HyperLynx can automate a DDR5 eye check. Who owns the pass?

Illustrative Ethernet eye diagram. Andrew D. Zonenberg / Wikimedia Commons · CC BY 4.0
Siemens’ HyperLynx Agentic Toolkit is designed to let an engineer describe a verification job in natural language, then assigns the work to agents for areas such as DDR, SerDes and power integrity. The existing HyperLynx solvers still do the simulation. What changes is who sets up each run, retrieves the results and checks them against the specified margin.
Take the DDR5 eye-mask example Siemens shows. If a layout change leaves a narrow margin, it is useful to rerun several routing variants and see which one recovers it. That is exactly the kind of repetitive work an agent ought to handle. The awkward bit comes after a pass: an obsolete stack-up or an incorrect package model can give a clean eye for a board that will never be built. A faster sweep cannot repair either assumption.
The most valuable feature here would be a record another engineer can reproduce: which models the agent used, which corners it ran and which mask it applied. Siemens shows the simulation and the mask, but its announcement doesn’t establish how much time the workflow saves or whether it catches a missed condition.
A fair test of the toolkit would give the manual and automated runs the same board revision, models and corners, then compare the time taken to reproduce a marginal result and explain why it passed. That is a harder benchmark than the number of simulations run overnight. If the agent saves set-up time but someone still has to reconstruct what it actually checked, some of the work has simply moved to sign-off.
Early access is planned for selected customers in the fourth quarter of 2026, with general availability to follow.
Sources: Siemens product announcement



