By Steve Robbins
Two years ago at SC07, I got all jazzed up about the speed and accessibility of multicore computing. I spent three days talking with companies about how their hardware enables engineers to leverage CAE to explore complex analyses and improve their company’s competitive advantage. Many manufacturers acknowledged simulation as a market driver, but seemed more interested in large data center applications for financial modeling. The world has changed.
At the recent HP CAE Symposium in Detroit, CAE had become the center of the universe for many of these same companies. Why? Because every simulation is a building block leading to more complex analyses that need faster, more powerful computers. For two, accessibility, scalability, and lower costs mean more engineers have access to powerful computers.
Cluster computing systems, for example, are more accessible because of ease of use. It might have taken weeks or months to set up and validate a cluster a couple of years ago, now it takes a few days. Accessibility means a cluster in a box, or you can lease the hardware. You can even rent time on a service provider to run simulations.
In a couple of years computing technology will be even more efficient and accessible. Moore’s Law applies; to more than just transistors. InfiniBand increases interserver speed by up to 2.5 times. Solid-state drives can increase performance by 25%. CPU architectures continue to increase speed via more cores, and intelligent processing, communications, and memory advances. And the algorithms at the heart of CAE evolve at similar rates.
All this explains why simulation is moving to the front of the design. Engineers are a skeptical lot. When simulation software first hit the scene, the cutting edge of engineering used it for validating real-world testing. Soon the results became more accurate, and it became a way to minimize the number of physical prototypes needed for testing.
Still, simulations could take weeks to complete and were complicated to set up. But now high-performance computing is more accessible, affordable, scalable, and easier to use. And easy-to-use simulation software takes advantage of multithreaded, multicore processing. The effect has been to reduce run times for increasingly complex analyses from weeks to minutes.
Today, you can run a multiphysics simulation on a multicore workstation and get results in a reasonable timeframe. If you need more processing power, you grab what you need off the cluster or some internal or external source. You can create simulations on direct models before you get into parametric designs, saving hours translating and simplifying CAD files, then analyzing and interpreting your results for the CAD expert. This in turn saves time bringing designs to market, produces more manufacturable designs, saves on material costs, and increases your ROI.
So what’s in the future? Model-based designs that are simulations of simulations. Drug companies now have a simulated model of the human body that’s used to simulate the effect of drug therapies. Aerospace companies simulating the entire engine on a simulation of the complete aircraft. The list knows no bounds.
Design and simulation are vital to remaining competitive. The ability to run multiple simulations of a complex design simultaneously provide engineers insights that weren’t available previously. Faster and faster computing has enabled faster, more complex simulations that crave more computing power. It’s a feedback loop that’s working for everyone’s benefit and changing the world of engineering.
Steve Robbins is the CEO of Level 5 Communications and executive editor of DE. Send comments about this subject to DE-Editors@deskeng.com.