A case study on adapting memory compilers and logic libraries for power-critical optical networking chips.
By Daryl Seitzer, Andrew Appleby, and Mohammad Tanveer
Building a new system-on-chip (SoC) starts with assembling the right foundational elements—pre‑verified IP for logic, memory, I/O, and other essential functions.
Standard IP solutions typically address most common design needs, but some projects call for more specialized approaches, especially when innovation is critical or when the power, performance, and area (PPA) targets are particularly aggressive.
As a result, the ability to customize IP is becoming increasingly valuable.
This is where Synopsys sets itself apart. Beyond offering the industry’s most extensive portfolio of silicon‑proven IP, we partner closely with customers to adapt and optimize our Foundation IP solutions to overcome unique design challenges.
One of our customers recently approached us with a unique situation and set of requirements. They wanted to develop a new chip to enable the optical network infrastructure supporting the surge in demand for Edge AI. The chip needed to perform reliably at just 0.4 volts and they couldn’t find IP solutions capable of supporting such low voltage. They also had an extremely aggressive, eight-month delivery timeline.
Always up for a challenge and backed by the industry’s most comprehensive, silicon-proven, application-enhanced, and adaptable portfolio of Foundation IP, our team quickly jumped into action.
Following a thorough evaluation of the customer’s PPA targets, it was clear the low-voltage requirements of the design would prove challenging for both memory compilers and logic libraries.
Memory bit cells — tiny storage elements within a chip — are typically unreliable at 0.5 volts or less. To solve this problem, our team developed a new memory compiler architecture. They refined the bit cell design, adjusted peripheral circuitry, and used advanced low-leakage transistors to ensure stability at lower voltages.
For other memory types, our team enhanced dual rail support — preferred for this technology node — which allowed different parts of the chip to operate at different voltages. Memory arrays, for example, typically need a higher minimum operating voltage than logic circuits. Our dual rail support enhancements ensured the logic circuits would not be restricted by memory voltage requirements. This flexibility helped minimize power across the entire chip.
In addition to enabling low-voltage operation, our team helped minimize the silicon area of the design while maximizing power efficiency. This included:
Together, these enhancements enabled every part of the chip to consume less power and occupy less space — crucial benefits for the cost- and power-constrained systems the chips will support.
Once the customer’s efficiency goals were met, our team developed a silicon test chip to validate the custom memory compiler and ensure real-world robustness. Comprehensive stress tests and benchmarks confirmed performance and stability, helping address issues early and build confidence in long-term reliability.
The design not only achieved all PPA targets — it was also completed within the customer’s demanding eight-month project timeline. The chip is currently pending tapeout, with production anticipated to follow.
Customizing IP modules is inherently complex—particularly when teams are striving to meet stringent PPA objectives under tight schedules. It demands deep collaboration, engineering adaptability, and the readiness to solve difficult technical problems—commitments that many IP providers are not equipped or willing to make.
Synopsys is different.
With a comprehensive and adaptable Foundation IP portfolio, backed by world‑class R&D and a highly engaged team of field engineers, we’re uniquely positioned to support customers as they navigate intricate design challenges and push toward their most aggressive PPA targets. Learn more: Synopsys Foundation IP.
Andrew Appleby is a principal product marketing manager for logic library IP at Synopsys.
Mohammad Tanveer is a principal applications engineer for embedded memory IP at Synopsys.
Leave a Reply