Product Design Strengthens HPC Ties
Increasingly, advanced computing technologies are embraced for their ability to bolster economic competitiveness. The advantages owed to a well-thought out and well-funded HPC strategy are embodied in the adage: to outcompute is to outcompete. One realm where the benefits of going digital have been especially prominent is product design.
Over the last decade or so, there has been a concerted effort in the United States and abroad to bring digital manufacturing tools into the hands of manufacturers who have traditionally been underserved. Often the companies, small to medium-sized shops, who would gain the most from these tools can least afford them. Thus, programs and technologies that lower the barrier to entry are sorely-needed as are campaigns that raise awareness of these benefits, thus helping to overcome cultural resistance associated with navigating the digital divide.
A recent article from London-based publishing house Raconteur Media makes it clear just how transformative this technology can be.
“While supercomputers aren’t mandatory for simulating the workings of products under development, there’s no doubt that the combination of high-powered computers and advanced visualisation technology is rapidly transforming the use of 3D product simulation and visualisation within product lifecycle management (PLM) and product design generally,” the author writes.
Swansea-based medical supply company Calon Cardio-Technology is focused on developing smaller and more efficient blood pumps for use as an alternative to heart transplants in patients with chronic cardiac failure. Designing these artificial hearts so that they are both safe and effective requires a thorough understanding of how blood flows through the pump. This is a sophisticated computational fluid dynamics problem that is beyond the scope of most desktop computers.
Calon researchers are carrying out their design work with the help of Swansea University’s Advanced Sustainable Manufacturing Technologies centre. New designs are simulated on a supercomputer cluster managed by HPC Wales. Using a well-equipped desktop computer, each 3D simulation, requiring a “mesh” of approximately two million elements, would take two to three days to complete, but the supercomputer shortens that time significantly.
“We need to model about a dozen scenarios for each pump design, after which we tweak and refine it,” explains Calon’s chief technology officer Graham Foster. “With a supercomputer, a scenario takes two to three hours, not two to three days, and we can send the dozen scenarios as a single batch. We’re saving time and also cost.”
Despite the competitive advantages conferred by HPC and a number of programs aimed at opening up access, the technology still remains out of reach to many “ordinary businesses with ordinary computers.”
This is problematic because as Peter Vincent, a lecturer in aerospace aeronautics at Imperial College London, points out, there are “a class of computational fluid dynamics problems that are currently intractable at an industrial level.” He adds that “they can be solved by supercomputers in national laboratories, but not at an industrial level.”
Hybrid computing using accelerators (NVIDIA GPUs) or coprocessors (Intel Phi) may offer one path to these much-needed FLOPS without breaking the bank. GPU maker NVIDIA is working with Imperial College to super-charge desktop computers with NVIDIA’s latest-generation Tesla parts. Each of these chips has about 2,500 compute “cores” while most desktop computers have perhaps two or four. The UK’s most powerful GPU-powered supercomputer, Emerald, has 372 Tesla M2090 GPUs.
“With the right computer algorithm and the right kind of simulation problem, there can be a five-to-tenfold increase in speed for a given expenditure on computing power,” observes Dr. Vincent on the topic of GPU-accelerated computing. “That can bring many more problems within reach.”