< Back to Blog

March 2021


Luma Bendini

Ingrid HPC is here: get to know the most powerful LES-technology in the market

Newly launched, the solver Ingrid HPC is an important achievement in the pursuit of making High-Fidelity flow simulations widely available across industries. And it is a milestone in Ingrid Cloud’s history. Let’s take a closer look into the features and improved performance of Ingrid HPC.

The new CFD solver, Ingrid HPC, is up and running since October 2020. It is a completely rewritten and modernised version of the underlying methods used in the previous solver. And the results are impressive: Ingrid HPC is running simulations twice as efficiently as its’ predecessor! This modernised solver significantly reduces the necessary turn-around time for a simulation, using fewer core hours, while still retaining the same high accurate results.

Developed over 2 decades of research


FEniCS was co-founded by Ingrid Cloud's co-founder, Johan Hoffman. It is a popular open-source platform for solving partial differential equations (PDEs) by the finite element method. 

Ingrid HPC is the result of almost 20 years of academic research. This research resulted in an innovation which enabled automated mesh optimisation for High-Fidelity simulations. The FEniCS project implemented this method in the form of different open-source components for solving partial differential equations (PDEs). Launched in 2007, Unicorn was the first open-source solver for CFD based on that innovation.

Since then, the technology has been developed both incrementally and in big leaps. In 2012 parallel computing was added to the framework, enabling the use of HPC-resources and thus increasing scalability. Later, in 2018, a modified version of Unicorn HPC was used as an important component when Ingrid Cloud was launched with a unique online interface for High-Fidelity simulations.

This modified version of Unicorn HPC kept evolving to meet Ingrid Cloud’s commercial needs and other industry standards. It was used exclusively by Ingrid Cloud and already offered great improvements in performance when compared to the 2012 code.


The company also made possible that a dedicated team would work on developing an even more robust and stable solver. Thanks to those efforts, in October of 2020, Ingrid HPC was up and running.

What's a High-Fidelity CFD Simulation? usp-idea-def@2x
Read the article here.

The fully rewritten solver has significantly improved performance and robustness. It also marked the shift from using an open source-based solver to using a proprietary solver. This has strategic implications, securing the unique capabilities of the Ingrid Cloud platform. Supported by years of research and development, Ingrid HPC is ready to challenge the industry incumbents and the legacy of low-fidelity CFD.

Faster and more stable 

So, what’s new in Ingrid HPC? We gathered part of Ingrid Cloud’s team to talk about this important achievement and what it represents for the future of time-dependent simulation solvers.


What are the highlights of the new solver? What’s new to Ingrid HPC?

Rodrigo: Many great features were implemented to Ingrid HPC. One feature that I can highlight is that Ingrid now has the capability to detect divergent conditions and restart the solver from a checkpoint with revised parameters. Also, we have cleaner file outputs now, more concise, less files written to disk. Lastly, the new solver enables easier implementation of more complex physics, such as Fluid-Structure Interaction (FSI), variable density flows, etc.

Julian: I’d add that by rewriting the solver we also had to revisit the numerical algorithms, which improved the numerical stability. The new solver is more stable overall, resolving larger time-steps with almost no disruptions.

What is so great about these changes?

Julian:The modernised solver significantly reduces the necessary turn-around time for a simulation and decreases the total core hour usage. And the best part: it maintains or even increases the high accuracy achieved by our adaptive methods based on unique goal-driven error estimation.

How much better is the solver now?

Sebastian: We have observed a performance improvement of at least 2X when comparing Ingrid HPC to Unicorn HPC (2018). That means half of the time to results and half of the computing cost per simulation.  For some cases the performance improvement can, of course, be lower, but for most cases we actually expect it to be even higher.  And in this context, it's interesting to point out that NASA has long ago recognised a shift towards LES-based solvers. With Ingrid HPC we are definitely a step closer to that.

2020 Ingrid HPC
Proprietary solver

  • 2X improved performance and efficiency
  • Significantly improved robustness
  • Improved automation mechanisms
  • Reduced need for file storage
  • Easier implementation of new physics

> 2X improvement

How is that important for Ingrid Cloud’s users?


It makes a powerful LES-technology available for virtually anyone, not only for the CFD community. We believe that this work will in the long term make Ingrid Cloud more accurate than physical wind tunnel experiments in specific use cases, such as for wind simulations in urban areas or vertiport assessment (a vertiport is a small airport for VTOL-aircrafts).

What’s unique about Ingrid HPC?

Rodrigo: It’s actually the same feature that was unique to the former solver, Unicorn HPC (2018): the adaptive mesh refinement technique. But now the new solver Ingrid HPC is bringing the adaptive algorithm to another level by improving performance, efficiency and robustness of the parallel mesh refinement.

What was the biggest challenges when developing Ingrid?

Rodrigo: The biggest challenge was to make sure that the hundreds of thousands of lines of new code actually did what they were supposed to do. Numerical validation and verification of such a comprehensive development project is a hard, difficult work.

Finally, what are the next great things we can expect from Ingrid Cloud?

Rodrigo: We will continue to add new physics – Fluid Structure Interaction is one of the top priorities, because we know Ingrid has a lot to contribute to the FSI problem. And we will continuously improve performance.

Niclas: Also, the new solver runs on the fastest supercomputers in the world and we are preparing it to the new generation of exascale systems.