Engineering’s Compute Power Appetite is Insatiable

Chad Jackson

For years, engineers needed more and more compute power. More memory and compute power has been needed to open and manipulate huge 3D assemblies in CAD applications. More compute power has been required to solve increasingly large finite element analysis (FEA) models with thousands, or even millions, of degrees of freedom. Over the span of decades, the need for more compute power continuously grows. That, of course, has been driven by software providers exploring the very edge of what can be done.

Today the need for more compute power is no different: More is needed. However, engineering organizations are facing some new questions: Do you put more power in the desktop or buy more capability in the cloud? In some cases, it is not an either-or type of answer; it’s both. To really figure out the right answer for your organization, figure out your software capability needs. Let’s run down the options.

Automated Simulation

Simulation-driven design, the idea of conducting analyses to make more informed design decisions, has been a longstanding vision in engineering organizations. In the past few years, a new technological paradigm has been developed to enable that pursuit. The idea is to create a “templatized” simulation that only needs a few inputs. Non-expert users can plug their designs into that template, tagging the right inputs for their situation, and press the “go” button. Some time later, the simulation is complete and the results can be reviewed. This process can be repeated, and even automated, to assess many different designs as a single batch.

From a compute perspective, heavy use of this approach will increase demand for more powerful computing resources. However, automated simulation capabilities can come in simulation applications that run on the desktop, can leverage compute farms or can operate in browser-based apps that run in the cloud. Investing in more powerful desktops or compute farms comes as a one-time capital expense. Supporting more compute power for cloud apps comes as an ongoing operational expense.

Instant Simulation Results

An intriguing new simulation capability has emerged in some analysis software applications: instant simulation results. The concept is that as soon as a non-expert identifies or supplies the minimum information needed to run the simulation, the software starts showing the lowest resolution result. As the software continues to refine the results, it is shown as an animation on the screen. As geometry is changed, removed or added to the design, the results update. It allows users to see the impact of changes almost in real time. This approach provides some instant gratification for engineers who want to play what-if games with their design.

Although this new approach leverages some significant advances in analysis software, it also has some unique computing requirements. It uses graphics processing units in workstations to run dramatically more iterations of simulations.

Big Data in the IoT Era

No discussion on modern product development would be complete without the inclusion of the internet of things (IoT). Streaming data from sensors and software off of products allows for more insight into how products operate. That, in turn, provides engineers with information to design better offerings in the future. It also allows them to feed that data back into simulation models, effectively creating a digital twin of the physical product.

The IoT carries some unusual compute demands. The data that streams off products can pile up quickly. Soon enough, organizations find themselves with big data, a set of information so large it is hard to even host it, much less analyze it. Desktop resources need to be dramatically improved if local software applications are used. Alternatively, some organizations employ cloud-based analytics tools that reside next to the data. Organizations must plan carefully when considering these options, as many frequently underestimate their needs.

Takeaways

Engineering will always need more compute power. However, the fundamental questions about computer resources are changing: Do you invest in desktop-based or cloud-based assets? The key to making the right choice is awareness of the tools the engineering organization is planning to use. With that, you have the right context to make the right choice.

More Info

Lifecycle Insights