Addressing Advanced Node Challenges with PrimeTime®: An Interview with Feroze Taraporevala

Q: Hi Feroze. Firstly, thank you for taking the time today. Please help outline how do you see the design and timing market evolving? And what new challenges do you see at advanced nodes?

Feroze Taraporevala

Well, this is a very exciting time for the semiconductor industry. With the increased level of design activity, the overall complexity has been increasing on multiple fronts.

On one hand, design size is becoming larger and larger, and the physics of the new process nodes are getting more and more complex. Also, there's been a ramp-up in the design methodology complexity. The market, on the other hand, is driving growth in new vertical segments, with AI, automotive, cloud, and so forth.

That brings up new challenges, because timing analysis can take up to 50% or more of compute time as designers try to close their ultra-large, high performance, advanced node designs. And our R&D team has focused on all these different areas to help designers manage their increasing challenges. There's a focus on scalable STA for runtime and memory, there's a focus on innovating to keep up with the new effects and the new process changes that are occurring, and also, to support designers and their new methodologies. So, a lot of exciting stuff is happening.

Q: Can you please share your vision on PrimeTime’s position to address these challenges and what role it plays in building confidence in the design flow?

Feroze Taraporevala

Absolutely. Let us take these challenges one by one. Let's start with the increasing design sizes. We have a high focus on what I call scalable static timing analysis. And here, we have been on a continuous tear of improving the tool runtime and memory, with 2X improvements in runtime and memory for the past two years.

There's also a focus on handling designs in a hierarchical fashion with our hyperscale technology, which has been used across the globe, across design implementation and also, in driving the last mile of signoff.

We have innovations in tackling the advanced node physical effects. There's a lot going on with process variation and shifts on the process side, Vt mistracking, wire mistracking, and so forth. So, we have been innovating to handle these effects. Process variation just gets more and more complex and nonlinear, so we have introduced technologies to manage all of that.

And finally, I'm pretty excited about helping designers make their designs more robust with things like voltage drop and process variations using our new product PrimeShield. These are just a few of the different ways in which we are helping manage and address the challenges that designers are facing.

Q: Can you further elaborate on the recent improvements of the tool to handle the ever-increasing performance and productivity requirements?

Feroze Taraporevala

Sure. As I mentioned before, our focus on managing the compute of static timing analysis is absolutely critical. We have been focused on making runtime improvements across the flow. And in particular areas such as AI with their associated newer design architectures, there's a lot of structural replication going on. We have performance improvements to leverage the replication and effectively handle these new structures.

Recently, we introduced some innovative ways of enhancing capacity with up to 50% improvement in memory. So, you can fit larger designs in a smaller memory footprint. In addition to the improvements we've made in flat timing analysis, we also have a development drive towards tiling designs and breaking them down hierarchically into parts with our HyperScale technology.

We also have a focus on reducing scenarios with the simultaneous multi-voltage analysis to handle cross-domain voltage. We have a focus on driving up the scalability of our solution so that we can run on a larger number of cores more efficiently. And last but not least, there's been a focus on path-based analysis where a lot of runtime is spent, because folks want to get a more accurate timing picture of the design, and they employ path-based analysis for that. We've made a lot of innovations in that area with HyperTrace and ML techniques providing more than 10X speed-up in runtime.

Q: How are new market segments driving the direction? And what are you doing to serve these new customer requirements?

Feroze Taraporevala

That's a great question. Actually, you know, there’s been a lot of activity in the AI space,  in the automotive space, high-performance computing, and even IoT applications. Each of them has a different footprint, in terms of the types of technology that we have to address those applications. Recently, there's also been a push towards 3D IC design, which promises a great way to keep up with Moore's law.

We have a lot of investment in all of these different areas. For example, for high performance computing applications, which tend to be at the higher voltages, we have methods to manage the back-end-of-the-line variability, where interconnect variation matters and the higher metal layers are employed for higher clock speed paths. On the other side, for IoT or other applications where we're driving the voltage lower to manage power, we have a lot of investment in the area of handling process variability at lower voltages. And these are just a few examples of the types of things we are doing to drive and help designers in these new market segments.

Q: Cloud is an exciting area, and the ecosystem is growing, how is PrimeTime enabling cloud-based design?

Feroze Taraporevala

Indeed, cloud is a very exciting area. It used to be the case where our customers would have their compute farms in-house on premises. But they are finding when driving their designs that compute demand is tricky to predict and manage, and they need to be able to leverage the cloud to manage that peak. Some other design houses just want to stay on the cloud, they don't want to own the compute.

PrimeTime is already deployed on the cloud by our customers. And we continually look for ways and opportunities to improve the user experience to drive further optimization on the cloud. Another aspect of cloud is, its robustness. Driving robustness in the cloud is an important area, and we are continually investing in all of these areas.

Q: Thank you for the overview, Feroze. It was an illuminating discussion.

Feroze Taraporevala

You are welcome! We are excited about Synopsys PrimeTime static timing analysis direction and look forward to helping our customers meet their needs for enhanced accuracy, performance, and productivity.