Because of the highly technical and extremely competitive nature of Formula One racing, high performance computing is now commonly used by most F1 programs worldwide. Aerodynamic engineering is a critical component of Formula One design, and CFD simulation on 3D virtual models greatly speeds up research and development by complementing the more expensive and time consuming wind tunnel testing. Besides accelerating aerodynamic design, CFD can also be used to optimize brake cooling, gas tank filling and other engine component cooling activities.
On Tuesday, Renault’s Formula One team announced it had purchased a 38 teraflop Appro Xtreme-X2 supercomputer to upgrade its compute infrastructure. The new Opteron-equipped Xtreme-X2, which is scheduled to be installed at the company’s Formula One CFD Centre in the UK at the end of June 2008, would represent the most powerful supercomputing cluster utilized by a Formula One team. In December of 2006, BMW Sauber F1 team announced it had deployed a 12 teraflop Intel Xeon-based system for its Formula One program; and an 8 teraflop supercomputer was installed by the AT&T Williams Formula One team in July 2007. If more powerful HPC systems are out there in the racing world, no one’s talking about them.
The new Appro machine will be used by the ING-sponsored Renault F1 Team to run full-car simulations for their 2008 F1 R28 and 2009 R29 racing cars. We asked Bob Bell, technical director of the Renault F1 Team, to help fill us in on how the group intends to use the new supercomputer to gain an edge on the competition.
HPCwire: How long has the Renault F1 Team been using high performance computing to help design Formula One cars?
Bell: We have been using CFD since 1995, but only since 2004 have we seriously invested in an HPC facility dedicated to CFD. The high performance computing systems will form part of a dedicated resource that will be used for the aero design development of our F1 cars. In addition they will be available as part of a CFD Centre of Excellence for use within the Renault Nissan alliance.
HPCwire: What will the new Appro system enable you to do that you couldn’t accomplish with your older HPC systems? How important are these systems becoming in Formula One racing?
Bell: The new system will increase by a factor of five our CFD computing capacity, allowing us to model the car’s aerodynamics in much more detail and with results delivered in greatly reduced time frames. These systems are now vital for the successful design of a modern F1 car as they start to rival a wind tunnel (which is the traditional development tool) for productivity.
The transition to our CFD solution using Appro systems will allow us to rival a wind tunnel in terms of development productivity but in a much more cost effective way. Traditionally 90 percent of our aero development has been conducted using empirical physical testing in the wind tunnel. The principal advantage of CFD is that it provides an in depth understanding of vehicle aerodynamics whereas a WT simply provides lined performance measurements. The clusters will primarily boost our CFD capacity from a current 10 percent contribution to our aero development to at least 50 percent. A sizeable step forward to us.
HPCwire: At 38 teraflops, the new system represents the most powerful machine for a Formula One team. What kind of edge do you think this will provide for coming up with winning Formula One designs?
Bell: It will be a significant advantage in terms of the productivity and accuracy of our aerodynamic development program. There is a direct correlation between the number of CFD simulations we can run in any given time period and the on-track performance development of the car.
HPCwire: Will the new system be used by other groups at Renault, outside the Formula One team?
Bell: It is hoped that the new facility will become a center of excellence within the Renault-Nissan alliance for CFD technology. This will directly benefit road car aerodynamic development.
HPCwire: What application software will you be using to drive the CFD simulations?
Bell: We will primarily be using commercial CFD software from CD-Adapco and specialized software solutions developed as part of our partnership with Boeing.
HPCwire: When can we expect to see the first Formula One cars on the track that were designed with the help of the new Appro system?
Bell: It will affect the end of season development with the R28 and strongly influence the entire development of the R29, where completely new aerodynamic regulations will benefit the increase in our computational capacity.
HPCwire: Using the following criteria: initial system cost, operational cost, computing power, ease of deployment and use, system vendor support, upgradeability, reliability, and fault-tolerance in mind — how would you rank them in importance when selecting an HPC system for your Formula One application?
Bell: Computing power – 1; reliability, fault-tolerance – 2; initial system cost – 3; operational cost – 4; upgradeability – 5; system vendor support – 6; and ease of deployment and use – 7.
In summary, it is performance, reliability, cost and operability.
One of the reasons to commit to a CFD facility like this was that it represents a higher rate of return on capital investment for an aerodynamic development tool. Although cost effectiveness was part of our decision-making process, the technical configuration of the cluster was paramount. Appro not only offered us a cost-effective solution, but in addition they improved on our required technical specification through better reliability, greater fault tolerance and redundancy, as well as more flexibility with regards to system scalability.