Nvidia’s role in the EDA industry

Not just GPU-based acceleration: the partnerships announced on occasion of this year’s GTC event demonstrate that Nvidia software, too, is a key technology in several existing or upcoming EDA tools

EDA and engineering software received quite a bit of attention at the recent Nvidia GTC event, with “rockstar” CEO Jensen Huang mentioning this theme during his two-hour long keynote, and with some of the major vendors (namely Ansys, Cadence, Siemens – the German parent company, not Mentor – and Synopsys) issuing GTC-related press releases to announce their extended collaboration with Nvidia. While partnerships between Nvidia and EDA vendors are not new, this level of emphasis from both sides seems unprecedented and deserves a closer look.

The announcements issued by EDA and engineering software vendors highlight three main areas of collaboration with Nvidia: GPU-based acceleration; Omniverse-based visualization; and the use of AI development tools encompassed by the Nvidia “AI Foundry” offering. A fourth area, the only one specifically related to the EDA flow, is optical proximity correction based on Nvidia cuLitho.

Software acceleration based on Nvidia GPUs

In GTC-related announcements, the use of Nvidia GPUs for software acceleration was highlighted by Ansys, Cadence and Synopsys. Ansys harnesses Nvidia H100 Tensor Core GPUs to boost multiple simulation solutions, and prioritizes the new Nvidia Blackwell-based processors and Nvidia Grace Hopper for products across its portfolio – including Ansys Fluent, Ansys LS-Dyna, and its electronics and semiconductor products. As for Cadence, previously announced collaborations with Nvidia include the GPU-optimized Fidelity CFD (computational fluid dynamics) software and the Millennium Enterprise Multiphysics Platform, a hardware box for the acceleration of CFD simulations based on Nvidia GPUs. Synopsys is applying Nvidia accelerated compute architectures, including the GH200 Grace Hopper, across its full EDA stack spanning design, verification, simulation and manufacturing. The tool list includes Synopsys VCS, Synopsys Fusion Compiler, Synopsys PrimeSim, Synopsys Proteus (see below).

Electronic Design Automation is mentioned in the announcement of Nvidia’s new Blackwell chip as one of the industries in which this new hardware platform will unlock breakthroughs. The press release specifically mentions Ansys, Cadence and Synopsys as some of the software makers which will use Blackwell-based processors to accelerate their software for designing and simulating electrical, mechanical and manufacturing systems and parts.

Use of Nvidia Omniverse in digital twins and physics-based simulations

Nvidia Omniverse is a platform of APIs, SDKs, and services that enable developers to easily integrate Universal Scene Description (OpenUSD) and RTX rendering technologies into existing software tools and simulation workflows for building AI systems. It enables the use of generative-AI techniques in visualization and is also available in a cloud-based version. In GCT-related announcements, current or future integration with Omniverse was highlighted by Ansys, Cadence, Siemens AG, and Synopsys. Ansys is using Omniverse in its AVxcelerate Autonomy solution – a simulation toolchain for the development and safety validation of ADAS/AD systems – through integration with Nvidia Drive Sim, a simulation platform for autonomous vehicles. Ansys also plans to investigate additional integrations of Omniverse across its product portfolio, including Ansys STK (digital mission engineering and systems analysis), Ansys LS-DYNA (multiphysics solver), Ansys Fluent (fluid simulation software), and Ansys Perceive EM (a 6G digital twin solution powered by Ansys HFSS, built on Nvidia 6G Research Cloud). Cadence has integrated Nvidia Omniverse in its Reality Digital Twin Platform for the design, simulation and optimization of data centers. Siemens AG will use Nvidia Omniverse in a new product later this year for Teamcenter X, Siemens’ cloud-based Product Lifecycle Management (PLM) software, part of the Siemens Xcelerator platform. It will provide the ability to create photorealistic, real-time, and physics-based digital twins for complex products such as ships. Synopsys will integrate its automotive digital twin solutions with the Nvidia Omniverse platform for the development, testing and safety validation of the car’s software and electronic systems. Omniverse will be used to deliver physically based visualization and simulation of environmental factors. Synopsys expects to begin engaging with lead customers on the solution in the second half of 2024, with general availability expected in 2025.

Use of Nvidia AI Foundry and other Nvidia AI development tools

Introduced in November 2023, the Nvidia AI foundry service pulls together three elements: a collection of Nvidia AI Foundation Models, Nvidia NeMo (a platform for developing custom generative AI), and Nvidia DGX Cloud AI supercomputing services. The goal is to give enterprises an end-to-end solution for creating custom generative AI models. Businesses can then deploy their customized models with Nvidia AI Enterprise software – a cloud-native platform for the development of production-grade co-pilots and other generative AI applications – to power generative AI applications. In GCT-related announcements, the future use of Nvidia AI Foundry, or other Nvidia AI development tools, was highlighted by Ansys and Synopsys. Ansys is examining the adoption of Nvidia AI Foundry to advance LLM (large language model) development, simplifying setup and use of simulation through expert virtual assistance. Additionally, Ansys intends to leverage the Nvidia NeMo platform, which offers a set of tools to develop generative AI capabilities. Lastly, Ansys is also exploring Nvidia Modulus – an open-source framework for building, training, and fine-tuning physics-ML models with a Python interface – for physics-based machine learning, to further boost its Ansys AI+ product family. Synopsys is extending its Synopsys.ai LLM-based capabilities, beginning with Synopsys.ai Copilot, to support Nvidia AI and compute platforms. Synopsys will utilize the Nvidia AI Enterprise software platform, which includes Nvidia NeMo framework, and Nvidia NIM inference and NeMo Retriever microservices deployment containers. Synopsys customers will be able to deploy Synopsys.ai Copilot on air-gapped on-prem environments, leveraging the accelerated computing performance of Nvidia DGX systems.

Use of Nvidia cuLitho

Announced one year ago, Nvidia cuLitho is a library with optimized tools and algorithms for GPU-accelerating computational lithography by orders of magnitude over current CPU-based methods. In GTC-related announcements, Synopsys reiterated that cuLitho is integrated in its Proteus tool, and Nvidia revealed that TSMC and Synopsys are now going into production with cuLitho. In addition to that, Nvidia also announced that it has developed algorithms to apply generative AI to further enhance the value of the cuLitho platform. The new generative AI workflow delivers an additional 2x speedup on top of the accelerated processes enabled through cuLitho. The application of generative AI enables creation of a near-perfect inverse mask or inverse solution to account for diffraction of light. The final mask is then derived by traditional and physically rigorous methods, speeding up the overall optical proximity correction process by a factor of two.

Differences among the “big three” EDA vendors in their relationship with Nvidia

Comparing the GTC-related announcements issued – or not issued – by the three major EDA vendors, some differences are quite evident. The most noticeable difference is that Mentor Graphics (Siemens EDA) did not issue any announcement related to the GTC event. Another difference concerns the use of Nvidia AI development tools: based on its GTC announcement, at least, it looks like Cadence prefers to rely on its own internal resources for developing AI-powered solutions. Overall, the EDA vendor with the strongest Nvidia ties seems to be the Synopsys-Ansys combination, which uses/will use Jensen Huang’s products for GPU-based acceleration, Omniverse-based visualization, the development of AI-powered tools, and its computational lithography tool.

Clearly, there can be a variety of reasons for these differences among the three major EDA vendors when it comes to their respective relationship with Nvidia. For example, one reason could be a different perception of what is a given company’s “core business” and what is not: core business requires differentiated solutions, while for non-core products it’s better to avoid reinventing the wheel. If such an explanation makes sense, then the appeal of Nvidia Omniverse for digital twins could mean that physics-based visualization tends to be perceived as a non-core technology in EDA – even though multiphysics simulation is increasingly important in the chiplet era. As for artificial intelligence, this technology has become essential in EDA so it’s no surprise that different EDA vendors have different approaches to it – in terms of using or not using Nvidia AI development tools.

EDA in the era of Nvidia empire

What impact will have an ever more powerful Nvidia on the EDA industry? Nvidia is undoubtedly the most significant incarnation of the AI positive feedback loop: powerful chips enable AI, and AI enables even more powerful chips. This year’s GTC highlighted that Nvidia’s essential role in this virtuous loop is twofold: on the hardware side, it provides raw GPU computing power for running AI-based EDA tools; and on the software side it is also instrumental in the development of EDA tools themselves, through its software offering. Is this tight relationship good or bad for the EDA industry? On the one hand, coupling the EDA car to the Nvidia locomotive – a relentless innovation engine – contributes to ensure that the EDA performance keeps up with the requirements of the most advanced chips. On the other hand, a superpower neighbor could potentially be a worrying presence. EDA is different from the other software industries as it “borders” to chipmaking, so to speak, which is Nvidia business. In an era of once unthinkable vertical integrations – with hyperscalers now developing their own chips – should EDA vendors start feeling “surrounded” by Nvidia on multiple fronts? Nvidia already provides key AI development software (for an increasingly AI-powered EDA toolchain), key visualization software (for an increasingly important digital twin offering), and a key OPC tool. As of today, however, Nvidia doesn’t seem to represent a threat to EDA incumbents. In his GTC keynote, Jensen Huang described Nvidia as an “AI foundry” and used TSMC as an example of what a foundry does: that is, serving its customers, without competing against them.

This entry was posted on Thursday, March 21st, 2024 at 5:21 pm. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.