Synopsys Supports AI-Bogged Data Centers With First 1.6T Ethernet IP

The modern data center experiences unprecedented stress when processing, analyzing, and storing vast amounts of data from artificial intelligence (AI), cloud computing, and high-performance computing (HPC) applications.

 

Data centers will soon require 1.6T Ethernet support.

The IEEE plans to finalize a new iteration of the 1.6TbE standard by 2026, with a baseline set of features slated for 2024.
 

1.6 T Ethernet holds significant promise for supporting the bandwidth and speeds of these growing data demands. Recently, Synopsys announced the “industry’s first” 1.6T Ethernet IP Solution to meet the high bandwidth requirements of AI and hyper-scale data center chips

 

Synopsys Reveals First 1.6T Ethernet IP Solution

Synopsys designed its Complete 1.6T Ethernet IP solution to meet the escalating demands for higher bandwidth and lower latency in data-intensive environments. The company hopes the solution will enable the industry’s fastest chips for AI and data center networking applications.

At the core of Synopsys’ offering are the multi-rate, multi-channel 1.6 T Ethernet media access control (MAC) and physical coding sublayer (PCS) controllers. These multi-channel/multi-rate Ethernet controllers support 1.6 T with up to 40% lower latency and up to 50% smaller area than current multi-rate 800-G IP solutions.

Synopsys’ 1.6T Ethernet IP Solution

Synopsys’ 1.6T Ethernet IP Solution. 
 

Another critical component of the solution is the silicon-proven 224-G Ethernet PHY IP. With support for pulse-amplitude modulation 4-level (PAM-4) and non-return-to-zero (NRZ) signaling, the physical layer interface provides robust link performance, high signal integrity, and interoperability across various channel lengths and connection types. It is customizable to support chip-to-chip, chip-to-module, and copper cable connections, optimizing power and performance trade-offs according to specific design requirements.

According to Synopsys, data centers can use this technology to reduce interconnect power consumption by up to 50% compared to existing solutions.

 

1.6 T Ethernet for Data Centers

The 1.6 T Ethernet networking standard is designed to significantly enhance bandwidth and reduce latency in data-intensive environments. As its name suggests,1.6 T Ethernet can transmit data at speeds of up to 1.6 terabits per second. This represents a sizable leap forward from previous generations, enabling the handling of the massive volumes of data generated by today’s AI algorithms, big data analytics, and other demanding applications.

1.6 T Ethernet promises to speed up reliable data transfer between servers, storage systems, and networking equipment, thereby ensuring that computational resources are utilized efficiently and without bottleneck-induced delays.

 

A full 1.6T Ethernet subsystem and the latency path

A full 1.6 T Ethernet subsystem and the latency path. 
 

A complete 1.6 T Ethernet solution encompasses not just the physical hardware that can support these high-speed data transfers but also the integrated MAC and PCS controllers, PHY interfaces, and comprehensive verification IP. This holistic approach ensures interoperability, reduces integration risks, and accelerates time to market for data center infrastructure components. Moreover, it significantly reduces power consumption and physical space compared to previous solutions, addressing two critical constraints in data center design and operation.

By providing the necessary bandwidth and efficiency, 1.6 T Ethernet enables data centers to support the ongoing evolution of computing and the digital economy.

 

Meeting the Data Center’s Evolving Needs

Synopsys believes its complete 1.6T Ethernet Solution can future-proof the modern data center. With a comprehensive suite of technologies designed to reduce latency, decrease power consumption, and minimize physical space requirements, the company’s solution addresses some of the most pressing challenges faced by data center operators today.  

 


 

All images used courtesy of Synopsys.


Source link

This site uses cookies to offer you a better browsing experience. By browsing this website, you agree to our use of cookies.