Hi, my name is David Scott with the high performance computing team of Intel Corporation. Today I would like to talk to you about HPC Applications.

To create an HPC application requires implementing 5 separate stages. You start with the Physical System. So for example, looking at air flow over an aeroplane. What you are trying to compute is how the air flows over the wings and fuse lodge, which would allow you to compute important quantities like lift, keeping the aeroplane up, and drag, how much fuel it takes.

From the Physical System, you go to a Mathematical Model. For the case of airflow over a wing, that's the Navier-Stoke's equation.

From the Mathematical Model, you go to a Numeric Model, something that can actually be solved by the computer. This typically involves the discretization of the domain with some solution techniques such as finite differences.

From there, you have to go to an algorithm, that is how you are actually going to solve the problem that you have created. For the airflow example over a wing, this is often done with a multigrid solver.

Once you have an algorithm, the last step is actually creating the application. That is, taking the algorithm and turning it into FORTRAN or C Code.

And when you're done, you have an application that runs on real computers and solves real problems of scientific and engineering importance.

With oil over $100 a barrel today, one of the economically most important HPC applications is seismic processing, which is trying to see what is under the ground.

The ground has in fact various layers under there, which may be shale or salt or sand or whatever. And the oil companies go out, and they set off explosions (boom!), which send sound waves down into the ground, which reflect off these layers but also defract and they would multiply reflect and so on and they eventually come back up to the surface and they have a whole bunch of microphones on the surface that record the acoustic traces.

Those traces get taken back to the computational lab and very expensive algorithms is run on it to try to recreate these layers very accurately. And the idea is to find places like these where oil which normally is light and would escape is trapped underneath rock.

And if you find the right shape of rock, that's a good place to drill a well and get the oil out and make lots of money.

There are many more application areas I can talk about today. But here are 5 of them.

Crash Analysis is using the computer to test the safety of cars by simulating them running into each other.

BioChemistry calculations are looking for new drugs, taking potential drugs and looking at the interactions between them and the proteins in your cells to verify A the drug works, and B that it is safe.

Animation is creating full length animated movies entirely inside the computer.

Chemistry calculations are looking at improving or creating new catalysts for industrial-sized chemical processes.

Formula 1 design is using the same calculations as the aeroplane that is fluid dynamics over the body of a car, but instead of trying to keep the airplane up, you are trying to keep the car down on the ground.

The HPC market is actually growing in 2-Dimension. It is growing wider as more companies use HPC for the first time. And it is also growing higher as traditional HPC users use more cycles with larger machines. In fact, the HPC market is growing much faster than traditional server market. HPC is already 25% of all servers in the world and may reach 30 or 35% in the near future.

I'm David Scott for Intel Corporation.