Semiconductor Engineering: Designing Chips That Can Explain Themselves
6.17.2026
Originally published on Semiconductor Engineering by Ann Mutschler | June 17, 2026.
Semiconductor Engineering continues its Experts At The Table roundtable on on-chip data analytics and resilience, with Satish Radhakrishnan, head of GTM at Vinci, joining panelists from Arteris, Axiomise, Baya Systems, Cadence, Keysight EDA, Movellus, Siemens EDA, and Synopsys. This installment focuses on where monitor data goes, how it gets analyzed, and whether the area cost of on-die visibility is justified.
The panel works through the path telemetry takes from on-die monitors to firmware to fleet-level analytics, and the tradeoffs at each hop. Radhakrishnan frames the requirement as a system that evaluates incoming data immediately and decides whether to act, rather than storing large volumes of it — an approach he describes as resembling a digital twin that runs in real time, predicts, and intervenes only when there is an issue. The core constraint, he notes, is that the inputs have to be processed extremely quickly for thermal problems, reliability concerns, or abnormal current spikes to be caught early enough for predictive action.
On the question of area cost for monitors and sensors, Radhakrishnan reframes it as a question of what the allocated space buys you: if the design can be simulated accurately as a digital twin with full detail included, the value of complete data and visibility at every location rises substantially. Other panelists connect the same thread to guard banding — Keysight EDA’s Pedro Merlo notes that on-die monitoring lets teams thin margins that would otherwise be set conservatively, and that the resulting field data feeds back to make the digital twin better over time.
"It operates in real-time, makes predictions, and only responds if there's an issue."
Key takeaways from the coverage:
Vinci’s Satish Radhakrishnan describes real-time evaluation of incoming data as a digital-twin approach that predicts and intervenes rather than accumulating stored telemetry.
Early detection of thermal issues, reliability concerns, and current spikes depends on inputs being processed at very high speed.
Accurate, fully detailed simulation raises the return on on-die visibility investment by making comprehensive data actionable.
Semiconductor Engineering reports panel agreement that on-die telemetry lets teams replace worst-case margin with measured silicon behavior, improving power, performance, and area without sacrificing resilience.
How does this relate to Vinci's platform?
The roundtable’s flywheel — field data improving the digital twin, and a better twin producing better designs — depends on the twin being accurate at the resolution where physical failures actually originate. That is the gap Vinci’s foundation model addresses: deterministic thermal and thermo-mechanical prediction on native, full-resolution geometry rather than simplified proxies.
For architects deciding how much margin to design in, that matters directly. Guard bands exist to cover physics the design flow cannot predict precisely. When thermal and mechanical behavior can be computed continuously and at manufacturing resolution, more of that uncertainty becomes measurable, and the tradeoff between conservatism and performance becomes a calculation instead of an assumption.
About Vinci
Vinci is a frontier lab building the foundation model for the physical world. Its deterministic, solver-grounded systems make physics continuously computable inside production engineering workflows and are already running on flagship programs, shifting physics from an episodic simulation bottleneck to continuous infrastructure for design, manufacturing, and reliability decisions.
Shift Left on Warpage: Thermo-Mechanical Analysis Before the Design Locks
September 16, 10am PDT
Stop discovering package warpage after your design choices are locked. Join our live demo to see how one semiconductor team cut a 16-hour thermal workflow to about an hour.
