By Steve Robbins
I’ve been traveling a lot recently, first to the Siemens CAE Symposium at the Joe Gibbs NASCAR Racing facility outside of Charlotte, NC, then off to Supercomputing 2011 in Seattle, WA, and finally to Las Vegas to attend Autodesk University. They say travel broadens the mind, and my trips were no exception to that rule.
I had the pleasure of meeting with engineers whose ability to create great designs has been positively changed by the adoption of new technologies. I saw the results of companies using new modeling tools, simulation software, PLM and engineering data management, rapid prototyping and 3D printing, all enabled by faster computer technologies. I’m lucky to be an editor at DE, because it allows me to see all that is happening in the design engineering world.
Many of you experience only the tools your companies or organizations use. You have seen new technologies that might be of great benefit to you and our team, but often, it takes years to adopt what has been available for decades. If engineers were left to their own devices, most would be early adopters of new technology, but often there is a layer of management asking you to justify an investment in new technologies. Well, the world is about to change. We are at an inflection point.
Change is Coming Fast
History has shown us that technology increases in performance as it decreases in cost. Think VCRs to DVDs and now, to streaming video; personal computers to workstations to local high-performance computing and the cloud. At Supercomputing 2011, Jen-Hsun Huang, NVIDIA’s CEO, discussed where we are today and where computing will be going in the future. The world’s fastest supercomputer is currently running at 10.51 Petaflop/s . That’s 10 quadrillion calculations per second. However, Huang said the power requirements of these machines is tremendous, with a 6-petaflop/s computer using about 9 megawatts of power. So how are supercomputer companies going to power the next generation of computers? These exaflop computers will need about as much power as a full-sized aircraft carrier if they use conventional CPUs. The A4W Reactor used by the U.S. Navy produces 104 megawatts. Most aircraft carriers have two of these. We’ll need more efficient floating point processors to reach exaflop computing levels.
Huang predicted reaching exascale computing somewhere between 2019 and 2022, and the goal is to reach this scale using about 20 megawatts of power. That still sounds like a lot, but if we can reach that for exaflop/s computing, a 5-watt device like a smart phone or tablet will have the power to run a few teraflop/s of compute power and today’s workstations, with 1,000 watts, will run at hundreds of teraflop/s. This translates into real-time multiphysics simulations on your desktop in about 10 years.
Speed Delivers on Technology’s Promise
Meanwhile, as software has been eyeing the cloud, our workstations and local HPC systems are getting much more powerful. CAE, simulation and analysis software are becoming increasingly easier to use while becoming exponentially more powerful. Being able to send models out to the cloud to be rendered, or sending a CFD simulation job to the cloud to run a large number of times works in certain situations, but keeping things local would be more efficient. In the next couple of years, we will see a dividing line in our work, with more compute power on our desks and access to very large amounts of data and compute power on the cloud.
Sharing and collaboration will become increasingly important and will produce unexpected results. Security will come to the forefront. PLM and data management will be able to deliver the promise of ease of use, data reuse, and sharing to the complete enterprise. It will be used by everyone in engineering, small or large companies, designers, manufacturing engineers, sales people and management.
Direct modeling and one-button analysis will be the norm during the creative processes. Running multiple simulations on large assemblies locally before prototyping the complete design using many different materials will change the way we work. And simulation isn’t limited to the mechanical model, it will include motors, electronics, and control systems. We will be able to determine manufacturability before handing the design off to the factory floor.
In 2012 we will see many changes. It will be a year of exponentially moving forward as software meets the challenge of using increasingly powerful hardware technologies now available. CAD/CAM/CAE is going where no engineer has gone before.
Steve Robbins is the CEO of Level 5 Communications and executive editor of DE. Send comments about this subject to DE-Editors@deskeng.com.