Hardware Podcasts

How 1.6T Ethernet will Enable the World's Fastest Datacenters

EE Times Current · 20 Mar 2024 · 28m

What this episode covers

Priyank Shukla, principal product manager for the Synopsys HPC IP portfolio, explains how Synopsys’s 1.6T Ethernet solution, featuring 224G SerDes and linear optical interfaces, addresses data bottlenecks in AI data centers. He details the convergence of digital logic (MAC/PCS) and physical layer technologies to reduce latency by 40-50% and power consumption by 50% compared to previous generations. The discussion covers the transition from 400G/800G to 1.6T, the role of the IEEE 802.3dj task force, and the Ultra Ethernet Consortium’s focus on cache coherence and tail latency for large language model training.

Most relevant to

large enterprises

Topics

Mentioned

Companies
Synopsys
Products
ChatGPT
Chips
224G SerDes
Also
Priyank Shukla, IEEE 802.3, PAM4, Ultra Ethernet Consortium

Related episodes from other shows

Other shows on the same ground, closest first.

  • The Circuit · 49m

    EP 146: The State of AI Networking with Austin Lyons

    Austin Lyons, writing as Chip Strat, joins Ben Bajarin and Jay Goldberg to discuss the evolution of AI networking, defining the critical distinctions between scale-up, scale-out, and scale-across topologies. The conversation details the technical trade-offs between copper and optical interconnects, explaining how SerDes technology enables higher speeds over shorter distances and why optical is necessary for longer ranges. They highlight the strategic importance of SerDes capabilities for companies like Broadcom and Marvell, and touch on emerging trends like Celestial's photonic fabric and the acquisition of Alpha Wave by Qualcomm.

    Wireless & RFChips, FPGAs & semiconductors
    4 chapters

  • The Circuit · 43m

    Ep 160: More good News for Intel, AI Model Trends, Future of Datacenter

    Ben Bajarin and Jay Goldberg analyze Intel's recent strategic moves, including a partnership with Tesla for its 'Terafab' project and co-design work with Google on its IPU, interpreting these as validation for Intel's advanced packaging and foundry capabilities. They discuss the broader semiconductor market's valuation disconnect, noting how investors are shifting focus from established compute giants to obscure optical and mineral supply chain bottlenecks. The hosts also examine the future of data centers, arguing that extreme co-optimization of power (800 VDC) and cooling is becoming essential to unlock compute capacity, and cover Anthropic's rapid product rollout, including its Word connector and 'Mythos' model, while highlighting rising token costs and capacity constraints.

    Chips, FPGAs & semiconductorsEnergy & climate tech
    6 chapters

  • OnTrack · 39m

    High-Speed Interconnects for AI & Physical AI with Samtec

    Matt Burns, Director of Technical Marketing at Samtec, discusses the evolution of high-speed interconnects for AI data centers and embedded systems. He explains how connector selection has shifted from an afterthought to a critical early-stage architectural decision to support data rates of 112G, 224G, and 448G. The conversation covers specific product families like Sci-Flex LP, Accelerated HP, and Sci-Flex HD, and details the move toward co-packaged copper and optics to reduce signal loss. Burns also explores how these high-speed technologies are trickling down to 'physical AI' applications in robotics and edge computing.

    PCB design & layoutWireless & RF
    5 chapters

  • Moore's Lobby · 51m

    Can Silicon Photonics Solve the Bandwidth Bottleneck in AI Data Centers?

    Steve Klinger, VP of Product at LightMatter, explains how the company uses silicon photonics to solve the interconnect bandwidth bottleneck in AI data centers. He details LightMatter's flagship product, Passage, a programmable optical fabric that acts as an active interposer to connect chiplets and ASICs with high density and low power. Klinger also discusses Envis, LightMatter's photonic AI accelerator, which performs matrix-vector multiplication using light, and shares insights on the company's growth from MIT research to a VC-funded scaling phase.

    PCB design & layoutChips, FPGAs & semiconductors
    7 chapters

  • The Circuit · 52m

    EP 163: Breaking the Memory Wall: Micron’s Strategy for the AI Era

    Jeremy Werner, SVP and GM of Micron’s Core Data Center Business Unit, explains how AI is shifting the memory industry from a cyclical model to sustainable growth driven by the 'memory wall' in inference. He details Micron's strategy to break this bottleneck through innovations like HBM4 and ultra-high-capacity SSDs, which address critical data center constraints of power and physical space. The conversation covers the technical hierarchy of memory, from HBM to context storage, and Micron's aggressive capacity expansion plans to meet surging demand for AI training and inference workloads.

    Chips, FPGAs & semiconductorsPCB design & layout
    7 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by EE Times Current. The episode itself belongs to its publisher.