Synopsys has announced the launch of Ultra Ethernet IP and UALink IP solutions that are designed to meet the demands required to scale AI architectures.
“For more than 25 years, Synopsys has been at the forefront of providing best-in-class IP solutions that enable designers to accelerate the integration of standards-based functionality,” said Neeraj Paliwal, senior vice president of IP product management at Synopsys. “With the industry’s first Ultra Ethernet and UALink IP, companies can get a head start on developing a new generation of high-performance chips and systems with broad interoperability to scale future AI and HPC infrastructure.”
Ultra Ethernet can support up to one million endpoints on a single network, and UALink allows AI clusters to scale up to 1,024 accelerators.
The Ultra Ethernet solution includes a MAC and PCS controller, PHY, and verification IP. The MAC and PCS controller IP offers up to 1.6 Tbps of bandwidth, and supports an interface to higher layers of the Ultra Ethernet stack, providing customers with a full silicon implementation for switches, AI accelerators, and smart NICs. Additionally, the Ultra Ethernet verification IP helps ensure that protocols are adhering to evolving standards.
“Built on silicon-proven technology, decades of experience, and compliant with evolving Ultra Ethernet standards, designers can reduce risk and accelerate time to market for the world’s most advanced SoCs,” Synopsys’ website states.
Synopsys’ UALink IP solution includes a PHY, controller, and verification IP, and is optimized for data-intensive AI workloads. The PHY IP provides 200 Gbps per land, and the Controller IP mitigates AI hardware bottlenecks by allowing accelerators to share memory. The verification IP also provides fast and reliable verification of AI hardware.
“Built on decades of design experience in high-speed PCIe and Ethernet IP, Synopsys’ UALink IP reduces integration risk and speeds time to market for AI and HPC chip designers unleashing next-gen AI infrastructures,” Synopsys’ website states.