CPU, Accelerators, and Switch vendors depend on Synopsys and others to provide the technology-certified IP blocks for MACs, Phys and verification tools. Artificial Intelligence, especially generative AI, has pushed the performance boundaries of computation, memory, and networking. Some vendors like Nvidia and AMD have developed their own networking between accelerators, and others like Intel use industry-standard Ethernet for simplicity at scale. Now Synopsys is enabling standard Ethernet networking to run at the speeds needed for AI with the first IP offering that deliver twice the bandwidth of today’s Ethernet at half the power to meet the growing demand for scaling AI. intellectual property sales rose to close to 25% of Synopsys’ business last quarter.

The CPUs, Accelerators, and Switches in today’s data centers depend on ethernet providers like Synopsys for the IP to build their chips. SYNOPSYS
Synopsys claims it 1,6T Ethernet is a complete IP solution that chip designers can build upon, and already has clients lined up to use this new platform. Each PHY, which translates the digital data to the analog signals, runs at 224Gb, and each controller IP can crank out 1.6 Tb of bandwidth, with a 40% reduction in latency and 50% reduction in chip area. That area is especially important for those companies that are building AI accelerators with built-in networking, which adds up to a lot of cost when chips contain dozens of Ethernet controllers.

Synopsys has rolled out the first complete 1.6T Ethernet solution SYNOPSYS
The IP and verification tools are available now, even though the Ethernet 1.6T standard will not be published as finalized for a couple of years. SoC designers can build on this platform now. And since 60% of data center infrastructure power involves Ethernet connections, this solution should find significant adoption for AI-centric workloads.

The Synopsys 1.6T solution SYNOPSYS
Conclusions
AI demands a lot of processing, fast memory, and very fast networking at low latency. Leading-edge adopters won’t wait for state of the art networking to materialize and become available from EDA and IP vendors, and like Nvidia and AMD; they develop it themselves. Hence NVLink and Infinity Fabric. Now the rest of the chip world has access to a complete IP solution they can depend on, so they can focus on whatever magic sauce they need to grab some the exploding market.