How a 1% Improvement in Fuel Efficiency Can Save $3 Billion a Year

by George Millington

Whether you’re building airplanes or microprocessors, computer simulation has become an indispensable tool. Products can be modeled operating in broiling heat or bone-chilling cold, in gale-force winds or lashing rain, without the need to step away from a server or computing cluster. Design cycles are shortened, money is saved and better products reach the market.

Now ANSYS, developer of ANSYS Fluent, the world’s top computational fluid dynamics (CFD) software, is taking things a step further by adding support for GPU acceleration to its latest production release. This puts the power to build sophisticated GPU-accelerated simulations – fast – into the toolbox of every one of its customers.

Go with the flow: Aerodynamic behavior of a passenger car from a CFD simulation
Go with the flow: Aerodynamic behavior of a passenger car from a CFD simulation

Twice as Fast, Half the Cost

ANSYS Fluent is used by thousands of businesses, including the world’s top automakers, industrial equipment manufacturers and aerospace companies.

It helps engineers design more aerodynamic cars and planes, improve the thermal management and reliability of integrated circuits, and enhance the design of a broad range of other products.

The latest version, Fluent 15, now supports GPU acceleration with its CFD flow solver. Multiple GPUs, in fact. This allows auto and aerospace engineers to run aerodynamic simulations twice as fast, helping them build more fuel-efficient cars and planes by reducing air drag.

GPU-accelerated simulation of truck body aerodynamics with ANSYS Fluent completed in half the time, compared to a CPU-only solution.
GPU-accelerated simulation of truck body aerodynamics with ANSYS Fluent completed in half the time, compared to a CPU-only solution.

More aerodynamic cars means extra cash for consumers. With U.S. drivers burning more than 74 billion gallons of gas a year – or $296 billion at $4 per gallon – improving fuel efficiency by just one percent adds up to almost $3 billion a year.

That’s some serious savings.

GPU-Acceleration Library Now Available to Other Developers

Once ANSYS decided to add GPU acceleration to Fluent 15, they realized they would benefit from NVIDIA’s expertise in GPU technology.

So NVIDIA developed a programming library that accelerates the core numerical CFD routines, known as algebraic multigrid (AMG). This made it much easier for ANSYS to add GPU support to Fluent. There was no need for their software team to become experts in GPUs.

NVIDIA is now making this library – AmgX, which stands for AMG aXelerator –available to other developers to accelerate similar calculations.

With AmgX, it takes only a few days to add GPU support to an application – far less time than if developers were to do it themselves from scratch. This means developers can cut months, even years, off the time it takes to bring a GPU-accelerated version of their product to market.

And, AmgX is not just for CFD. It can add GPU acceleration to a wide range of applications used for energy exploration and research, computational physics and nuclear research.

A beta version of AmgX is now available. To learn more, visit http://developer.nvidia.com/AmgX or send an email to AmgX@nvidia.com.

You can also learn more about AmgX and Fluent at the Supercomputing 2013 (SC13) conference in Denver (Nov. 18-21). NVIDIA’s Joe Eaton will be discussing AmgX in the GPU Technology Theater, booth #613, on Thursday, Nov. 21 from 12:30 to 1pm MT.

Not going to SC13 this year? No problem. All of our booth presentations will be streamed live on our SC13 web page.