• Subscribe

Supercomputing from the bargain bin

Speed read
  • Complex modeling and simulations usually require supercomputers
  • Supercomputers are expensive and only available to a fraction of researchers
  • Using GPUs as cheaper alternative advances research more quickly

Powerful graphics cards, also known as graphic processing units (GPUs), are usually used to create ultrafast gameplay and realistic visuals for games consoles, personal computers, and laptops. But recently, the GPU has emerged as a technology to accelerate scientific simulations, running some applications over 100 times faster than conventional computers.

<strong>Violent flows</strong> such as ocean waves overtopping a ship's deck require lots of computing power to simulate due to their complexity. Courtesy NOAA. Using this technology, Alex Chow, from the School of Mechanical, Aerospace and Civil Engineering at the University of Manchester, is now creating largescale simulations of ‘violent fluid flows’ including powerful ocean waves crashing against offshore wind turbines to predict their potential impact forces on the structures.

“Traditional computational fluid dynamics (CFD) methods have difficulties dealing with the flow of free surfaces,” says Chow. “Anything where you’ve got a body of water and some of it breaks off, that’s very difficult to model with the conventional CFD methods.”

Creating these complex and accurate computer simulations is usually performed on a supercomputer. Rather than being an individual machine, a supercomputer is actually made up of hundreds of central processing units (CPUs) connected with up to thousands of computing cores.

The models of violent flows that Chow works with have millions of data points, and require the solving of millions of simultaneous equations with millions of variables. And each of these matrices must be solved thousands of times in a single simulation.

“And it has to be done quickly because you don’t want engineers to be waiting several days or weeks for you to design something,” says Chow.

<strong>Traditional supercomputers</strong> are expensive, use a lot of energy, and are available to only a fraction of the world's researchers. But Chow's software uses inexpensive GPUs to run complex simulations. Courtesy Argonne National Laboratory.But supercomputers are very expensive, with even small clusters ranging from hundreds of thousands to millions of dollars. They use large amounts of energy and are only accessible to a small number of researchers and scientists.

The benefit of using GPUs is that they’re much cheaper and more energy efficient compared to supercomputers. Some GPUs are compact enough to fit in a laptop whereas supercomputers may require a whole room or dedicated facility.

A GPU, a very good one, costs less than $1,000, says Chow. That accessibility factor means that engineers can run simulations without having access to a supercomputer. Multiple researchers could even run different scenarios simultaneously, thanks to the low cost and portability of the devices.

Sharing the knowledge

To that end, Chow has developed new software from the open-source code “DualSPHysics” for the scientific simulation of Incompressible Smoothed Particle Hydrodynamics (ISPH) to run on a GPU for simulating complex, violent hydrodynamic flows.

Simulation of waves propagating towards a cylindrical structure analagous to a wind turbine column. The largest wave breaks just before it hits the structure. Courtesy Alex Chow.

The new code is capable of computing millions of data points on a single device for real 3D engineering applications. According to Chow, using affordable technology like GPUs reduces the costs of complex scientific simulations from hundreds of thousands of pounds to just a couple of thousand.

Most researchers and small engineering companies are able to afford a relatively powerful laptop or computer with a quality GPU, making this kind of simulation and research even more accessible.

“The project is all based on open-source code, so anyone can download it, anyone can use it,” says Chow. “That’s what we aim for—the ease of re-usable research, to accelerate and allow the field to develop faster.”

Cheaper computing, greener energy

“My main motivation is offshore engineering and helping us move towards this renewable energy movement,” says Chow.

<strong>Chow's research</strong> will help engineers safely maneuver and construct the giant turbines of offshore wind farms in waters as great as 60m deep. Courtesy SteKreuBe. <a href='https://creativecommons.org/licenses/by-sa/3.0/deed.en'>(CC BY-SA 3.0)</a>The UK, for example, generates more electricity from offshore wind than any other country in the world with around 5% of annual electrical energy coming from the sector. This is expected to grow to 10% by 2020 and is also growing fast on the global level.

But shipping a wind turbine out to sea is a risky business. The industry is moving towards ever-larger turbines with columns as much as 200 meters high that need to stand in water depths of at least 60 meters.

ISPH modeling allows engineers to design these turbines to withstand tsunamis and extreme waves of up to 30 meters high.

“The amount of energy produced from offshore environments is increasing as the world tries to meet energy targets, but the ocean environment can be very violent and harsh,” says Chow.

“Efficiently designing structures for these environments is a difficult task. Using physical experiments can be extremely impractical and not representative of the problem. These simulations allow engineers and researchers to make important decisions about the design of a structure without having to invest in site visits and costly experiments.”

Read the original article on the University of Manchester's website here.

Join the conversation

Do you have story ideas or something to contribute? Let us know!

Copyright © 2018 Science Node ™  |  Privacy Notice  |  Sitemap

Disclaimer: While Science Node ™ does its best to provide complete and up-to-date information, it does not warrant that the information is error-free and disclaims all liability with respect to results from the use of the information.


We encourage you to republish this article online and in print, it’s free under our creative commons attribution license, but please follow some simple guidelines:
  1. You have to credit our authors.
  2. You have to credit ScienceNode.org — where possible include our logo with a link back to the original article.
  3. You can simply run the first few lines of the article and then add: “Read the full article on ScienceNode.org” containing a link back to the original article.
  4. The easiest way to get the article on your site is to embed the code below.