Siggraph Briefings

JPR Press Luncheon Briefing: Intel - High Fidelity Visualization

High Fidelity Visualization

Digital graphics visualization has come a long way since Siggraph audiences applauded chrome tea pots, but the fact of the matter is that the industry as has a very long way to go. The advantage of the computer is that it can model more than the things we see, it can model the things we think.

For example, Intel is working with Autodesk and Audi to simulate realistic lighting and reflection for cars. Audi says that as a result of the work they're doing, Audi's engineers and designers have access to real-time predictive rendering enabling them to reduce their dependence on physical prototypes. Ironically, this was one of the very early uses of high end visualization. After all these years, it's still not good enough for car designers who want to have a realistic idea about what a car is going to look like on the road before its even built. 

Intel's Director of Workstation Strategic Marketing, Wes Shimanek talked with Paul Navrátil of the Texas Advanced Computing Center (TAAC), about the challenge of high fidelity visualization. Navrátil says, "at TACC, we want to give users the power to explore massive simulation data with the same ease as conducting a bench experiment." Navrátil thinks of high fidelity visualization, as "visualization without limits" because by bringing the full capabilities of modern processors to the task, it's possible to simulate "real" lighting and material properties on problems too large or too small to test in lab. As an example, says Navrátil, suppose you want to assess the efficiencies of an aircraft carrier hull. If it were possible, you could place the hull in a rheoscopic fluid and observe the flow of the current as the ship moves through the liquid. "That is just not something you'd do," says Navrátil, "but it's something we can simulate."

Dr. Paul Navrátil, Research Associate
Texas Advanced Computing Center

Take it another step: what if you could visualize complex data using photorealistic rendering techniques. High fidelity visualization will allow you to combine simulation and visualization, where the act of "seeing" the data combines with the simulation to produce deeper insight. "Current rendering techniques produce images in a fast but ad hoc way. Instead, if we render by simulating how light actually travels, we can efficiently provide both stunning visualization and more interesting representations."   

Navrátil makes the comparison to old Hollywood westerns and describes the western main street, we're all familiar with. "Those towns were nothing but facades. This is the way video games work today, and the optimized rasterization pipeline was built with video games in mind. Visualizations done this way might be fast, but they cannot easily simulate accurate light or material properties."

Navrátil thinks there's a lot more that can be done with ray tracing when it's  used for visualizing huge amounts of data. Ray tracing is an idea as old as computer graphics itself, but new advances in algorithm design and hardware capability make it attractive for today's largest and most complex visualization challenges, says Navrátil. But, dependence on rasterization is quickly becoming visualization's limiting factor.
"There has to be a better way," says Navrátil. He believes ray tracing will scale more efficiently when it comes to visualizing huge data sets, particularly where complex lighting and materials are used.

Scaling Processors
Intel, thinks they do have a better way. The work Intel is doing with the Xeon processor and Xeon Phi array of processors puts the power of many processors to work on a problem like ray tracing.
Democratization, is it a good thing? JPR discusses big problems at the annual Siggraph Press Luncheon
The annual Siggraph luncheon hosted by Jon Peddie Research for press and analysts during the Siggraph conference provides a forum to discuss ideas like these. This year, the theme of of the luncheon discussion is: Democratization, which we're seeing play out in a number of different ways. Democratization is great for small companies as powerful computers and software come down in price, but democratization also brings in more competition and can cause price wars. At this year's conference, the luncheon discussion promises to be wide ranging and passionate.

Intel and the TACC are dealing with democratization on a very large scale, they're tackling the very largest problems in the world with off the shelf technology. And yes, they see democratization as a very good thing.

Press and analysts may register at: JPR SIGGRAPH Luncheon