OFC 2024: Marvell Displays Accelerated Infrastructure Portfolio Gems

OFC 2024: Marvell Displays Accelerated Infrastructure Portfolio Gems

The News: Marvell outlined its broad-ranging strategy to optimize optical technology to meet the rising performance and power requirements of AI and cloud data centers at OFC 2024. Read the full press release on the Marvell website.

OFC 2024: Marvell Displays Accelerated Infrastructure Portfolio Gems

Analyst Take: Marvell unveiled new products, portfolio development breakthroughs, and partnerships at OFC 2024 targeted at ecosystem-wide scaling of optical technology for the accelerated infrastructure age. Marvell adroitly used OFC 2024 to underscore the major drivers of rapid optical growth and Marvell’s portfolio development approach for enhancing coherent digital signal processors (DSPs), optical PAM4 DSPs, silicon photonics, and other optical technologies aimed at data centers to serve a growing range of architectures, applications, use cases, performance characteristics, and customer profiles.

The key products and use cases underscored at OFC 2024 included Marvell Nova 2, Marvell 3D SiPho Engine, Spica Gen2-T, COLORZ 800, O-band Optics, collaboration with Juniper Networks, and Coherent Collaborate on ZR/ZR+ transport solution, as well as customer momentum for Orion with InnoLight and HG Genuine. The key takeaways consisted of:

Marvell Nova 2: Breakthrough PAM4 Optical DSP Integrates 200 Gbps Electrical and Optical Interfaces

Marvell debuted Marvell Nova 2, a 1.6 Tbps PAM4 optical DSP featuring 200 Gbps electrical and optical interfaces aimed at fulfilling the intensifying performance demands of accelerated infrastructure, generative AI, and high-performance computing. 1.6T Nova optical DSPs enable module manufacturers to develop a diverse array of industry-standard form factor optical modules that deliver twice the bandwidth than current 800 Gbps optical modules for the transition to 200G interfaces in accelerated infrastructure.

From my view, doubling optical module and interface bandwidth can substantially increase the amount of network traffic that can be managed within the same physical space, blazing an ecosystem-wide trail to scale capacity and increase return on infrastructure investments. Of key importance, the Nova family uses four generations of Marvell’s proven PAM4 optical DSP technology. Nova 2 features eight 200 Gbps electrical lanes to the host device and eight 200 Gbps optical lanes to interface with a wide range of optical components to enable 1.6T total bandwidth that can fit inside standardized module form factors using eight electrical lanes.

Nova 2 based optical modules support backward compatibility for up to three generations. Nova 2 is designed for the fast-growing AI networking fabric infrastructure where switches, network interface cards (NICs), and accelerators will be based with 200 Gbps I/O interfaces. The first flagship Nova device is designed for connecting devices based with 100 Gbps electrical I/O interfaces, used in clouds presently.

Marvell 3D SiPho Engine: Vanguard of AI and Accelerated Infrastructure Progress

Marvell demonstrated its 3D Silicon Photonics (SiPho) engine featuring 32 channels of 200G electrical and optical interfaces for connecting next generation AI clusters and cloud data centers at multi-terabit speeds. The Marvell 3D SiPho engine combines hundreds of components such as waveguides and modulators, photodetectors, modulator drivers, trans-impedance amplifiers, microcontrollers, and a host of other passive components into a single, unified device to bolster the performance, bandwidth, and energy efficiency of optical interconnects. The 200 Gbps device delivers twice the bandwidth, twice the input/output (I/O) bandwidth density, and 30% lower power per bit versus comparable devices with 100 Gbps electrical and optical interfaces.

I find that integral to the solution’s new capabilities is the integration flexibility of the laser technology into the package as well as offering it off-package to a pluggable module. As such, the light engine can be used as a co-packaged optics solution or as a pluggable optical module and can support Tbps-class performance at sub 10 picojoules per bit, which is vital to attaining energy efficiency benefits across AI and accelerated infrastructure environments.

Spica Gen2-T: Spearheading New Wave of Energy Efficient Modules

Marvell unveiled Spica Gen2-T, a 5nm 800 Gbps transmit-only PAM4 optical DSP. Designed for transmit retimed optical modules (TRO modules), Spica Gen2-T can reduce the power consumption of 800 Gbps optical modules by more than 40% while maintaining interoperability with conventional optical modules and IEEE 802.3 compliant host devices.

The new offering is important as TRO modules use transmit-only DSPs to digitize, process, equalize, and amplify data from host devices, performing electrical-to-optical conversion on the egress path; and use transimpedance amplifiers (TIAs) only to perform the optical-to-electrical conversion on the ingress path. As a result, TROs with transmit-only DSPs can reduce power consumption yet maintain the auto adaptation and compensation capability to minimize part-to-part, system-to-system variation when deployed in scale, which has become critical for large AI system deployment.

COLORZ 800: Expanding the ZR/ZR+ Opportunity in the AI Era

Marvell’s COLORZ 800 family of optical modules, incorporating Orion 800 Gbps coherent DSP and field-proven silicon photonics platform, which integrates multiple discrete components into a single die, provides up to 800 Gbps of bandwidth for DCI links up to 500 km. COLORZ 800 can also be enabled for connections up to 1,200 km operating at 400/600 Gbps to replace traditional transport equipment across a greater range of cloud infrastructure.

COLORZ 800 can lower the capital cost of data center interconnect (DCI) by up to 75% compared to traditional rack-based DCI boxes with embedded DSPs while reducing operating expenses such as power, rack space, and real estate. From my viewpoint, such competitive advantages can accelerate adoption of COLORZ 800 across intensely competitive DCI environments. At OFC 2024, Marvell demonstrated an innovative probabilistic constellation shaping (PCS) capable 800 Gbps ZR/ZR+ small form factor pluggable module.

From my perspective, PCS support is a key portfolio breakthrough as it continually adapts the modulation of coherent signals to maximize transmission capacity or minimize spectrum and energy. The technology enables optical fiber to approach the Shannon Limit, the theoretical maximum transfer rate achievable. With PCS enabled, the reach of a Marvell COLORZ 800 module at full bandwidth doubles from 500 km to 1000 km.

I expect that Marvell’s COLORZ 800 offering can energize uptake of coherent dense-wavelength division multiplexing (DWDM) by offering the same size as 800G gray optics and providing an integrated amplifier in ZR+ that can deliver top competitive cost and energy savings per GB. Through enabling more versatile, speedier DCI capabilities, COLORZ 800 is primed to play an instrumental role in the expansion of virtual hyperscaler data centers, especially in ensuring more efficient AI processing as well as improved dynamic load balancing.

From my view, COLORZ 800 alongside Orion 800 strengthen Marvell’s overall data infrastructure proposition, also consisting of its Teralynx 10, a 51.2T switch chip for powering cloud switch systems, and the Nova/Nova 2 1.6T PAM4 DSP capable of transmitting 200G per wavelength of light to double the capacity of PAM4 modules. Marvell also offers the 400 Gbps and 800 Gbps Alaska A DSPs for active electrical cable (AEC) capabilities. Moreover, Marvell produces drivers and amplifiers included in optical modules. Taken together, Marvell offers the data infrastructure silicon portfolio required to accelerate ecosystem adoption of coherent pluggable technology, especially across cloud environments.

Marvell, Juniper, and Coherent: 800ZR Power Trio

Marvell partnered with Coherent Corp., a provider of compound semiconductors and high-speed optical networking technology, and Juniper Networks, a supplier of secure AI-Native Networking, to present the first public demonstration of their 800ZR solution featuring Juniper Networks PTX10002-36QDD Packet Transport Router, Coherent Corporation’s 800ZR transceiver and the 800G Marvell Orion coherent DSP.

From my view, the solution highlighted significant ecosystem progress in transport technology and the constant innovation, suggesting a new benchmark of capabilities and efficiencies. As such, hyperscalers are likely to lead the deployment of 800G in data center and interconnect environments. Accelerating the release of ZR+ versions can power the adoption of these technologies in transport networks, enhancing operator and user experiences.

The 2nm Accelerated Infrastructure Dimension

Underpinning Marvell’s accelerated infrastructure portfolio-wide innovation is the recent TSMC collaboration focused on 2nm infrastructure silicon, which follows Marvell innovative breakthroughs in 5nm and 3nm working with TSMC. Marvell has developed and demonstrated high-speed, ultra-high-bandwidth silicon interconnects produced on TSMC 3nm process. Marvell’s silicon building blocks in this node include 112G XSR SerDes (serializer/de-serializer), Long Reach SerDes, PCIe Gen 6/CXL 3.0 SerDes, and a 240 Tbps parallel die-to-die interconnect.

The building blocks are part of Marvell’s continued execution of its strategy to develop a comprehensive silicon intellectual property (IP) portfolio for designing chips that can increase the bandwidth, performance, and energy efficiency of rapidly evolving data infrastructure. These technologies also support all semiconductor packaging options from standard and low-cost Redistribution Layers (RDLs) to silicon-based high-density interconnect.

For fast expanding AI/GenAI workloads, 2nm will prove essential for providing the massive gains required across transistor density, area, power, and performance factors. Accordingly, I find that Marvell can offer sharply differentiated analog, mixed-signal, and foundational IP to deliver accelerated IP innovation. Fundamentally, Marvell’s modular approach to semiconductor research and design that focuses first on qualifying foundational analog, mixed-signal IP and advanced packaging is well-suited for a vast array of devices.

Key Takeaways: Marvell Optical Portfolio Ready for Accelerated Infrastructure Prime Time

I believe that Marvell’s OFC 2024 announcements demonstrate that its portfolio can provide the technological and business foundation to meet the massive scaling demands of optical data rates alongside delivering the energy per bit savings essential to transforming data infrastructure. Through Marvell’s optical-related portfolio, data infrastructure can play an even more integral role in driving ecosystem-wide innovation across key application domains such as AI, cloud computing, smart devices, and streaming media.

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 Industry Analyst Day 2023: Accelerated Computing Takes Off

Marvell TSMC: Stimulating 2nm Accelerated Infrastructure Innovation

The Futurum Group at Marvell’s Industry Analyst Day 2023 with Sandeep Bharathi – Futurum Tech Webcast

 

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.

SHARE:

Latest Insights:

Brad Shimmin, VP and Practice Lead at The Futurum Group, examines why investors behind NVIDIA and Meta are backing Hammerspace to remove AI data bottlenecks and improve performance at scale.
Looking Beyond the Dashboard: Tableau Bets Big on AI Grounded in Semantic Data to Define Its Next Chapter
Futurum analysts Brad Shimmin and Keith Kirkpatrick cover the latest developments from Tableau Conference, focused on the new AI and data-management enhancements to the visualization platform.
Colleen Kapase, VP at Google Cloud, joins Tiffani Bova to share insights on enhancing partner opportunities and harnessing AI for growth.
Ericsson Introduces Wireless-First Branch Architecture for Agile, Secure Connectivity to Support AI-Driven Enterprise Innovation
The Futurum Group’s Ron Westfall shares his insights on why Ericsson’s new wireless-first architecture and the E400 fulfill key emerging enterprise trends, such as 5G Advanced, IoT proliferation, and increased reliance on wireless-first implementations.

Book a Demo

Thank you, we received your request, a member of our team will be in contact with you.