbps optical module power by over 20%. The energy efficiency improvement reduces operational costs and enables new AI server and networking architectures to address the need for higher bandwidth and performance for AI workloads, within the significant power constraints of the data center.
Ara, the industry’s first 3nm PAM4 optical DSP, builds on six generations of Marvell leadership in PAM4 optical DSP technology. It integrates eight 200 Gbps electrical lanes to the host and eight 200 Gbps optical lanes, enabling 1.6 Tbps in a compact, standardized module form factor. Leveraging 3nm technology and laser driver integration, Ara reduces module design complexity, power consumption and cost, setting a new benchmark for next-generation AI and cloud infrastructure.
“Ara sets a new industry standard by leveraging advanced 3nm technology to deliver significant power reduction, driving the volume adoption of 1.6 Tbps connectivity for AI infrastructure,” said Xi Wang, Vice President of Product Marketing for Optical Connectivity at Marvell. “With a co-optimized companion TIA, our next-generation PAM4 optical DSP platform empowers customers to scale generative AI and large-scale compute applications with best-in-class performance and unmatched energy efficiency.”
“Ara is another Marvell optical connectivity industry-first solution, delivering the power efficiency required for the most demanding AI workloads,” said Osa Mok, chief marketing officer, InnoLight Technology. “The Ara platform combined with InnoLight’s advanced high-speed optical transceiver design and manufacturing expertise, offers the industry a state-of-the-art pluggable module optimized for next-generation AI and cloud infrastructure.”
“We anticipate unit shipments of PAM4 DSPs will more than triple from 2024 to 2029 to nearly 127 million units a year and remain the primary optical technology for connecting assets inside data centers for the foreseeable future,” said Bob Wheeler, Analyst at Large, LightCounting. “Ara marks another first for Marvell and demonstrates that PAM4 technology continues to evolve to meet the challenges of AI infrastructure.”
Optimized for next-generation AI and cloud infrastructure, Ara is designed to support high-density 200 Gbps I/O interfaces across switches, network interface cards (NICs) and XPUs, while ensuring backward compatibility with prior generations. With best-in-class power efficiency and integration, Ara addresses the growing requirements of hyperscale data centers, to deliver high-performance accelerated infrastructure with best-in-class total cost of ownership (TCO).
Ara Platform Key Features
- 200 Gbps per channel support, providing high bandwidth for next-generation AI-driven applications.
- 200 Gbps per lane line-side receiver with companion Marvell TIA CB11269TA, providing best-in-class linearity and low noise for AI applications.
- PAM4 modulation for efficient high-speed data transmission – critical for AI and cloud applications.
- Integrated, high-swing laser driver to improve performance while reducing overall transceiver module design complexity, power consumption, and TCO.
- Enhanced crossbar switching capabilities within a streamlined architecture, improving routing flexibility across channels.
- InfiniBand and Ethernet support for versatile interconnect flexibility across diverse network topologies, enhancing adaptability for accelerated infrastructure.
Availability
Marvell Ara will sample to select customers in Q1 2025.
About Marvell
To deliver the data infrastructure technology that connects the world, we’re building solutions on the most powerful foundation: our partnerships with our customers. Trusted by the world’s leading technology companies for over 25 years, we move, store, process and secure the world’s data with semiconductor solutions designed for our customers’ current needs and future ambitions. Through a process of deep collaboration and transparency, we’re ultimately changing the way tomorrow’s enterprise, cloud, automotive, and carrier architectures transform—for the better.
|
Related News
- Marvell Announces Industry's First 2nm Platform for Accelerated Infrastructure Silicon
- Cadence Unveils Millennium Platform - Industry's First Accelerated Digital Twin Delivering Unprecedented Performance and Energy Efficiency
- Marvell Demonstrates Industry's First 3nm Data Infrastructure Silicon
- Marvell Extends Data Infrastructure Leadership with TSMC 3nm Platform
- Renesas Unveils Industry's First Automotive Multi-Domain SoC Built with 3-nm Process Technology
Breaking News
- Ultra Accelerator Link Consortium (UALink) Welcomes Alibaba, Apple and Synopsys to Board of Directors
- Breaking Ground in Post-Quantum Cryptography Real World Implementation Security Research
- RIKEN adopts Siemens' emulation and High-Level Synthesis platforms for next-generation AI device research
- CAST to Enter the Post-Quantum Cryptography Era with New KiviPQC-KEM IP Core
- InPsytech Announces Finalization of UCIe IP Design, Driving Breakthroughs in High-Speed Transmission Technology
Most Popular
- Eighteen New Semiconductor Fabs to Start Construction in 2025, SEMI Reports
- InPsytech Announces Finalization of UCIe IP Design, Driving Breakthroughs in High-Speed Transmission Technology
- Imagination pulls out of RISC-V CPUs
- Chip Interfaces Successfully Completes Interlaken IP Interoperability Test with Cadence 112G Long-Reach PHY
- RISC-V in AI and HPC Part 2: Per Aspera Ad Astra?
E-mail This Article | Printer-Friendly Page |